WO2008151570A1

WO2008151570A1 - Method, device and system for coding and decoding

Info

Publication number: WO2008151570A1
Application number: PCT/CN2008/071255
Authority: WO
Inventors: Xiaozhen Zheng; Yannan Wu; Jianhua Zheng; Yun He
Original assignee: Huawei Technologies Co., Ltd.; Tsinghua University
Priority date: 2007-06-13
Filing date: 2008-06-10
Publication date: 2008-12-18

Abstract

A coding method comprises the steps of: receiving the data to be transformed; implementing a first transform to the received data and obtaining a first transformed data; implementing a second transform to the received data and obtaining a second transformed data; determining an adjusting parameter based on the first transformed data and the second transformed data; adjusting the first transformed data according to the adjusting parameter and the parameter of the second transform. A method for decoding and a device, system for coding and decoding are also provided. The value range of the transformed data is adjusted so that the value range ofimage data transformed by different transforms can keep consistent and the influence caused by the transforms can be reflected truly when the adaptive block transform technique is adopted. The transform with preferred effect can be selected to improve the coding efficiency.

Description

Coding and decoding method and device and system

The present invention relates to video compression coding techniques, and in particular, to a coding and decoding method, apparatus and system. Background technique

In order to reduce the amount of data of video data as it is transmitted or stored, it is generally necessary to compress and encode the video data. In the field of video compression coding, transform is an important technology, which is to transform an image, image content and information in an area into a specific area, so that the video compression algorithm can more effectively perform this part of the content. Compression. Then, after the transform is performed, the data is quantized, entropy encoded, etc., and then the compressed and encoded video data is formed.

In video coding and decoding standards, such as Moving Pictures Experts Group (MPEG) MPEG-2, H.264, and Digital Audio Video Codec Standard (AVS) use transform technology. In these standards, an area of an image or image is divided into small blocks or sub-areas called sub-blocks, and the transformation is performed in units of sub-blocks. In general, the sub-block size can be 4x4 or 8x8, where 4 and 8 are in image pixels.

A video file is composed of multiple video images, and an image usually contains a lot of content, and different parts of the image have different characteristics. Therefore, if all the images in a video or an image is divided into sub-blocks of the same size (such as 8x8 size) and then transformed, the effect may not be optimal, that is, all the sub-blocks cannot be effectively The contents of the block are effectively transformed into a specific area after being transformed. Based on this, an adaptive block transform technique is proposed. The principle is: divide a specific region according to different sub-block sizes, and then perform different transforms for sub-blocks of different sizes (for example, dividing the image into 4x4 and 8x8 blocks, 4x4 transforms for 4x4 blocks, 8x8 transforms for 8x8 blocks), and then according to certain criteria, which transforms can more effectively focus block information on specific regions in different transform situations. Finally, the better transformation results are stored. When the decoding end decodes the image transformed by the above manner, the decoding end acquires the information of the transform scale (such as 4x4 or 8x8) according to the corresponding information in the code stream, and then uses the corresponding inverse transform (such as 4x4 inverse transform or 8x8 inverse transform). The area is processed to obtain the original video data.

However, in the compression coding process of video, it may be necessary to fuse different changes for various purposes. For example, the 4x4 transform matrix is based on Discrete Cosine Transform (DCT), and the 8x8 transform is based on wavelet. The two sets of transform matrices probably do not have too many identical transform features, and the same data passes. The degree of change in the range of values of these transformed data is inconsistent. Since the quantization may result in the damage of the data information, whether the same or different quantization tables are used for quantization in the quantization process, since the same data is inconsistently changed after the different values are converted, the data after the transformation is quantized. The degree of loss is inconsistent. In this case, it is impossible to determine a better conversion method by using certain judgment criteria, so that the data coding efficiency cannot be effectively improved. Summary of the invention

In view of this, the embodiments of the present invention provide a coding and decoding method, apparatus, and system, so that the numerical ranges of data after different transformations are substantially consistent.

In order to achieve the above object, the following technical solutions are used in the embodiments of the present invention:

An encoding method, including:

Receiving data to be transformed;

Performing a first transformation on the data to be transformed to obtain first transformed data;

Performing a second transformation on the data to be transformed to obtain second transformed data;

And determining an adjustment parameter according to the first transformed data and the second transformed data, and adjusting the first transformed data according to the adjustment parameter and the second transformed parameter;

The adjustment parameters are written into the encoded code stream.

A decoding method, including:

Receiving a code stream, decoding the code stream, obtaining first transformed data and adjusting parameters; and adjusting the first transformed data according to the adjustment parameter and the second transformed parameter.

An encoding device comprising:

a data receiving unit, configured to receive data to be transformed;

a transform unit, configured to perform first transform on the data to be transformed received by the receiving unit, to obtain first transformed data, and perform second transform on the data to be transformed to obtain second transformed data;

The first adjusting unit determines an adjustment parameter according to the first transformed data and the second transformed data, and adjusts the first transformed data according to the adjustment parameter and the second transformed parameter.

A decoding device comprising: a code stream receiving unit, configured to receive a code stream;

a decoding unit, configured to decode the code stream to obtain first transformed data and adjustment parameters, and a second adjusting unit, configured to adjust the first transformed data according to the adjustment parameter and the second transformed parameter.

A codec system, comprising: an encoding device and a decoding device;

The encoding device includes:

a data receiving unit, configured to receive data to be transformed;

The first adjusting unit determines an adjustment parameter according to the second transformed parameter, the first transformed data, and the second transformed data, and adjusts the first transformed data according to the adjustment parameter and the second transformed parameter.

The decoding device includes:

a code stream receiving unit, configured to receive a code stream;

An encoding method, including:

Receiving data to be transformed and encoded parameter information;

Determining an adjustment parameter according to the encoded parameter information and the second transformed parameter, and adjusting the first transformed data according to the adjustment parameter and the second transformed parameter;

The adjustment parameters are written into the encoded code stream.

An encoding device comprising:

a third receiving unit, configured to receive data to be transformed and encoded parameter information;

a third transforming unit, configured to perform first transform on the data to be transformed received by the third receiving unit, to obtain first transformed data;

a third adjusting unit, configured to determine an adjustment parameter according to the encoded parameter information received by the third receiving unit and the second transformed parameter, and adjust the third according to the adjustment parameter and the second transformed parameter The first transformed data obtained by the transform unit;

And a third writing unit, configured to write the adjustment parameter obtained by the third adjusting unit into the encoded code stream. It can be seen that, in the embodiment of the present invention, the adjustment parameter is first determined according to the first transformed data and the second transformed data, and the first transformation is adjusted according to the adjustment parameter and the second transformed parameter. The data is such that the first transformed data is consistent with the numerical range of the second transformed data, so that when the adaptive block transform technique is applied, the influence of the transform on the data can be truly reflected to select a better transform, and then Improve coding efficiency. The codec method and device provided by the present invention can ensure that the first transformed data range and the second transformed data range are consistent, and the adjustment parameters are written into the code stream, so that the decoding end can be adjusted according to the adjustment parameter. The received data is processed accordingly. DRAWINGS

1 is a general flowchart of a method for processing transformed data according to an embodiment of the present invention;

2 is a schematic structural diagram of a transform data processing apparatus according to an embodiment of the present invention;

3 is a specific flowchart of a method for processing transformed data according to Embodiment 1 of the present invention;

4 is a detailed structural diagram of a transform data processing apparatus according to Embodiment 1 of the present invention;

FIG. 5 is a specific flowchart of a method for processing transformed data according to Embodiment 2 of the present invention; FIG.

6 is a specific flowchart of a method for processing transformed data according to Embodiment 3 of the present invention;

7 is a detailed structural diagram of a transform data processing apparatus according to Embodiment 3 of the present invention;

8 is a specific flowchart of a method for processing transformed data according to Embodiment 4 of the present invention;

9 is a detailed structural diagram of a transform data processing apparatus according to Embodiment 4 of the present invention;

10 is a schematic flowchart of an encoding method according to Embodiment 5 of the present invention;

FIG. 11 is a schematic flowchart of a decoding method according to Embodiment 6 of the present invention;

12 is a schematic structural diagram of an encoding apparatus according to Embodiment 7 of the present invention;

13 is a schematic structural diagram of a decoding apparatus according to Embodiment 8 of the present invention;

14 is a schematic flowchart of an encoding method provided in Embodiment 9 of the present invention;

FIG. 15 is a schematic structural diagram of an encoding apparatus according to Embodiment 10 of the present invention. Detailed ways

In order to make the objects, technical means and advantages of the embodiments of the present invention more clear, the following The embodiments of the present invention are further described in detail.

In the embodiment of the present invention, firstly, the data is subjected to two different transformed data value ranges, and then the transformed data is compensated according to the different transformed and quantized data value ranges to ensure the same. The range of values of the data to be encoded after the transformation and quantization is substantially consistent with the range of values of the data after another transformation and quantization to improve coding efficiency.

FIG. 1 is a general flowchart of a method for processing transformed data according to an embodiment of the present invention. As shown in Figure 1, the method includes:

Step 101: Calculate the range of values of the image data after the two transformations according to the transformation matrix required by the preset two transformations.

Step 102: Calculate a difference value range characteristic value of the two transforms according to the two transformed numerical ranges.

According to different compensation methods, the values of the difference values of the above two types of transformations are also different, and the difference between the two types of image data and the corresponding quantized value range may be obtained, or the image data may be subjected to two types respectively. The difference in the range of values after the transformation. When the feature difference is a difference between the two types of transforms and the corresponding quantized value range, when calculating the feature difference, the calculation includes the quantization step size of each of the two transforms found according to the quantization point corresponding to the image data. The two transformed numerical range feature difference values.

Step 103: Apply one of the two transformations to the data to be transformed, and compensate the transformed data according to the difference of the calculated numerical range.

In the embodiment of the present invention and the following embodiments, the transform data processing method may specifically include an encoding method, a decoding method, and the like.

FIG. 2 is a general structural diagram of a transform data processing apparatus according to an embodiment of the present invention. As shown in FIG. 2, the apparatus includes a first value range estimating unit, a second value range estimating unit, and a numerical range difference unit. And transform compensation unit.

The first value range estimating unit is configured to calculate a range of values of the image data after the first transformation according to a transformation matrix required by the first one of the preset two transformations, and provide the value range difference unit .

