WO2009056071A1

WO2009056071A1 - A multiple description coding and decoding method, system and apparatus based on frame

Info

Publication number: WO2009056071A1
Application number: PCT/CN2008/072842
Authority: WO
Inventors: Lizhong Peng; Hualin Wan; Jun Zhang
Original assignee: Huawei Technologies Co., Ltd.
Priority date: 2007-10-26
Filing date: 2008-10-27
Publication date: 2009-05-07
Also published as: CN101420607B; CN101420607A

Abstract

A coding and decoding method based on frame is provided, the method includes: executing phase frame decomposing on an original image according to a determined phase matrix group 1 to form several descriptions; coding the descriptions formed respectively, the data being coded is sent to a decoding end; the row number of the matrix in the phase matrix group 1 is a half of that of the original image, and the line number is the same as that of the original image, the phase matrix group 1 is represented by:Wherein, bi,i+bi,i+1=1, the phase matrix group 1 includes 3 matrixes, which are represented by T1, T2 and T3, T1, T2 include a free parameter a (0<=a<=1). The embodiments of the invention also provide another kind of coding and decoding method and coding and decoding system, coding apparatus and decoding apparatus based on frame. The embodiments of the invention improve the quality of the rebuilt image.

Description

Framework-based multi-description codec method, system and device

The present invention relates to the field of codec technologies, and in particular, to a frame-based multiple description codec method, system and apparatus.

Background technique

In the field of Internet information transmission, network bandwidth fluctuations, noise interference, bandwidth changes, network congestion, etc., will cause packet loss, decoding error and delay phenomenon sent by the encoding end to the decoding end, which seriously affects the recovery of the decoding end. Image Quality. Therefore, codec technology needs to consider data errors and losses that occur in unreliable channels. Multi-description coding technology (MDC, Mul t iple Descr ipt ion Cod ing) is an effective method to reduce the impact of transmission errors, and is now a hot spot in image and video transmission research.

With the rapid development of the network, especially the wireless network, the research on MDC has gradually evolved from a theoretical to a practical MDC system. The MDC method has also become a hot spot in the field of image and video coding research. So far, MDC has no universal international standards, and there is still room for development and improvement.

The codec end of the MDC is independent. After the encoding end forms each description, the formed descriptions are encoded, and the encoded data is sent to the decoding end. The decoding end decodes the received data to obtain various descriptions, and then each Description for integration.

In MDC, the framework-based multi-description codec scheme is currently the hot research field. The frame-based multi-description codec method includes a multi-description codec method based on phase frame, and a multi-description codec method based on multi-wavelet frame. The following description will be respectively made by the prior art 1 and the prior art 2 respectively.

The prior art one is a multi-description codec method based on a phase frame. The method specifically includes: grouping the original image into two frames according to an odd frame and an even frame. Then, the two descriptions are encoded: 时 The time domain method is used to calculate the even frame prediction sequence and the odd frame prediction sequence, and the two motion vector sequences are respectively calculated, and then the prediction sequence and the generated two descriptions are respectively used as residuals, and finally Will The corresponding residual information, the motion vector sequence and the generated two descriptions are packed to obtain the encoded data. The encoded data is sent to the decoding end, and the decoding end decodes and then uses the time domain method to obtain the restored image according to the odd frame and the even frame.

The prior art has the following disadvantages: The method is not applicable to a case where the packet loss rate is high and the network environment is poor. When a certain frame is lost, the image information corresponding to the frame is lost, resulting in a poor image quality.

Summary of the invention

Embodiments of the present invention provide a frame-based multi-description codec method that can improve the quality of a restored image.

Embodiments of the present invention provide a frame-based multi-description codec system that improves the quality of recovered images.

Embodiments of the present invention provide a frame-based multiple description encoding apparatus that is capable of improving the quality of a restored image.

Embodiments of the present invention provide a frame-based multiple description decoding apparatus that is capable of improving the quality of a restored image.

A framework-based multiple description encoding and decoding method, the method comprising:

Performing phase frame decomposition on a pair of original images of the determined phase matrix group to form a plurality of descriptions;

Each of the formed descriptions is separately encoded, and the encoded data is sent to the decoding end; the number of rows of each matrix in the phase matrix group 1 is half of the original image, the number of columns is the same as the original image, and the phase matrix group is Expressed as:

Wherein, b _u + b _u+1 = i , the phase matrix group one comprises three matrices, denoted as T1, Τ2 and Τ3, and the T1 and Τ2 have a free parameter a (0<= a <=1) .