And a second value range estimating unit, configured to calculate a range of values of the image data after the second transformation according to a transformation matrix required by the second transformation in the preset two transformations, and provide the value range difference unit.

The numerical range difference unit is configured to calculate the numerical range characteristic difference values of the two transformed values according to the numerical ranges after the first and second transformations, respectively, and provide the same to the transform compensation unit.

And a transform compensation unit, configured to apply a first transform to the data to be transformed, and compensate the first transformed data according to the calculated numerical range feature difference value.

It should be noted that, in all embodiments of the present invention and the following embodiments, the transform data processing apparatus may be specifically an encoding apparatus or a decoding apparatus.

In the above method and apparatus, the manner of compensating the transformed data may be: determining an appropriate adjustment factor according to the relationship between the numerical ranges of the two transformed data, and multiplying the data obtained by one of the transformations by the adjustment factor. Then, the quantization is performed according to the corresponding quantization points, so that the numerical range of the data obtained after the transformation and quantization is consistent with the numerical range of the data obtained by another transformation and quantization.

Alternatively, the method for compensating the transformed data may be: adjusting the quantization points corresponding to one of the transformations according to the relationship between the numerical ranges of the two transformed data, and then adjusting the data according to the adjusted quantization points. Quantization is performed so that the numerical range of the data obtained by the transformation and quantization can be made to coincide with the numerical range of the data obtained by another transformation and quantization.

Alternatively, the manner of compensating the transformed data may be: according to the relationship between the numerical ranges of the two transformed data, re-establishing the quantization table for one of the transformations, and then changing the quantization table according to the reconstructed quantization table. The data obtained is quantized, so that the numerical range of the data obtained by the transformation and the quantization can be made to be consistent with the numerical range of the data obtained by another transformation and quantization.

The specific embodiments of the present invention in the above three different processing modes will be described below through different embodiments. The first embodiment is directed to the first processing mode, the third embodiment is directed to the second processing mode, and the fourth embodiment is directed to the third processing mode.

Embodiment 1:

In the present embodiment, it is assumed that there is an encoded data block C, and a transform A and a transform B. The scale of transformation A is nxn, and the scale of transformation B is mxm. The quantization point is used when encoding the data block C, and the quantization step sizes of the transform A and the transform B found at the quantization point are QTAB1 [s] and QTAB2[s], respectively. Here, when the data is quantized, each different coding region (for example, the data block C) corresponds to one quantization point, and the corresponding quantization step size is found according to the quantization point, and the quantization process is performed. The range of values of the transformed A and the corresponding quantized data is adjusted to be consistent with the range of values of the transformed B and the corresponding quantized data. In this embodiment, the difference value range of the two transformed values is the difference between the two types of image data and the corresponding quantized value range.

FIG. 3 is a specific flowchart of a method for processing transformed data according to Embodiment 1 of the present invention. As shown in Figure 3, the method includes:

Step 301: Calculate the range of values of the image data after the two transformations according to the transformation matrix required by the preset two transformations.

In this step, the coding block C is divided into sub-blocks according to the scale of the transform and B, and the transforms A and B are respectively applied to the respective sub-blocks. Calculating the transformed range of values of each pixel data in the divided sub-blocks according to the transformation and the transformation matrix corresponding to B. For any type of transformation, the method of calculating the converted value range is the same. The following is an example of calculating the numerical range of each pixel data in the sub-block after the transformation A, and the specific calculation method is explained. Assuming that the coding block C size is MxN, T is the transformation matrix corresponding to transform A, and T' is the transposed matrix of T, then applying transform A to coding block C includes: calculating [TOT']®S, where S is a scaling factor Matrix, which is used to normalize the data after [ΤΟΤ'], multiply the symbols by the matrix, and @ to represent the respective elements of the matrix. According to the matrix multiplication, the numerical range increment of [Τ· Τ,] can be calculated.

First calculate the increment of the range of [Τ· Τ,] relative to C: Τ, the sum of the absolute values of the coefficient in the middle column is dp, then the value range of the data obtained by multiplying the column with the ith row in C, the maximum becomes The multiple of the value range of the data in the zth row and column column in C. Since both the computer and the hardware processor store information in binary units, the numerical range is represented in binary form herein. Here, the maximum increment of the numerical range of the data after multiplying the column by the zth row in C is expressed as log ₂ ( ). As above, you can calculate the maximum value range increment for each element in C after [Τ· Τ,].

Then calculate the increment of the value range of [TOT']®S relative to [Τ· Τ']: If the absolute value of the element at (k, h) in S is S( ), then after [TOT']®S, The value range increment for point (k,h) is log ₂ (S( _{t 3⁄4} )) relative to [ΤΟΤ'].

Adding the numerical range increments calculated in the above two steps, you can calculate the maximum value range increment for each point in C after [TOT,]®S. According to the numerical range of the coding block C itself, the maximum value range of the values of each point in each sub-block can be obtained after the transformation.

Calculated in the above manner is the maximum value range after the coding block C is transformed, and in the actual coding, a common coding block C, because the data in the coding block C obeys a certain mathematical distribution, it is difficult to reach its maximum The range of values, so the converted range of values is likely to fall short of the maximum value range. Therefore, preferably, the mathematical distribution of the coded block C is further considered when performing the numerical range calculation.

Specifically, it is assumed that the value in the block C has a certain mathematical distribution, and the transformation A is applied to the distribution, and then the above-mentioned method can be used to calculate the region where the numerical range of the C-transformed data is most likely to occur. example For example, suppose the mathematical distribution model of the data of the M x N size data block C is P(x, y), and the maximum possible numerical range of each point of the i-th row of the data block is expressed as

Where /(·) is the mapping relationship obtained according to the mathematical distribution model, and its meaning is the maximum possible occurrence of the point (x, y), (0<=x<=M, 0<=y<=N) The range of values, that is, the most likely value of the point (x, y) at the time of encoding is 2 ^{/ (p(} w» , 2 ^{/ (p(} w» is set to ^. Also, assume that transform T' j-th column The coefficient value of the point is T'G,y), ( 0<=y<=N ), then the maximum increment of the data range of the ith row of C is multiplied by the i-th row of C. Log ₂ (|∑T j'x2 ^v ' |) , from which the maximum possible range of values for each point of the data block C after the transformation [Τ· Τ,] ® S can be calculated.

After the calculation of the numerical range in the above manner, the sub-block in the coding block C can be transformed.

The range of values of the data after each point of A:

All A12 Α1(η-1) Α1η

A21 A22 A2(n-l) A2n

Anl An2 An(n-l) Ann

The value range of the data of each point after the sub-block in the coding block C is transformed is:

Bll B12 B13 Bl(m-2) Bl(m-l) Blmm

B21 B22 B23 B2(m-2) B2(m-1) B2mm

B31 B32 B33 B3(m-2) B3(m-1) B3mm Bml Bm2 Bm3 Bm(m-2) Bm(ml) Bmm

Step 302: Calculate the transformed average value range for each of the two transformations.

In this step, the calculation method for the average value range of the two transformations is the same. Still taking the transformation A as an example, the manner of calculating the transformed average value range may be:

The numerical range of the pixel data of the first row and the row in the sub-block after the transformation A obtained in 301. Similarly, the average value range after the transformation is calculated as:

The range of values of the pixel data of the first row and the jth column in the sub-block.

Step 303: Calculate the difference between the two types of transforms and the corresponding quantized value range according to the transformed average value range obtained in step 302.

In this step, the range of values after two transformations and corresponding quantization is first calculated. The transformation A and the corresponding quantization are still taken as an example. As described above, the quantization block is used for the coding block C corresponding to the quantization point, the quantization step size of the transformation A is QTAB1[s], and the A r calculated in step 202 is used to represent the data block obtained after the transformation A. The average numerical range, the numerical range is expressed in binary form, and the average value of the encoded data after transformation A is 2 ^Av .

The quantization process of data usually in video codec can be expressed as:

( XxQTAB[s] ) »shift[s] ( 1 ),

Where X is the value to be quantized, s is the quantization point, QTAB[n] is the quantization step size found in the quantization table according to the quantization point n, and shift[s] is the shift corresponding to the quantization point s. According to the above steps, the average value obtained by quantization using the quantization step QTAB1[s] is (QTABl[s].2 ^Av ) »shift[s] , and the numerical range of the data is expressed in binary form (A ^ + log^TMSlW ^hif^s Similarly, by After over-transform B and quantification with QTAB2[s] as the quantization step size, the value range of the obtained data is ( Avr _B + log ₂ QTAB2[s] )»shift[s]. Next, the two transformations and the corresponding quantized numerical range difference are calculated. The specific calculation method is: Calculate the multiple relationship between the two transformations and the corresponding quantized values, ie (QTAB2[ _S ]x 2 ^A » _S hift[ _S ] ^

(QTAB\[s]x 2 ^AvrA ) » shift[s] The difference between the range of values expressed in binary form is: τν , QTAB2[s]x2 ^Av3⁄4 »shift[s] _{τΑ Ώ ν}

D = log, ―—— τ ― ,

TAB2[s,] , )-

⁵² QTABl[s]x2 ^A3⁄4 »shift[s] ^L ”

(A ^ + log^r^SlW Step 304, determining the adjustment factor of the transform A according to the difference in step 303.

In this step, according to the numerical range difference obtained in step 303, the adjustment factor of the transformed A is determined to be q = 2 ^D '.

The data required in the above calculation process is a priori data, that is, data that can be obtained before encoding, so the above steps can be completed before encoding and the results are saved without calculation in the encoding process.

In the actual coding, the adjustment factor of the transform A may be adjusted due to the difference of the encoded data, so the adjustment factor of the transform A at the time of encoding can be appropriately adjusted according to the characteristics of the encoded data. The encoded data characteristic may be the actual average range of values after the data block has been transformed A. The adjustment factor of transform A is written into the encoded code stream. The adjustment factor of transform A can be written in the sequence header or image header or slice header or macroblock header of the code stream. The adjustment factor of the transformation A is used in the same manner as the above method, and will not be described here.

Step 305, applying a transform A to the coding block C, obtaining the transformed data, and multiplying the obtained data by the adjustment factor of the transform A.

In this step, the transform A is applied to the coding block C, and the specific process is the same as the existing one, and will not be described here. Multiply the transformed data by the adjustment factor in step 204, that is, after the transformation The obtained data are: C, ' = C, .2 ^D ', where C is the pixel data value of the i-th row and the j-th column after the transformation, and is the reconstructed value of the pixel data of the i-th row and the j-th column after the adjustment. .

Step 306, quantizing each by using the quantization step size found by the transform A.

Each of the adjusted ones after the transformation A is quantized in accordance with the existing method. Since each C is reconstructed by using the adjustment factor in step 205 to obtain C, ', and the adjustment factor is determined according to the difference between the two types of transformations and the corresponding quantized value ranges, therefore, this step is performed by quantizing each The numerical range of the data can be consistent with the numerical range of the data obtained by the transforming block B and the corresponding quantized data.

The specific derivation process is as follows: Since the shift in the quantization process has the same effect on the transformation A and the transformation B, the influence of the displacement is not considered below.

As described above, the average value range after the coding block C is applied to transform A is A r , and the value obtained by transforming the encoded data A is , then the result of multiplying the adjustment factor is

The value range expressed in binary is A _V r _B + log ₂ QTAB2[s] - log ₂ QTAB\[s]; The result obtained by quantizing the quantization step QTAB1 [S] is 2 ^{Α +1} °^ TM ^β2 Μ_^βTM ] · ρ?3⁄4 1 = 2 ^{Av +1} ° ^g f ^w , the range of values in binary is Αν + log ₂ QTAB2[s]. The average value range after the coding block C is applied to the transform B is Avr _B , and the value obtained by transforming the encoded data is A, and the result obtained by quantizing the quantization step QTAB2[s] is 2 ^Αν · β73⁄452 , in binary. The range of values represented is Αν + log ₂ QTAB2[s]. It can be seen that the same coding block C has the same value range as the data obtained after the transformation B and the corresponding quantization after the operations of steps 205 and 206. The purpose of making the final numerical range of the transformation A and the transformation B consistent is achieved.

So far, the flow of the method for processing the transformed data provided by the embodiment ends.

Performing transform data processing on the encoding side according to the above method, thereby selecting transform A and transform B Excellent transformation method. If the finally selected transform mode is transform A, the data sent to the decoder is data processed by transform A and corresponding quantization and entropy coding. At this time, after the data is inverse quantized according to the quantization point corresponding to the transform A at the decoding end, the inverse quantized result is divided by the adjustment factor of the transform A, and then inversely transformed to reconstruct the encoded end data. The adjustment factor of the transform A can be obtained at the decoding end according to the procedures of the foregoing steps 301-304. If the decoding end avoids the use of division, the decoded data can be multiplied and shifted to achieve the purpose of dividing the decoded data by the adjustment factor.

If the coded stream already contains the adjustment factor of the transform A at the time of encoding, the decoding end obtains the adjustment factor of the transform A from the place where the adjustment factor is included in the code stream, such as the sequence header, the picture header, the slice header, and the macroblock header. The adjustment factor of the transformation A is used in the same manner as the above method, and will not be described again here.

This embodiment further provides a specific implementation of the transform data processing apparatus, which can be used to implement the method shown in FIG. Fig. 4 is a view showing a specific configuration of the down conversion data processing apparatus of the embodiment. As shown in Fig. 4, the apparatus includes a first value range estimating unit, a second value range estimating unit, a numerical range difference unit, and a transform compensating unit. The transform compensation unit includes an adjustment factor determining sub-unit, a transform sub-unit, a post-transform processing sub-unit, and a quantized sub-unit.

The first value range estimating unit is configured to calculate a range of values of the image data after the first transformation according to a transformation matrix required by the first one of the preset two transformations, and provide the value range difference unit.

a value range difference unit, configured to calculate, according to the first and second transformed value ranges and the first and second transforms respectively corresponding to the second quantization point, the image data is subjected to the first and second transformations and the corresponding quantized values respectively. The difference in range is provided to the transform compensation unit.

In the transform compensation unit, the adjustment factor determining subunit is used to calculate a difference value according to the numerical range The difference in the range of values provided by the element determines the adjustment factor and is provided to the post-transformation subunit. A transform subunit that applies a first transform to the data to be transformed and provides the transform result to the post-transform processing sub-unit. The post-transform processing sub-unit is configured to multiply the first transformed data by a determined adjustment factor, and provide the result to the quantizing sub-unit. The quantizing sub-unit quantizes the result obtained by multiplying the adjustment factor by the quantization point corresponding to the image data.

If the encoding end writes the adjustment factor into the code stream, the corresponding adjustment factor writing unit needs to be added to write the adjustment factor into the code stream.

In the present embodiment, the data after the transformation A is first multiplied by an adjustment factor, and then quantized. As can be seen from the quantization process represented by the formula (1), in the quantization process, it is first multiplied by the quantization step size, and then the right multiplication is performed on the multiplication result, that is, the low-order information is discarded. Then, after the data after the transformation A is multiplied by the adjustment factor and the result is multiplied by the quantization step QTABl [s], the lower bit may not be 0, but some information is still concentrated, and when the shift is performed, This will result in a loss of data accuracy. Based on this, in order to reduce the loss of data precision in the quantization process, an evolution processing method is also proposed based on the present embodiment, and the value of the data after the transformed A is changed by multiplying the adjustment factor of the transform A and increasing the shifting manner. range.

Specifically, corresponding to the flow shown in FIG. 3, the operations of steps 301 to 303 are unchanged. In step 304, the adjustment factor of the determined transformation A becomes q' = 2 ^ΰ ' χ 2", which is called an adjustment coefficient. Wherein, the value of "the value is determined according to the hardware storage range of the encoding end. After the data after the guaranteed transformation is multiplied by the adjustment factor q', the data of the subsequent operation does not exceed the hardware storage range, and the value of "the value" The bigger the better, this will be able to focus the information in the data after the transformation A as high as possible.

Next, in step 305, the transformed data is multiplied by the above adjustment factor q'.

Finally, when the data is quantized in step 306, the bit number corresponding to the quantization point is further increased by "bit", that is, after the quantization operation is performed, the quantization result is shifted to the right by n bits. In this way, it will be able to In step 304 after the surface modification, the 2" times of the numerical range is expanded due to the change of the adjustment factor. Briefly, the operations of the above modified steps 305 ~ 306 are: ( 2 ^D x2" xXxQTAB[s] )»shift[s]+n, where X represents the transformed data. Intuitively, the operation is the same as the operation in the original steps 305 ~ 306 (2 ^ΰ ' xXxQTAB[s])»shift[s], but when implemented on a hardware platform, the accuracy loss of the former is better than The latter is small. This is because, in the modified step 305, after multiplying the adjustment factor q', the data X is expanded by 2 ^ΰ ' χ 2" times, and the useful information is enlarged relative to the data after the original step 305 operation, More emphasis is placed on the upper bits, so that during the shifting process involved in the quantization step 306, the useful information lost by the shift is relatively reduced. Thus, the data accuracy can be more effectively improved.

When the encoding end uses the evolution mode of the first embodiment, when the decoding end performs inverse transform and inverse quantization on the data processed by the transform, the corresponding inverse operation is also performed. Specifically, in inverse quantization, the number of shift bits is reduced by n, and the inverse quantization result is divided by the adjustment factor q' of the transform A before inverse quantization, and then inversely transformed to more accurately reconstruct Encoding data.

Of course, the present invention also provides a corresponding evolved device structure corresponding to the above evolution mode, and the structure is similar to the structure shown in FIG. 4, except that the transform compensation unit includes an adjustment coefficient and a shift offset quantum unit, a transform subunit, and a transform. Post-processing sub-unit and quantization sub-unit. The structure and function of the first numerical range estimating unit, the second numerical range estimating unit and the numerical range difference unit are the same as those shown in Fig. 4.

In the transform compensation subunit of the evolved device, the adjustment coefficient and the shift offset quantum unit are configured to determine the adjustment coefficient and the shift offset amount during quantization according to the calculated numerical range feature difference value, and respectively provide the transformed offset amount Processing subunits and quantizing subunits. A transform subunit that applies a first transform to the data to be transformed and provides the transform result to the post-transform processing sub-unit. The post-transform processing sub-unit is configured to multiply the first transformed data by the determined adjustment coefficient, and provide the result to the quantizing sub-unit. The quantizing sub-unit quantizes the result multiplied by the adjustment factor according to the shift offset and the quantization step size corresponding to the first type of transform.

In the above embodiment, when calculating the difference between the two kinds of transforms and the corresponding quantized value ranges, the averaged range of values after the two transforms is used. In order to make the numerical range calculation more accurate, it can also be carried out in the manner of the second embodiment.

Embodiment 2:

In this embodiment, similar to the first embodiment, the difference value range of the two transformed values is the difference between the two types of image data and the corresponding quantized value range, but the specific calculation method of the difference is compared with the first embodiment. Different in the middle.

FIG. 5 is a specific flowchart of a method for processing transformed data provided in the embodiment. As shown in Figure 5, the method includes:

Step 501: Calculate the range of values of the image data after the two transformations according to the transformation matrix required by the preset two transformations.

The calculation method of this step is the same as that of step 301 in the first embodiment, and details are not described herein again.

In step 502, the average value range after the transformation B is calculated.

The manner of calculating the average value range after the transformation B in this step is the same as that in the first embodiment, and includes Avr _B = f Bij , where is the numerical range of the pixel data of the z-th row and the column in the sub-block after the transformation B.

Step 503: Calculate the two types of transforms and the corresponding quantized value range difference values according to the value range of each data after the transformed A obtained in step 501 and the average value range after the transformed B obtained in step 502.

In this step, the method principle when calculating the difference of the numerical range is the same as that in the first embodiment, and the difference is The difference of the data range in this embodiment is calculated according to the numerical range of each data after the transformation A and the average value range after the transformation B. Specifically, after the coding block C is divided into sub-blocks according to the scale of the transform A, the pixel data of the i-th row and the j-th column in the sub-block is transformed into A, and the data range difference corresponding to the transformed B is: d' _{( )} =( Avr _B + log ₂ QTAB2[n] )- ( Aij + log ₂ QTAB\[n] ) , (K=i<=w, \<=}<=m) , where is the transformed A The range of values of the pixel data of the zth row and column in the subblock.

Step 504: Determine, according to the value range difference obtained in step 503, an adjustment factor of the transform Α. In this step, the adjustment factor is determined in the same manner as in the first embodiment, and is determined for the pixel data of the i-th row and the j-th column in the sub-block, including =.