Multi-wavelet transforming the original image to obtain each sub-band of the low-frequency part and each sub-band of the high-frequency part; combining the sub-bands of the low-frequency part and the sub-bands of the high-frequency part to form a plurality of descriptions; The description is separately encoded, and the encoded data is sent to the decoding end.

A frame-based multi-description codec system, the system includes an encoding end and a decoding end; the encoding end is configured to perform phase frame decomposition on a pair of original images according to the determined phase matrix group to form a plurality of descriptions, The number of rows of each matrix in the phase matrix group 1 is half of the original image, and the number of columns is the same as the original image; each of the formed descriptions is separately encoded, and the encoded data is sent to the decoding end;

The decoding end is configured to receive the encoded data sent by the encoding end, perform inverse phase frame synthesis inverse transformation on the missing description according to the determined phase matrix group 2, and obtain a restored image; Obtained according to the phase matrix group one;

The number of rows of each matrix in the phase matrix group 1 is half of the original image, and the number of columns is the same as the original image, and the phase matrix group one is expressed as:

A frame-based multi-description codec system, the system comprising an encoding end and a decoding end; The encoding end is configured to perform multi-wavelet transform on the original image to obtain each sub-band of the low-frequency part and each sub-band of the high-frequency part; combining the sub-bands of the low-frequency part and the sub-bands of the high-frequency part to form a plurality of Decoding the plurality of descriptions, and transmitting the encoded data to the decoding end; the decoding end is configured to receive the encoded data sent by the encoding end, and decode the encoded data to obtain no Lost description; according to the description without loss, the subband of the low frequency part, and the subband of the high frequency part are obtained, and the two are combined to obtain a combined signal; the inverse wavelet transform is performed on the combined signal to obtain the recovered image.

A frame-based multiple description encoding device, the device comprising a plurality of description forming modules, an encoding module and a transmitting module;

The plurality of description forming modules are configured to perform phase frame decomposition on a pair of original images according to the determined phase matrix group, and form a plurality of descriptions and send the same to the encoding module, where the number of rows of each matrix in the phase matrix group is original The half of the image and the number of columns are the same as the original image; the number of rows of each matrix in the phase matrix group one is half of the original image, and the number of columns is the same as the original image, and the phase matrix group one is expressed as:

Wherein, b _u + b _u+1 = i , the phase matrix group one comprises three matrices, denoted as T1, Τ2 and Τ3, and the T1 and Τ2 have a free parameter a (0<= a <=1) ;

The encoding module is configured to separately code each formed description, and send the encoded data to the sending module;

The sending module is configured to send the encoded data to the decoding end.

A frame-based multiple description decoding device, the device includes a receiving module and a decoding module, and the receiving module is configured to receive the encoded data sent by the encoding end; The decoding module is configured to perform inverse phase transform synthesis on the missing description according to the determined phase matrix group 2 to obtain a restored image; and the phase matrix group 2 is obtained according to the phase matrix group, phase The number of rows of each matrix in matrix group one is half of the original image, and the number of columns is the same as the original image;

The phase matrix group one is expressed as:

The plurality of description forming modules are configured to perform multi-wavelet transform on the original image to obtain each sub-band of the low-frequency part and the sub-bands of the high-frequency part, and combine the sub-bands of the low-frequency part and the sub-bands of the high-frequency part, Forming multiple descriptions and sending them to the encoding module;

The encoding module is configured to separately encode the formed plurality of descriptions, and send the encoded data to the sending module;

The sending module is configured to send the encoded data to the decoding end.

A frame-based multiple description decoding device, the device includes a receiving module and a decoding module, and the receiving module is configured to receive the encoded data sent by the encoding end;

The decoding module is configured to decode the encoded data to obtain a description without loss; according to the description that is not lost, obtain a subband of a low frequency part, and a subband of a high frequency part, Combining, obtaining a combined signal; performing multi-wavelet inverse transform on the combined signal to obtain a restored image. As can be seen from the foregoing solution, the embodiment of the present invention performs phase frame decomposition on a pair of original images of the determined phase matrix group, forms a plurality of descriptions, and encodes the multiple descriptions; or performs multi-wavelet transform on the original image. The subbands of the low frequency portion and the subbands of the high frequency portion are obtained, and the subbands of the low frequency portion and the subbands of the high frequency portion are combined to form a plurality of descriptions, and the plurality of descriptions are encoded. The solution of the embodiment of the present invention provides a complete frame-based multi-description codec technology. Moreover, in the transmission process, if a certain frame data is lost, a higher quality image can be recovered according to other frames.