Thus, the obtained adjustment factor is varied according to the change in the pixel position.

Step 505, applying a transform A to the coding block C, obtaining the transformed data, and multiplying the obtained data by the adjustment factor of the transform A.

In this step, when multiplied by the adjustment factor, the result of transforming the pixel data of the i-th row and the j-th column is multiplied by the corresponding adjustment factor q, _y , that is, ς.' = ς. · 2 The adjustment of the transformation result is more accurate.

Step 506, quantizing each of the quantization step sizes found by transform Α.

At this point, the compensation for the data obtained by the transformation A is completed, and the compensation principle is the same as that of the first embodiment, and will not be described here. The difference is that in the compensation process, the calculation of the adjustment factor changes for the change of the position of the pixel data, so that the compensation result is more accurate, so that the final two kinds of transformations and the corresponding quantized data have a higher degree of consistency. Thereby, a better transformation can be selected more effectively, and the coding efficiency is further improved.

In the actual encoding, the adjustment factor of the position of the pixel according to the position of the pixel may be adjusted due to the difference of the encoded data, so the adjustment factor of the transform A at the time of encoding can be appropriately determined according to the characteristics of the encoded data. Adjustment. The encoded data characteristic may be a value of each pixel location after the data block is transformed A. The adjustment factor of transform A is written into the encoded code stream. The adjustment factor of transform A can be written in the sequence header or image header or strip header or macroblock header of the code stream. The adjustment factor of the transformation A is used in the same manner as the above method, and will not be described again here.

The method in this embodiment can also be implemented by using the apparatus shown in FIG. 4 in the first embodiment.

Considering the problem of compensation accuracy, the embodiment also provides an evolved compensation method, including: determining an adjustment factor f( ) and a coefficient offset according to the obtained two transforms and the difference between the corresponding quantized value ranges s^) , when compensating, first multiply the transformed data by f( ), and then add the coefficient offset _s( ) to the product result, so that the following relationship holds f ₍ ) x Aij+ s _(! . ₎ = Avr _B ,

(K=i<=n, K=j<=n), thus ensuring the accuracy of the compensation. Finally, the result of adding the offset is quantized according to the quantization step corresponding to the transform A. Correspondingly, at the decoding end, f ₍ ) and s^) are set corresponding to the encoding end, and the received data to be dequantized is inversely quantized, subtracted and divided by f( ), and then transformed to A correspondingly. The inverse transform to accurately reconstruct the encoded image data.

For the above-mentioned evolved compensation mode, the present invention also provides a corresponding evolved device structure, which is similar to the structure shown in FIG. 4, wherein the transform compensation unit includes an adjustment factor and an offset quantum unit, a transform subunit, and a transform. Post-processing sub-unit and quantization sub-unit. The structure and function of the other first numerical range estimating unit, the second numerical range estimating unit, and the numerical range difference unit are the same as those shown in Fig. 4.

In the transform compensation unit, the adjustment factor and the coefficient offset quantum unit are configured to determine an adjustment factor and a coefficient offset according to the value range difference value provided by the value range difference unit, and provide the converted processing subunit. A transform subunit that applies a first transform to the data to be transformed and provides the transform result to the post-transform processing sub-unit. The post-transform processing sub-unit is configured to multiply the first transformed data by a determined adjustment factor plus an offset, and provide the result to the quantizing sub-unit. Quantized sub-list The element quantizes the result obtained by multiplying the adjustment factor by the quantization step size corresponding to the first type of transform and adding the coefficient offset.

Similar to the first embodiment, the adjustment factor of the transform A can be calculated in advance at the decoding end. When the transform A is used as the better transform mode, the decoding end divides the inverse quantization result by the adjustment of the corresponding transform A after performing inverse quantization. The factor is then inverse transformed to reconstruct the encoded data.

Embodiment 3:

In this embodiment, the compensation method for the data obtained by transforming A is different from the previous two embodiments, and the method of adjusting the quantization points is used for compensation. In addition, similar to the first embodiment, the numerical value difference values of the two transformed values in the embodiment are the difference between the two types of image data and the corresponding quantized value ranges. Specifically, FIG. 6 is a specific flowchart of a method for processing transformed data provided in Embodiment 3. As shown in Figure 6, the method includes:

Steps 601 ~ 603, according to the transformation matrix required by the preset two transformations, calculate the range of values of the image data after the two transformations respectively; calculate the average range of values after the transformation for the two transformations; and calculate the two transformation sums The corresponding quantized value range difference.

The operations in steps 601 to 603 are the same as the operations in steps 301 to 303 in the first embodiment, and will not be described here.

Step 604, determining a quantization point offset according to the difference in step 603.

In this embodiment, by adjusting the quantization points corresponding to the transform A, the data obtained by the transform A is compensated, so that the numerical range consistent with the transform B is obtained, and the optimal transform mode is prepared.

The quantization point corresponding to the transform A may be a preset quantization point, or may be obtained from the transform B. Quantization point. In the video standard, the quantization step size corresponding to two adjacent quantization points usually has a certain multiple relationship, including: Every A quantization points, the corresponding quantization step size is numerically changed to the original one-half. Therefore, the multiple relationship between the quantization step sizes corresponding to the adjacent quantization points is such that the multiple relationship between the quantization step sizes corresponding to the quantization points is (1/2) and the quantization step size and the quantization are utilized. The value obtained by quantizing the step size is proportional. Therefore, the method of adjusting the offset of the quantization point can also adjust the range of data values. According to the transformed Α and the subsequent numerical range difference D′ obtained in step 603, the quantization point offset required to compensate the difference of the numerical range can be obtained. Specifically, the specific numerical multiple corresponding to the numerical range difference D' is 2 ^D ', and after the quantization point plus the offset Δ ρρ, the corresponding quantization step is the original quantization step.

"door ( _γ β ^ρ

The number of the number is (1/2; ^ = 2 times, so that the two are equal, then the quantization point offset can be obtained.

- ”, J

'\ , where II is the rounding operator. Because the size of the quantization step will affect the actual coding effect of the data, A QP can be adjusted according to the results of the experimental data based on the size of the quantization step. Of course, the calculation of the quantization point offset can also use other formulas as long as the difference of the numerical range obtained in step 603 can be reflected. Step 605: Apply transform A to the coding block C to obtain the transformed data, subtract the quantized offset corresponding to the transform A by the determined quantization offset, and perform quantization according to the quantization step size found by the adjusted quantization point. Deriving Δ from step 604

The process of '| can be seen that after the quantization point is adjusted, the adjusted quantization step size has a relationship with the multiple of the original quantization step size of 2 ^D ', that is, the adjusted quantization step size becomes 2 ^D -QTABl[s] , where QTABl [s] is the original quantization step size. Then, using the result of the quantization step size 'the father ^ ί positive quantized to ^{2 Α ^ · ρΣ4 1 · 2} ϋ = 2 Avr ^ eTAB2ls], a numerical range is represented in binary Αν + log ₂ QTAB2 [s]. Obviously, the numerical range is the same as the numerical range of the transformed B and the corresponding quantized data, and the purpose of making the final numerical range of the transform A and the transform B consistent is achieved.

The transform data processing on the encoding side is performed in accordance with the above method, thereby selecting the preferred transform method in transform A and transform B. If the final selected transform mode is transform A, the data sent to the decoder is the data processed by transform A and corresponding quantization and entropy coding. At this time, the decoding end needs to calculate the quantization point offset corresponding to the transform A according to steps 601 to 604 in advance, and compensate the quantization point offset, including adding Δ^ to the quantization point corresponding to the original transform A. ρ, the inverse quantization of the data and the inverse transformation of the data are performed by using the quantization step size corresponding to the changed quantization point to accurately reconstruct the data of the encoding end.

In the decoding step, when the quantization point offset is compensated, the Δβ 减 may be subtracted from the quantization point corresponding to the original A of the transformation A, and the data is inversely quantized by the quantization step corresponding to the changed quantization point. And inverse transformation. Adding or subtracting Δρρ from the quantization point corresponding to the transform primitive is determined by the quantization table characteristic, which is a preset step, and the purpose is that the data of the encoding end can be accurately recovered at the decoding end after being processed by the transform.

In the actual coding, the transform A quantization offset may be adjusted due to the difference of the encoded data, so the adjustment factor of the transform A at the time of encoding can be appropriately adjusted according to the characteristics of the encoded data. The encoded data characteristic may be an actual average range of values after the data block has been transformed A. The quantized offset of transform A is written into the encoded code stream. That is, the quantization offset subtracted by the quantization point corresponding to the transform A may depend on the range of values after the data is transformed A by encoding a sequence, an image, a strip or a macroblock, and the subtracted quantization offset The amount is written to the encoded code stream. The quantization offset of transform A can be written in the sequence header or image header or slice header or macroblock header of the code stream. If the decoding end knows that the coded code stream contains quantization points The offset information is obtained by subtracting the quantization offset from the sequence header or the image header or the slice header or the macroblock header in the code stream, and subtracting the quantization offset from the quantization point corresponding to the original of the transform A. Here, the quantization point corresponding to the coding end transform A and the quantization point corresponding to the decoding end transform A may be preset quantization points or the same quantization points as the transform B.

This embodiment further provides another embodiment of the transform data processing apparatus, which can be used to implement the method shown in Fig. 6 above. Fig. 7 is a view showing a specific configuration of the down conversion data processing apparatus of the embodiment. As shown in FIG. 7, the apparatus includes a first numerical range estimating unit, a second numerical range estimating unit, a numerical range difference unit, and a transform compensating unit. The transform compensation unit includes a quantization point correction subunit, a transformation subunit, and a quantization subunit.

In the transform compensation unit, the quantization point correction sub-unit is configured to determine the quantization point offset according to the difference of the numerical range provided by the numerical range difference unit, and subtract the quantization point corresponding to the first type of the quantization point The offset, and the adjusted result is provided to the quantized subunit. A transform subunit applies a first transform to the data to be transformed, and provides the transformed result to the quantized subunit. And a quantization subunit, configured to quantize the transformed result according to the adjusted quantization point provided by the quantization point correction subunit. In the specific implementation process of the above embodiment, the difference of the numerical range after the transformation of A and B is calculated according to the average value range. In fact, the calculation of the difference can be performed in the manner of the second embodiment, that is, the difference of the numerical range is calculated for each pixel in the sub-block after the transformation A. Correspondingly, when the quantization point corresponding to the transform A is adjusted in step 604, the pixels in the sub-block after the transform A are also adjusted, specifically, AQP lwx Ad'^) ^ Then, according to different pixels in step 605 , quantizing according to the quantization step size corresponding to the adjusted quantization point. In this way, the result of the compensation can be made more accurate, and the numerical range of the final two kinds of transforms and the corresponding quantized data is more consistent, so that the better transform can be selected more effectively, and the coding efficiency is further improved.