DRAWINGS

1 is an exemplary flowchart 1 of a frame-based multi-description codec method according to an embodiment of the present invention; FIG. 2 is a first flowchart of a multi-description codec method based on a frame according to an embodiment of the present invention; FIG. The five images obtained correspond to the corresponding images;

FIG. 4 is a second exemplary flow chart of a multi-description codec method based on a framework according to an embodiment of the present invention; FIG. 5 is a second flowchart of a multi-description codec method based on a frame according to an embodiment of the present invention; FIG. Image;

FIG. 7 is a third flowchart of a multi-description codec method based on a frame according to an embodiment of the present invention; FIG. 8 is a fourth flowchart of a multi-description codec method according to an embodiment of the present invention; FIG. FIG. 9b is a schematic structural diagram 2 of a frame-based multi-description codec system according to an embodiment of the present invention; FIG. 10 is a compression ratio obtained by using the method of FIG. 1 without losing description. - signal noise power ratio (PSNR, Power S igna l-to-Noi se Ra t io );

Figure 11 is a schematic diagram of the compression ratio - PSNR value obtained when the description of Figure 2 is lost;

Figure 12 is a schematic diagram showing the compression ratio - PSNR value obtained when the three methods in the first four descriptions are lost by the method of Figure 2;

Figure 13 is a schematic diagram showing the compression ratio - PSNR value obtained by the method of Figure 5 without losing the description; Figure 14 is a diagram showing the compression ratio - PSNR value obtained when the method of Figure 5 is lost. Figure

Figure 15 is a schematic diagram of the compression ratio - PSNR value obtained when the three descriptions are lost by the method of Figure 5;

Figure 16 is a diagram showing the number of frames - PSNR values obtained by the method of Figure 1 and the method of Figure 4 for a video sequence;

17 is a schematic diagram of a compression ratio - PSNR value obtained by using the method of FIG. 1 and the method of FIG. 4 when no packet is lost;

18 is a schematic diagram showing a compression ratio - PSNR value obtained by using the method of FIG. 1 and the method of FIG. 4 respectively;

FIG. 19 is a schematic diagram showing a compression ratio - PSNR value obtained by using the method of FIG. 1 and the method of FIG. 4 respectively when the description 2 is lost;

20 is a schematic diagram of a compression ratio - PSNR value obtained by using the method of FIG. 1 and the method of FIG. 4 respectively when the description is lost;

21 is a schematic diagram of a compression ratio - PSNR value obtained by using the method of FIG. 1 and the method of FIG. 4 respectively when the description is lost;

22 is a schematic diagram of a compression ratio - PSNR value obtained by using the method of FIG. 1 and the method of FIG. 4 when three descriptions are lost;

FIG. 23 is a schematic diagram showing the compression ratio - PSNR value when a description is lost by using the method of FIG. 1 for the video sequence, respectively;

Figure 24 is a schematic diagram of the compression ratio - PSNR value in the case where no double packet is generated by the method of Figure 4;

Figure 25 is a diagram showing the compression ratio - PSNR value for a description when a double description is generated using the method of Figure 4.

detailed description

In order to make the objects, the technical solutions and the advantages of the present invention more comprehensible, the present invention will be further described in detail below with reference to the embodiments and drawings. FIG. 1 is an exemplary flowchart 1 of a frame-based multiple description encoding and decoding method according to an embodiment of the present invention, where the method includes:

Step 101: Perform phase frame decomposition on a pair of original images of the determined phase matrix group to form a plurality of descriptions; wherein the number of rows of each matrix in the phase matrix group 1 is half of the original image, and the number of columns is the same as the original image.

Step 102: Code each formed description separately, and send the encoded data to the decoding end.

In the embodiment of the present invention, the phase matrix group one can be expressed as:

2 is a specific process example of a multi-description codec method based on a frame according to an embodiment of the present invention. This embodiment combines the descriptions of the multiple descriptions into five descriptions, and specifically describes the method of FIG. Includes the following steps:

Step 201: Determine a phase matrix group 1. The number of rows of each matrix in the phase matrix group 1 is half of the original image, and the number of columns is the same as the original image.