Correspondingly, after the coding end selects the transform A as the preferred transform mode, the quantization point offset corresponding to each pixel data after the transform A is obtained in advance according to the above method, and the corresponding data of the data to be inverse quantized is respectively quantized. The offset is added to the point, and then the quantization step size corresponding to the changed quantization point is used for inverse quantization and inverse transformation to accurately reconstruct the encoded data.

When the quantization offset needs to be written into the code stream, the code stream can be written, parsed, and quantized according to the foregoing method. The quantization point offset is processed at the encoding end and the decoding end.

In the above three embodiments, the transformed data is compensated in two different ways, so as to achieve the purpose of adjusting the range of values. Specifically, one way is to multiply the transformed data by an adjustment. Factor, another way is to change the quantization point corresponding to the transform. In fact, the above two methods can be combined to compensate the transformed data. For example, multiplying the transformed data by an adjustment factor f, and offsetting the quantization point by s, the purpose is to make the following relationship true: fx A ^ +s A ^ , (l<=i<=n, l<=j <=n). Since A ^ and Αντ _β are a priori data, f and s can be preset values, and the values are not unique. The specific f and S can be set according to the A ^ and Α τ _β data size relationship and the codec implementation conditions. If f and S need to be adjusted according to the image encoded data, f and S can be written to the sequence header, the image header, the slice header, the macroblock header, and the like in the encoded code stream. If the coded stream contains f and S, the decoding end is encoded. The corresponding position of the code stream, such as the sequence header, the image header, the slice header, the macroblock header, etc., is parsed to obtain f^. S. Of course, all the pixel data after sub-block transformation can also be multiplied by a corresponding adjustment factor, and the quantization point is offset by s ₍ ), so that the following relationship holds: f ₍ ) x Aij+ s _(! . ₎ = Avr _B , (K=i<=n, l<=j<=n). Among them, f ₍ ) and s ₍ ) can be set according to the data size relationship of A ^ and ^ and the codec implementation conditions. If .) and ^) need to be adjusted according to the image coded data, then .) and can be written to the sequence header, image header, slice header, macroblock header, etc. in the code stream.

If the data obtained by the transformed A is compensated in the above manner, the same f ₍ os ₍ ) may be set in advance at the decoding end. After receiving the encoded data corresponding to the transformed A, the transform A is adjusted before the inverse quantization of the data. For the corresponding quantization point, add s^) to the original quantization point, and divide the inverse quantized data by the corresponding adjustment factor f( ), and then use the quantized compensation pair corresponding to the changed quantization point by the adjustment factor. The inverse anti-quantized data is inverse quantized to accurately reconstruct the encoded data.

If the coded stream contains f() and then the decoder extracts f() and s ₍ ) from the corresponding locations of the encoded stream, such as sequence headers, picture headers, slice headers, macroblock headers, and the like.

The present invention also provides a corresponding device structure in response to the combination of the above two compensation methods. The structure of the apparatus is similar to that shown in FIG. 7, except that the transform compensation unit includes an adjustment factor and a quantization point correction subunit, a transformation subunit, a post-transformation processing sub-unit, and a quantization sub-unit. The structure and function of the other first numerical range estimating unit, the second numerical range estimating unit, and the numerical range difference unit are the same as those shown in Fig. 7.

In the transform compensation unit, the adjustment factor and the quantization point correction sub-unit are configured to determine an adjustment factor and a quantization point offset according to a difference value range provided by the value range difference unit, and quantize the first type of transform The quantization point offset is subtracted from the point, the adjusted quantization point is provided to the quantization sub-unit, and the determined adjustment factor is provided to the transformed processing sub-unit. A transform subunit that applies a first transform to the data to be transformed and provides the transform result to the post-transform processing sub-unit. After the transformation a subunit, configured to multiply the first transformed data by a determined adjustment factor, and provide the result to the quantized subunit. The quantizing sub-unit quantizes the result multiplied by the adjustment factor according to the quantization step size corresponding to the adjusted quantization point.

In the above three embodiments, the transform A and the transform B may use different quantization tables, that is, the quantization step sizes found according to the quantization points are different, but for the same case of the quantization table, the above method is used. The loss of data accuracy is large.

For the same case of the quantization table, the present invention proposes an embodiment of the fourth embodiment. For one of the transformations, a set of quantization tables is redesigned so that the numerical ranges of the two transformations and the corresponding quantized data are identical.

Embodiment 4:

Different from the foregoing three embodiments, in this embodiment, the difference value range of the two transformed numerical ranges is the difference between the numerical values of the two transformed images of the image data.

FIG. 8 is a specific flowchart of a method for processing transformed data according to Embodiment 4 of the present invention. As shown in Figure 8, the method includes:

Steps 801 to 802, according to the transformation matrix required for the preset two transformations, calculate the image data to undergo two converted numerical ranges respectively; for the two transformations, respectively calculate the transformed average numerical range; in steps 801-802 The operation is the same as the operations of steps 301 to 302 in the first embodiment, except that the above operations are performed separately for different quantization points. That is to say, for all the different macroblocks, the average value range of the two transformed macroblocks is calculated separately: Avr and Α^^μ where η is the quantization point index.

Step 803: Calculate the difference between the two transformed numerical ranges.

The two transformed value range differences are calculated for different quantization points, specifically, Ad(«) = Avr _B {n) - Avr _A {n). In fact, as mentioned above, since the video coding and decoding theory usually uses the homo-quantization method, that is, every few quantization points, the quantization step size is reduced to half, so the value of Δ(1(«) There is also a certain regularity. Based on this, a simplification can be made here. Specifically, it is assumed that every m quantization points, the quantization step size is reduced to half, then a continuous m Δ(1(«) can be selected. For example, this step only calculates Ad(l), Ad(2), L, Ad(m), and the following is the calculation of the m differences.

Step 804, setting a quantization step size and a shift bit number corresponding to each quantization point, and forming a quantization table. In this step, when the quantization step size is set, it is performed according to the difference calculated in step 803. Specifically, corresponding to the m differences calculated in step 803, the quantization step corresponding to the first to mth quantization points is set to be: 2 ^AdW xR, 2 ^Ad(m) xR, where R is a constant coefficient The value is to make the value of the quantization step size an integer and improve the calculation accuracy (the numerator is large, naturally more information can be retained), but considering the limitation of the hardware storage range, the R value has an upper limit. According to the principle of uniform quantization, the quantization step size of the other quantization points can be derived from the quantization step corresponding to the 1st to mth quantization points.

Of course, when setting the quantization step size, all the quantization step ^sizes can also be set, and the quantization step size corresponding to the nth quantization point is 2 ^AdW xR.

Set the shift bit number to a fixed value t. When the t is large, the coding effect is poor, and the compression coding rate is high; when the t is small, the coding effect is good, and the compression coding rate is low. Therefore, the specific value of t needs to consider the coding effect and compression coding rate comprehensively, and find a t corresponding to a better balance point.

The quantization table is formed according to the shift bit number and the quantization step size set as described above, and the process is the same as the existing implementation, and will not be described here.

Step 805: Receive data to be transformed, and apply transform A to obtain transformed data, and quantize the transformed data according to the quantization table.

After the above processing, the value range of the data after the encoding end data is quantized by the transform A and the quantization table is substantially consistent with the transformed B and the corresponding quantized numerical range. The principle of the present invention is the same as that of the first embodiment. Specifically, since the quantization step size is proportional to the quantized data in the quantization process, the data obtained by the transform A is quantized according to the quantization step size. The process is in fact substantially equivalent to the process of multiplying the data obtained by transform A by an adjustment factor in the first embodiment, which is multiplied by 2 ^AdW . Therefore, this embodiment can also achieve the purpose of adjusting the value range so as to be consistent with the transformation B and the corresponding quantized value range.

Correspondingly, an inverse quantization table corresponding to the quantization table of the encoding end is also designed at the decoding end. In the inverse quantization table, considering the method of uniform quantization at the encoding end, the quantization step size of the nth quantization point is twice the quantization step size of the (n+m)th quantization point, and is utilized in the inverse quantization table. The shift compensates for the multiple relationship between the quantization step sizes, that is, the number of shift bits of the nth quantization point in the inverse quantization = the shift number of the (11+111)th quantization point +1. The number of shift bits of the 1st to mth inverse quantization points can be set to k, and the number of shift bits of the m+1th to 2m inverse quantization points is correspondingly k-l. As for the setting of the k value, it is also considered in consideration of the storage range of the hardware, and the principle is the same as the case of considering the hardware storage range.

Next, according to the setting of the shift bit number described above, the inverse quantization step size of the inverse quantization table is calculated as IQ(n), where n is the quantization point. IQ(n) needs to satisfy the relationship 2 ^AdW xRxIQ(n)=

, where, reflects the change in the number of shift bits, ie every

m quantization points, the number of shift bits is increased by 1. According to this formula, IQ(n)=

2 ^Ad (") . R where h is a constant constant, the larger the value, the better, but it is also limited by the hardware storage range. The inverse quantization table is established as described above, so that at the decoding end, the data of transform A is used. The reconstructed data can be recovered using the established inverse quantization table after the inverse transformation.

This embodiment further provides a specific implementation of the transform data processing apparatus, which can be used to implement the method shown in FIG. 8 above. Fig. 9 is a view showing a specific configuration of the down conversion data processing apparatus of the embodiment. As shown in FIG. 9, the apparatus includes a first numerical range estimating unit, a second numerical range estimating unit, and a number Value range difference unit and transform compensation unit. The transform compensation unit includes a quantization table establishment subunit, a transformation subunit, and a quantization subunit.

The value range difference unit is configured to calculate a difference between the two transformed value ranges according to the first and second converted value ranges respectively, and provide the difference as a numerical range characteristic difference value to the The transformation compensation unit.

In the transform compensation unit, the quantization table establishing subunit is configured to establish a corresponding quantization table for the first type of transform according to the numerical range feature difference value, and provide the quantized subunit. A transform subunit, configured to apply a first transform to the data to be transformed, and provide the transform result to the quantized subunit. The quantizing subunit quantizes the first transformed result according to the quantization table corresponding to the first transform.

In the above embodiments of the present invention, the data of the transformed A is compensated as an example to illustrate the specific implementation of the present invention. In fact, the data of the transformed B can also be compensated, and the implementation manner is the same, but only The parameters corresponding to the transformation A and the transformation B are interchanged in the corresponding formula, and will not be described here.

The foregoing method for transforming data processing can be used in an encoding method. FIG. 10 is a schematic flowchart of an encoding method provided in Embodiment 5 of the present invention. The encoding method includes the following steps:

Al, receiving data to be transformed;

The data to be transformed includes a predicted image block residual.

A2, performing a first transformation on the data to be transformed to obtain first transformed data; Performing a second transformation on the data to be transformed to obtain second transformed data;

The first transform is here a 4x4 size transform;

The second transform is here an 8x8 size transform;

A3. Determine an adjustment parameter of the first transformation according to the first transformed data and the second transformed data, and adjust the first transformed data according to the adjustment parameter and the second transformed parameter.

The first transformed data and the second transformed data herein include the first converted numerical range and the second transformed numerical range. For convenience of description, the first transformed data and the second transform are used. The name of the post data.

Adjusting the parameter may include determining an adjustment factor according to the first transformed data and the second transformed data, and then adjusting the first transformed data according to the adjustment parameter and the second transformed parameter includes : multiplying the first transformed data by the adjustment factor; and quantizing the data multiplied by the adjustment factor according to the corresponding quantization step size.

The adjustment parameter may include a quantization offset determined according to the first transformed data and the second transformed data, the second transformed parameter being a second transformed quantization point, and the adjusting parameter and the second Adjusting the first transformed data by the transformed parameter includes: subtracting the quantized offset corresponding to the second transform as the quantized point of the first transform; using the adjusted first transformed quantized point pair The first transformed data is quantized.

The adjustment parameter may include a quantization offset and an adjustment factor determined according to the first transformed data and the second transformed data; the parameter of the second transformation is a quantization point corresponding to the second transformation, and then according to the Adjusting the parameter and the parameter of the second transform to adjust the first transformed data comprises: multiplying the first transformed data by the adjustment factor, and subtracting the quantization point corresponding to the second transform from the quantization The offset is used as a quantization point of the first transform; the data multiplied by the adjustment factor is quantized by the quantization point of the first transform. The adjustment parameter may further include an adjustment coefficient and a quantization shift offset determined according to the first transformed data and the second transformed data; the parameter of the second transformation includes: shifting a quantization point of the second transformed data Adjusting the first transformed data according to the adjustment parameter and the second transformed parameter includes: multiplying the first transformed data by the adjustment coefficient, and using the first transform And shifting the bit number of the quantization point to quantize the data multiplied by the adjustment coefficient; the shift bit number of the quantization point corresponding to the first transform is a shift bit number of the quantization point corresponding to the second transform The sum of the quantized shift offsets is described.

The adjustment parameter may include an adjustment factor and a coefficient offset determined according to the first transformed data and the second transformed data, and then adjusting the first transformation according to the adjustment parameter and the second transformed parameter. The subsequent data includes: multiplying the first transformed data by the adjustment factor, adding the coefficient offset and performing quantization.

A4. Write the adjustment parameter into the encoded code stream.

The adjustment parameter is written into a sequence header or an image header or a slice header or a macroblock header in the encoded code stream for use by the decoding end. After the above processing, the numerical range of the first transformed data is substantially the same as the numerical range of the second transformed data, and the effect of the transform on the data can be more effectively reflected in the encoding to select a better transform, thereby improving encoding. effectiveness. FIG. 11 is a schematic flowchart of a decoding method according to Embodiment 6 of the present invention, where the method includes:

Bl, receiving the code stream, decoding the code stream, obtaining the first transformed data and adjusting parameters; the first transformed data is obtained by entropy decoding if the currently decoded image block is determined to use the first transform The data after the image block residual is inverse transformed by the first transform; The first transform is here a 4x4 size inverse transform;

The second transform is here an inverse transform of 8x8 size;

The adjustment parameter corresponds to an adjustment parameter written to the code stream at the time of encoding.

B2. Adjust the first transformed data according to the adjustment parameter and the second transformed parameter. This step includes a variety of methods, namely:

(1) determining, according to the adjustment parameter and the parameter of the second transformation, an adjustment factor for the first transformed data: performing inverse quantization on the first transformed data according to a quantization step size corresponding to the first transformation, and The inverse quantization result is inversely transformed with the quotient of the adjustment factor.

(2) if the received adjustment parameter is a quantization offset, the parameter of the second transformation is a quantization point corresponding to the second transformation, and the quantization point of the first transformation is a quantization point corresponding to the second transformation minus The quantizing offset, the adjusting the first transformed data according to the quantized offset and the quantization point corresponding to the second transform comprises: using the quantization point corresponding to the first transform to the first transform The latter data is inverse quantized and the inverse quantization result is inversely transformed.

(3) if the adjustment parameter includes a quantization offset and an adjustment factor for the first transformed data; then the parameter of the second transformation is a quantization point corresponding to the second transformation, and the quantization point of the first transformation is the The quantized point of the second transform is subtracted from the quantized offset; the adjusting the first transformed data according to the adjusting parameter and the second transformed parameter comprises: using a quantization point corresponding to the first transform Performing inverse quantization on the first transformed data; inversely transforming the inverse quantization result and the quotient of the adjustment factor.

(4) if the adjustment parameter includes a first transformed adjustment coefficient and a quantization shift offset for the first transformed data; the second transformed parameter includes a shift number of the quantized point of the second transformed data And adjusting the first transformed data according to the adjusting parameter and the second transformed parameter, comprising: subtracting the quantized shift of the quantized point corresponding to the second transform by the quantization shift An offset amount as a shift bit number of a quantization point corresponding to the first transform; a shift bit number of the quantization point corresponding to the first transform Performing inverse quantization on the data received by the first transform; and inversely transforming the inverse quantization result and the quotient of the adjustment coefficient.

(5) determining an adjustment factor for the first transformed data according to the adjustment parameter and the parameter of the second transformation; determining a coefficient offset for the first transformed data according to the adjustment parameter and the parameter of the second transformation The adjusting the first transformed data according to the adjusting parameter and the second transformed parameter comprises: performing inverse quantization on the received data according to the quantization step size corresponding to the first transform, and The quantized result is subtracted from the coefficient offset, and the subtraction result is inversely transformed with the quotient of the adjustment factor.

Corresponding to the encoding end, the adjustment parameters are obtained from a sequence header or an image header or a strip header or a macroblock header in the code stream.

FIG. 12 is a schematic structural diagram of an encoding apparatus according to Embodiment 7 of the present invention, where the encoding apparatus includes: a data receiving unit, configured to receive data to be transformed;

a transform unit, configured to perform first transform on the data to be transformed received by the data receiving unit, to obtain first transformed data, and perform second transform on the data to be transformed to obtain second transformed data ;

The first adjusting unit determines an adjustment parameter according to the second transformed parameter, the first transformed data, and the second transformed data, and adjusts the first transformed data according to the adjustment parameter and the second transformed parameter;

And a writing unit, configured to write the adjustment parameter into the encoded code stream.

13 is a schematic structural diagram of a decoding apparatus according to Embodiment 8 of the present invention, where the decoding apparatus includes: a code stream receiving unit, configured to receive a code stream;

Decoding unit, decoding the code stream to obtain first transformed data and adjusting parameters; and second adjusting unit, configured to adjust the first transform according to the adjusting parameter and the second transformed parameter The data.

The above encoding device and decoding device can be used in a system, and are not described here. Embodiment 9: An encoding method

Embodiment 9 provides an encoding method, which can adaptively obtain a quantization offset (the aforementioned adjustment parameter) at the encoding end. As shown in FIG. 14, the method includes:

Cl, receiving data to be transformed and encoded parameter information;

C2, performing a first transformation on the data to be transformed to obtain first transformed data;

C3. Determine an adjustment parameter according to the encoded parameter information and the second transformed parameter, and adjust the first transformed data according to the adjustment parameter and the second transformed parameter;

C4. Write the adjustment parameter into the encoded code stream.

The following force p is described in detail.

Let the scale of transform A be 4x4, and the scale of transform B be 8x8.

In this embodiment, the data to be transformed and the encoded parameter information are first received. Performing a first transformation on the data to be transformed to obtain first transformed data, and determining an adjustment parameter according to the encoded parameter information and the second transformation parameter. The second transform parameter is a quantization point corresponding to the second transform, and the adjustment parameter is a quantization offset.

The encoded parameter information includes the number of intra-predicted image blocks using the first transform, the number of intra-predicted image blocks using the second transform, all using the number of intra-predicted image blocks, and the first in the P-frame or B-frame. The number of transformed intra-predicted image blocks, the number of intra-predicted image blocks using the first transform in P and B frames, the number of second transformed intra-predicted image blocks in P-frames or B-frames, P-frames and B-frames The number of intra-predicted image blocks used in the second transform, the number of intra-predicted image blocks in the P-frame and the B-frame, the number of image blocks in the P-frame using the skip mode or the direct mode, and the skipping in the P-frame Mode and The number of image blocks in the direct mode, the number of image blocks in the B frame using the skip mode or the direct mode, the number of image blocks in the B frame using the skip mode and the direct mode, and the inter prediction mode image block in the P frame. The number, or one or more of the number of inter prediction mode image blocks used in the B frame. The skip mode is a commonly used technique in video encoding and decoding, which means that the current image block motion vector is obtained according to the encoded or decoded image block information, and there is no coding residual in the image block; the direct mode meaning is based on the encoded or already The image block information is decoded to obtain a current image block motion vector, but the image block contains an encoded residual.

It is noted that the number of intra prediction image blocks using the first transform is img->intra4x4num, the number of used intra prediction image blocks is img->intra-num, the skip mode is used in the P frame and or directly The number of image blocks of the mode is img->pskip-num, the number of intra prediction image blocks used in the P frame is img->p_in, and the number of image blocks in the P frame using the inter prediction mode is Img->pnum, the quantization point corresponding to the second transform is i _mg -> qp, and the quantization offset is img->qp_shift.

In this embodiment, three tables for determining the quantization offset are stored at the encoding end, and each table contains 64 elements, which are recorded as QPshift_table[0], QPshift_table[l], and QPshift. — table[2] three arrays, each containing 64 elements, represented as follows:

QPshift_table[0]={0,l, 1,1, 1,2,2,2,2,3,3,3,3,4,4,4,4,5,5,5,5,6, 6,6,6,7,8,8,8,8,8,8,8,8,9,9,9,10,10,10,10,10,10,10,10,10,10, 10,10,10,10,10,10,10,10,10,10,10,10,10, ^,10,10}

6,6,6,6,6,6,6,6,6,6,6,6

QPshift_table[2]={0,l, 1,1,2,2,3,3,3,4,4,4,5,5,6,6,7,8,8,9,9,10, 10,10,10,10,11,11,11,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12, 12,12,12,12,12,13,13,13,1 3,13,13,13,13,13,13,13, When the adaptive block transform technique is used, if the current coded image is an I frame and is the first image of the current sequence or the current image group, img->intra4x4num, img->intra-num, img->pskip-num, Img -> p - intra and img -> pnum is 0. Use QPshift_table [0] to determine img -> QP-shift. The method is: img->QP_ shift=QPshift_table[0][img- >qp]. The image group referred to herein includes a plurality of I frames, P frames, B frames, or a combination of the above types.

If the current encoded image is an I frame and is not the current sequence or the first image of the current image group but there are no images already encoded in the P frame in the current sequence or the current image group, then two parameters are set, which is recorded as the first The determination value vl and the second determination value v2, vl and v2 are calculated as follows:

Vl =img->intra4x4num * 5.0/ img -> intra- num/ (img->QP_shift+ 12)

V2=0.2

Where the symbol '*' stands for multiplication and the symbol ' /' stands for division.

After obtaining vl and v2, the quantization offset is obtained according to the following logic: img->QP_ shift:

(1) If vl is less than 0.3 and v2 is less than 0.4, or vl is less than 0.4 and v2 is less than 0.15, or vl is less than 0.5 and v2 is less than 0.1, the value of img->QP_shift is indexed by img->qp, in the table Find in QPshift_table[l];

If condition (1) is not satisfied, the following judgment is made:

(2) If vl is greater than 0.5 or v2 is greater than 0.5 or vl *v2 is greater than 0.2, the value of img -> QP_shift is indexed by img->qp and found in the table QPshift_table[2];

If the conditions (1) and (2) are not satisfied, the value of img->QP_shift is indexed by img->qp and found in the table QPshift_table[0].

If the current encoded image is not the first image of the current sequence or the current image group and the image in the current sequence or the current image group has been encoded in the P frame, two parameters are set, which are recorded as the first determined value vl and the second The determined values v2, vl and v2 are calculated as follows: Vl =img->intra4x4num * 5.0/ img -> intra- num/ (img->QP_shift+ 12);

V2=img->pskip_num* 1.0/img -> pnum

If condition (1) is not satisfied, the following judgment is made:

After the quantization offset img->QP_shift is obtained, the weight coefficient lambda corresponding to the first transform in the rate-distortion optimization is obtained according to the table pre-stored by the encoding end. This table is denoted as QP-lambda-table-4x4, which contains 16 elements in this embodiment. In the first transformation, the lambda using the intra-coded image block is recorded as img->lambda4x4I, and the lambda of the image block using the inter-prediction mode is img->lambda4x4p, img->lambda4x4I and img->lambda4x4 :

Img->lambda4x4I=QP— lambda—table— 4x4[img->QP— shift]

Img->lambda4x4p=img->lambda4x4I* 0.9

QP—lambda—table—4x4 behaves as follows:

QP—lambda—table—4χ4[16]={1.0,1.1,1.3, 1.5,1.8,2.1,2.6,3.1,3.5,4.0,4.6,5.1,5.7,6. 3,100,100} If the current image to be encoded is a P frame, then img->QP_shift is set to 5 and img->lambda4x4p is set to 1.6.

If the current image to be encoded is a B frame, img->QP_shift is set to 5.

After the img->QP_shift is obtained by the above method, the quantization point of the first transform is calculated according to the parameter of the second transform (ie, the quantization point of the second transform), and the specific method is to subtract the quantized point of the second transform The quantization offset is used as the quantization point of the first transform. The first transformed data is quantized using the calculated first transformed quantized points. The quantization offset img->QP_shift is written into the encoded code stream.

The tenth embodiment further provides an encoding device, which can be used to perform the encoding method provided in the ninth embodiment. As shown in FIG. 15, the encoding device includes:

a third adjusting unit, configured to determine, according to the encoded parameter information and the second transformed parameter received by the third receiving unit, an adjustment parameter, and adjust the third transformation unit according to the adjustment parameter and the second transformed parameter First transformed data;

And a third writing unit, configured to write the adjustment parameter obtained by the third adjusting unit into the encoded code stream. In this embodiment, the third receiving unit receives the data to be transformed and the encoded parameter information. The third transform unit performs a first transform on the data to be transformed to obtain first transformed data, and the third adjusting unit determines the adjustment parameter according to the encoded parameter information and the second transform parameter. The second transform parameter is a quantization point corresponding to the second transform, and the adjustment parameter is a quantization offset.

The encoded parameter information includes the number of intra-predicted image blocks using the first transform, the number of intra-predicted image blocks using the second transform, all using the number of intra-predicted image blocks, and the first in the P-frame or B-frame. Number of transformed intra prediction image blocks, intra prediction using the first transform in P and B frames Number of image blocks, number of second transformed intra-predicted image blocks used in P-frame or B-frame, number of second transformed intra-predicted image blocks in P-frame and B-frame, intra prediction used in P-frame and B-frame The number of image blocks, the number of image blocks using the skip mode or the direct mode in the P frame, the number of image blocks using the skip mode and the direct mode in the P frame, and the image block using the skip mode or the direct mode in the B frame The number of image blocks in the B frame using skip mode and direct mode, the number of inter prediction mode image blocks in the P frame, or the number of inter prediction mode image blocks in the B frame or A variety. The skip mode is a commonly used technique in video encoding and decoding, which means that the current image block motion vector is obtained according to the encoded or decoded image block information, and there is no coding residual in the image block; the direct mode meaning is based on the encoded or already The image block information is decoded to obtain a current image block motion vector, but the image block contains an encoded residual.

The third adjusting unit calculates the quantized offset according to the encoded information and the quantized point of the second transform, and calculates the quantized point of the first transform according to the quantized offset and the quantized point of the second transform, and obtains the obtained quantized point. The first transformed quantized point quantizes the first transformed data (the first transformed data obtained by the third transform unit is adjusted according to the adjustment parameter and the second transformed parameter). The steps of the specific calculation can be referred to the description of the embodiment 9.

The third write unit writes the quantization offset into the encoded code stream to obtain a quantized offset for decoding operation when decoding. The ninth embodiment and the tenth embodiment obtain the adjustment parameters (quantization offset) according to the encoded parameter information and the second transformation parameter, so that the coding end has better flexibility. Since the encoding end has written the calculated adjustment parameters into the encoded code stream, the decoding end only needs to obtain the adjustment parameter from the code stream. The method described in this embodiment takes into account the coding performance without additionally increasing the burden on the decoding end. Through the description of the above embodiments, those skilled in the art can clearly understand that the present invention can be implemented by means of software plus a necessary hardware platform, and of course, can also be implemented entirely by hardware. Based on such understanding, all or part of the technical solution of the present invention contributing to the background art may be embodied in the form of a software product, which may be stored in a storage medium, such as

ROM/RAM, diskette, optical disk, etc., comprising instructions for causing a computer device (which may be a personal computer, server, or network device, etc.) to perform the methods described in various embodiments or portions of the embodiments of the present invention. .

The above are only the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and scope of the present invention are intended to be included within the scope of the present invention.

Claims

Rights request

1. An encoding method, comprising:

Receiving data to be transformed;

And adjusting the adjustment parameter according to the first transformed data and the second transformed data, and adjusting the first transformed data according to the adjustment parameter and the second transformed parameter.

2. The method according to claim 1, wherein the determining the adjustment parameters according to the first transformed data and the second transformed data comprises:

The numerical range characteristic difference values of the two transformations are calculated based on the numerical range of the first transformed data and the second transformed data.

3. The method according to claim 2, wherein the calculating the numerical range characteristic difference values of the two transformations comprises:

Calculating the numerical range feature difference values of the two transforms according to the respective quantization step sizes of the two transforms found by the quantization points corresponding to the image data; wherein, the two types of transformed numerical range feature difference values are image data respectively passing through two The difference between the transform and the corresponding quantized value range.

The method according to claim 3, wherein the calculating the difference between the two types of transformed image data and the corresponding quantized value range includes:

For the two transformations, calculate the average range of values after transformation and

Avr _B = TBij , where, for the image data, the value range of the pixel data of the first row after the first transformation is obtained according to the scale of the first transformation, ^ / is the image data according to two changes The value range of the pixel data of the first row and the second row after the second transformation in the sub-block obtained by the second transformation of the second transformation, the scale of the first transformation is wx«, and the scale of the second transformation is w xw;

According to the two transformed average value ranges and the corresponding quantization step sizes, the difference D'= ( Aw _B + log ₂ QTAB2[n] ) - ( ) is calculated for the image data after two transformations and corresponding quantized values. Avr _A + log ₂ QTABl[n] ) , where 2∑4 [w] and 2∑432["] are quantization step sizes respectively found according to the quantization points corresponding to the first and second transforms.

The method according to claim 3, wherein when the image data is video data, the difference between the two types of transformations and the corresponding quantized value ranges of the calculated image data is: The second transformed average value range in the transformation Aw _B = ∑∑ Bij;

'■=1 =1 Calculated according to the average value range after the second transform and the value range of all the pixel data of the sub-block obtained according to the scale of the transform, and the corresponding quantization points of the two transforms, The difference between the value range of the pixel data of the first row and the second column after the first transform and the corresponding quantization and the average value range after the second transform and the corresponding quantization d _u) = ( Avr _B + log ₂ QTAB2[n] ) -

( Aij + log ₂ QTABl[n] ) , where is the value range of the pixel data of the _th column of the z-th row after the first transformation of the sub-block, ρ∑^ι[«] and ρ∑^2[ «] is the quantization step size found by the quantization points corresponding to the first and second transforms, respectively.

The method according to any one of claims 1 to 5, wherein the adjustment parameter comprises an adjustment factor determined according to the first transformed data and the second transformed data, Adjusting the parameter and the parameter of the second transform to adjust the first transformed data includes:

Multiplying the first transformed data by the adjustment factor;

The data multiplied by the adjustment factor is quantized according to the quantization step size corresponding to the first transform.

The method according to any one of claims 1 to 5, characterized in that the adjustment parameter And including a quantization offset determined according to the first transformed data and the second transformed data, where the parameter of the second transformation is a quantization point of the second transformation, and the parameter according to the adjustment parameter and the second transformation Adjusting the first transformed data includes:

And subtracting, by the quantized point corresponding to the second transform, the quantized offset as the quantized point of the first transform; and quantizing the first transformed data by using the adjusted quantized point of the first transform.

The method according to any one of claims 1 to 5, wherein the adjustment parameter comprises a quantization offset and an adjustment factor determined according to the first transformed data and the second transformed data; The parameter of the second transformation is the quantization point corresponding to the second transformation, and the adjusting the first transformed data according to the adjustment parameter and the parameter of the second transformation includes:

Multiplying the first transformed data by the adjustment factor, and subtracting the quantization offset from the quantization point corresponding to the second transform as the quantization point of the first transform;

The data multiplied by the adjustment factor is quantized using the quantized points of the first transform.

The method according to any one of claims 1 to 5, wherein the adjustment parameter comprises an adjustment coefficient and a quantization shift offset determined according to the first transformed data and the second transformed data. The parameter of the second transform includes a shift bit number of the quantization point corresponding to the second transform;

The adjusting the first transformed data according to the adjusting parameter and the second transformed parameter includes:

Multiplying the first transformed data by the adjustment coefficient, and quantizing the data multiplied by the adjustment coefficient by using a shift bit of the quantization point corresponding to the first transform;

The shift bit number of the quantization point corresponding to the first transform is the sum of the shift bit number of the quantization point corresponding to the second transform and the quantized shift offset.

The method according to any one of claims 1 to 5, wherein the adjustment parameter comprises an adjustment factor and a coefficient offset determined according to the first transformed data and the second transformed data. The adjusting the first transformed data according to the adjusting parameter and the second transformed parameter comprises: multiplying the first transformed data by the adjustment factor, and adding the coefficient offset And quantify.

The method according to any one of claims 1 to 5, wherein the method further comprises:

The adjustment parameters are written into the encoded code stream.

The method according to claim 11, wherein the writing the adjustment parameter to the encoded code stream comprises:

The adjustment parameters are written to a sequence header or image header or strip header or macroblock header in the encoded code stream.

13. The method according to claim 1, wherein the first transform is a 4x4 transform and the second transform is an 8x8 transform.

14. A decoding method, comprising:

Receiving a code stream, decoding the code stream, obtaining data corresponding to the first transform, and adjusting parameters; and adjusting data corresponding to the first transform according to the adjustment parameter and the second transformed parameter.

The method according to claim 14, wherein the adjusting the data corresponding to the first transform according to the adjusting parameter and the second transformed parameter comprises:

Determining, by the adjustment parameter and the parameter of the second transformation, an adjustment factor of data corresponding to the first transformation:

And dequantizing the data corresponding to the first transform according to the quantization step size corresponding to the first transform, and performing inverse transform corresponding to the first transform on the quotient of the inverse quantization result and the adjustment factor.

The method according to claim 14, wherein the adjustment parameter is a quantization offset, the parameter of the second transformation is a quantization point corresponding to the second transformation, and the quantization point of the first transformation is the Subtracting the quantization offset from the quantization point corresponding to the second transform, according to the adjustment parameter and the The data adjustment of the parameter of the second transformation to the first transformation includes:

And performing inverse quantization on the data corresponding to the first transform by using the quantization point corresponding to the first transform, and performing inverse transform corresponding to the first transform on the inverse quantization result.

17. The method of claim 14 wherein:

The adjustment parameter includes a quantization offset and an adjustment factor of data corresponding to the first transformation; a parameter of the second transformation is a quantization point corresponding to the second transformation, and a quantization point of the first transformation is the second transformation a quantized point minus the quantized offset;

The adjusting the data corresponding to the first transformation according to the adjusting parameter and the second transformed parameter includes:

And performing inverse quantization on the data corresponding to the first transform by using the quantization point corresponding to the first transform; and performing inverse transform corresponding to the first transform on the quotient of the inverse quantization result and the adjustment factor.

18. The method of claim 14 wherein:

The adjustment parameter includes an adjustment coefficient of the first transformation and a quantization shift offset of the first transformed data;

The parameter of the second transform includes a shift bit number of the quantization point corresponding to the second transform;

And subtracting the quantized shift offset from the shift bit number of the quantization point corresponding to the second transform as a shift bit number of the quantization point corresponding to the first transform;

Performing inverse quantization on the data corresponding to the first transform by using the number of shift bits of the quantization point corresponding to the first transform;

An inverse transform corresponding to the first transform is performed on the quotient of the inverse quantization result and the adjustment coefficient.

19. The method of claim 14 wherein: Determining an adjustment factor for the first transformed data according to the adjustment parameter and the parameter of the second transformation;

Determining a coefficient offset for the first transformed data according to the adjustment parameter and the second transformed parameter;

The adjusting the data corresponding to the first transform according to the adjusting parameter and the second transformed parameter includes:

And performing inverse quantization on the received data according to the quantization step size corresponding to the first transform, and subtracting the coefficient offset from the inverse quantization result, and performing the first quotient of the subtraction result and the adjustment factor Transform the corresponding inverse transform.

20. The method according to claim 14, wherein the obtaining the adjustment parameter is specifically obtainable by using a sequence header or an image header or a strip header or a macroblock header in the code stream.

The method according to claim 14, wherein the obtaining the adjustment parameter comprises: receiving data to be decoded;

Performing a first transformation on the data to be decoded to obtain first transformed data;

Performing a second transformation on the data to be decoded to obtain second transformed data;

The adjustment parameter is determined based on the first transformed data and the second transformed data.

The method according to claim 14, wherein the first transform is a 4x4 transform and the second transform is an 8x8 transform.

23. An encoding device, comprising:

a data receiving unit, configured to receive data to be transformed;

a transform unit, configured to perform first transform on the data to be transformed received by the receiving unit, to obtain first transformed data, and perform second transform on the data to be transformed to obtain second transformed data; The first adjusting unit determines an adjustment parameter according to the first transformed data and the second transformed data, and adjusts the first transformed data according to the adjustment parameter and the second transformed parameter.

The device of claim 23, further comprising:

25. A decoding apparatus, comprising:

a code stream receiving unit, configured to receive a code stream;

a decoding unit that decodes the code stream to obtain data and adjustment parameters corresponding to the first transform; and a second adjusting unit, configured to adjust data corresponding to the first transform according to the adjustment parameter and the second transformed parameter.

26. A codec system, comprising: an encoding device and a decoding device; the encoding device comprising:

a data receiving unit, configured to receive data to be transformed;

a transform unit, configured to perform first transform on the data to be transformed received by the data receiving unit, to obtain first transformed data; perform second transform on the data to be transformed, to obtain a second transform The data;

The decoding device includes:

a code stream receiving unit, configured to receive a code stream;

a decoding unit, configured to decode the code stream to obtain first transformed data and adjustment parameters, and second adjustment unit, configured to adjust the first transformed data according to the adjustment parameter and the second transformed parameter.

27. An encoding method, comprising:

Receiving data to be transformed and encoded parameter information;

The adjustment parameters are written into the encoded code stream.

The method according to claim 27, wherein the adjustment parameter comprises a quantization offset determined according to the encoded parameter information and the second transformed parameter, and the second transformed parameter is a second And the adjusting the quantized points, the adjusting the first transformed data according to the adjusting parameter and the second transformed parameter includes:

The method according to claim 27, wherein the encoded parameter information comprises statistical data that has been obtained before encoding the current image in the current sequence or the current image group, and the statistical data includes a frame using the first transform. Number of intra-predicted image blocks, number of intra-predicted image blocks using the second transform, number of all intra-predicted image blocks used, number of intra-predicted image blocks using the first transform in P-frame or B-frame, P-frame and B The number of intra-predicted image blocks using the first transform in the frame, the number of second transformed intra-predicted image blocks in the P-frame or B-frame, and the number of second transformed intra-predicted image blocks in the P-frame and the B-frame, The number of intra prediction image blocks used in P frames and B frames, the number of image blocks using skip mode or direct mode in P frames, the number of image blocks using skip mode and direct mode in P frames, in B frames The number of image blocks using the skip mode or the direct mode, the number of image blocks using the skip mode and the direct mode in the B frame, the number of image blocks in the interframe prediction mode or the B frame in the P frame use

The method according to claim 28 or 29, wherein the step of determining the quantization offset based on the encoded parameter information comprises:

Performing a mathematical operation on the encoded parameter information to obtain a first determined value and a second determined value; performing logical determination using the first determined value and the second determined value and a predetermined condition, and the predetermined result is determined according to the logic Select a table in the table to determine the quantization offset;

A quantization offset amount is determined based on the parameter of the second transform in the table for determining the quantization offset.

The method according to claim 30, wherein the step of performing mathematical operations on the encoded parameter information to obtain the first determined value and the second determined value comprises:

Calculating a proportional relationship between the number of the intra-predicted image blocks to obtain a first relationship value, and calculating a proportional relationship between the number of image blocks using each skip mode or direct mode and the number of image blocks of each inter-prediction mode, or Calculating a proportional relationship between the number of image blocks using each skip mode and the direct mode and the number of image blocks of each inter prediction mode to obtain a second relationship value;

The first determined value and the second determined value are obtained by performing mathematical operations on the first relationship value, the second relationship value, and the second transformed parameter.

32. The method according to claim 28 or 29 or 31, wherein after calculating the quantized offset, the weight parameter lambda in the rate-distortion optimization model is obtained according to the quantization offset and a preset table. .

33. An encoding device, comprising:

a third adjusting unit, configured to receive the encoded parameter information and the second transform according to the third receiving unit The parameter determining the adjustment parameter, and adjusting the first transformed data obtained by the third transform unit according to the adjustment parameter and the second transformed parameter;

And a third writing unit, configured to write the adjustment parameter obtained by the third adjusting unit into the encoded code stream.