The following is an example in which the phase matrix group includes three matrices, which are T1, T2, and T3. The transposition indicating ^ is used to indicate the transposition of T ₂ and 7f is the transposition of Τ ₃ . . The expressions of Tl, Τ2, and Τ3 can be: a,\-a,0, 0, 0, 0, ···, 0, 0

0, 0, a,l-a, 0, 0, ···, 0, 0

τ 0, 0, 0, 0, a, l-a, ···, 0, 0

0, 0, 0, 0, 0, 0, ···, α,\-α,)

--\,2-α, 0, 0, 0, 0, ···, 0, 0

0, 0, α-\,2-α, 0, 0, ···, 0, 0

Τ 0, 0, 0, 0, a-Vl-α ··· 0, 0

0, 0, 0, 0, 0, 0, ■•■ a- 2-α

0, 0, 0, 0, ···,(), 0

0, 0, -,-, 0, 0, 0, 0

twenty two

1 1

0, 0, 0, 0. 0, 0 (where ≤α≤1,α≠- )

twenty two'

0, 0, 0, 0, 0, 0,

Step 202: Perform phase frame decomposition on a pair of original images of the determined phase matrix group to form a plurality of descriptions.

Taking the phase matrix group one in step 201 as T1, Τ2, and Τ3 as an example, the five descriptions formed in this step are respectively represented as: Sl, S2, S3, S4, and S5, and then: then: ^S i ^{= T} i ^ST i ^T , ^S i ^{= T} i ^ST i ^T ,

^ ^{= 1} 2^ι , 2丄丄1 , = ^ ₃ , where S represents the original image. The image representation of the five descriptions is shown in Figure 3.

Step 203: separately code each formed description, and send the encoded data to the decoding end.

In this step, the coding may be implemented by using an existing coding technique. For example, a joint image expert group (Jpeg, Joint Photographic Experts Group) 2000 technology may be used for images, and H.264 technology may be used for each frame image of the video. Wait.

Step 204: The decoding end decodes the received data to obtain a description that is not lost. In this step, if the description obtained after decoding is not lost, there are five descriptions. In the actual transmission process, data loss often occurs due to network congestion, etc., that is, some descriptions are often lost.

The decoding uses a technique corresponding to the encoding in step 203, such as Jpeg2000 technology for images, and H.264 technology or the like for each frame image of the existing video. Based on the description that is not lost, it is possible to determine which descriptions have been lost and the number of missing descriptions.

Step 205: Determine the phase matrix group 2, and according to the determined phase matrix group 2, perform phase frame synthesis inverse transformation on the description without loss, and obtain the restored image.

If the description of the loss is one, the step 205 is replaced by: the decoding end interpolates the description that is not lost, and obtains the restored image.

The phase matrix group 2 is obtained according to the phase matrix group 1. The number of rows of each matrix in the phase matrix group 2 is half of the original image, and the number of columns is the same as the original image. Similarly, in step 201, the phase matrix group one is T ₂ and T ₃ , and the phase matrix group two includes two matrices, which are 'and Γ ₂ ', 'obtained by the 2-a and 1-a exchange positions of 2 , T; obtained by 7; a and a - 1 exchange positions, respectively expressed as:

2-α,\-α, 0, 0, 0, 0, ···, 0, 0

0, 0, 2-a, l-a, 0, 0, ···, 0, 0

Where T;= 0, 0, 0, 0, 2-αΛ-α··· 0, 0

0, 0, 0, 0, 0, 0, 2 - a, 1 - a a -I a, 0, 0, 0, 0, , 0, 0

0, 0, a-l a, 0, 0, , 0, 0

0, 0, 0, 0, a-l, a, , 0, 0 (where ≤a≤l,a≠- )

0, 0, 0, 0, 0. al, a where T; ^T denotes the transposition of Τ; Τ; ^τ denotes the transposition of Γ ₂ '. This step is described in two cases: 1) no description of the lost condition; 2) a description of the lost condition.

1) There is no description of the lost situation. At this point, the recovered image is represented as:

, where S' represents the restored image. The image recovered without loss of description is identical to the original image.

2) There is a description of the lost situation. At this time, it is divided into three cases: a, missing 1 description; b, losing 2 or 3 descriptions; c, losing 4 descriptions. The following description will be respectively made.

a. If the missing description is due to; the average information of the first four descriptions, this time does not affect the image effect obtained by using the first four descriptions, and the recovered image is restored as in the case where there is no missing description. The image is represented as:

S' = T; ^T ST; + T; ^T S ₂ T + T %T; + T ^T SJ . If the description of the loss is description s, description 2 description s or description s, then ^{4x is} subtracted from the result of the 3 descriptions that are not lost as the missing description, and the restored image is expressed as: S' = T; ^T ST; + T; ^T S ₂ T + T %T; + T ^T SJ . b. If the description of the loss is 2 or 3 descriptions, it may include the 5th description, or may not include the 5th description. If the fifth description is lost, inter-row or inter-column mean interpolation is performed on the description adjacent to the description, respectively, resulting in two or three missing descriptions, ie, two adjacent rows of the description adjacent to the description or The column elements are added and averaged as a row or column corresponding to the missing description. Thus, the description of the interpolation process is respectively taken as two or three missing descriptions, and the restored image is represented as:

S' = T; ^T ST; + T; ^T S ₂ T + T %T; + T ^T SJ

If the description of the loss is 2 or 3, and the 5th description is not included, the above-mentioned inter-row or inter-column mean interpolation processing method may be used, or ^4x may be subtracted from the description without loss to obtain a subtraction result. Dividing the subtraction result by the number of missing descriptions, and dividing the result of the division as each of the missing descriptions; the recovered image is represented as:

S' = T; ^T ST; + T; ^T S ₂ T + T %T; + T ^T SJ c. If the description of the loss is four, the decoding end performs interpolation processing on the description that is not lost, and obtains the restored image.

For a video sequence, a video sequence is composed of a plurality of image groups (GOP, Group Of Picture), each GOP is composed of a plurality of image frames, and the image shown in Fig. 1 is processed for each image frame in the GOP.

Referring to FIG. 4, which is a second exemplary flowchart of a frame-based multiple description codec method according to an embodiment of the present invention, the method includes the following steps:

Step 401, performing multi-wavelet transform on the original image to obtain each sub-band of the low-frequency part and each sub-band of the high-frequency part.

Step 402: Combine each subband of the low frequency portion obtained in step 401 with each subband of the high frequency portion to form a plurality of descriptions.

Step 403: Encode the multiple descriptions, and send the encoded data to the decoding end. Figure 4 is exemplified below by Figures 5, 6, 7, and 8.

Referring to FIG. 5, it is a specific process example of a multi-description codec method based on a frame according to an embodiment of the present invention. The method includes the following steps:

Step 501: Perform multi-wavelet transform on the original image to obtain each subband of the low frequency and each subband of the high frequency.

This step can specifically include:

Step 5011: Preprocess the one-dimensional discrete initial image to obtain an original image that needs to be multi-decoded. The one-dimensional discrete image is represented as S, S = (", where n is an integer. The preprocessing is: transforming a one-dimensional discrete image into a 2 x 1 vector group signal, and the specific implementation method can be Yes, two adjacent ones of the one-dimensional discrete image are put together to form a vector of 2 X 1 , and the preprocessed signal is called an original image for performing multiple description encoding and decoding, which is S1 , S'

. Step 5012: Perform multi-wavelet transform on the obtained original image to obtain low frequency sub-bands and high-frequency sub-bands. The step may specifically include: 对 low-passing the original image with a low-pass filter Through the filtering, the low-frequency coefficient ⁵ - -i ^l l ^L "'J ^j , _ "'^, the vector with the low-frequency coefficient of 1 χ 1 is obtained; 釆 the high-pass filter is used to high-pass filter the original image to obtain the high-frequency coefficient.

The high frequency coefficient is a vector of 2 x 1; each subband of the low frequency and each subband of the high frequency are obtained according to the low frequency coefficient and the high frequency coefficient.

The selection of the low pass filter and the high pass filter has a large impact on the quality of the image recovered in step 508. Here, the low-pass filter and the high-pass filter have a weight of 2, and both are 2 x 2 matrices. The low-pass filter and the high-pass filter can be a set of orthogonal balanced multi-wavelet frames. Corresponding filter, where the low pass filter is expressed as: L =, l. , l _x ,...,l _N _ _x ,, The high-pass filter is expressed as: , 'Λ are all 2 x 2 matrices. Here is a set of better-performing low-pass filters and high-pass filters:

_( 0.8279 0.5117 θ.1954 -0.1208 —!

Low pass filter: -0.1208 0.1954j ⁺ to.5H7. .8279 / +

(-0.51Π 0.8279 -0.1208 -0.1954 —!

High-pass filter: ) -0.1954 -0.120 ₈ J ⁺ t 0.8279 -0.5117/ ·. It is assumed that after performing step 501, 16 sub-bands as shown in Table 1 are obtained. Among the 16 sub-bands, L1L1, L1L2, L2L1, L2L2 are low frequency sub-bands, and the remaining 12 sub-bands are high frequency sub-bands. The images of the respective sub-bands corresponding to Table 1 are as shown in FIG. 6. See Figure 6 for the 16 self-contained corresponding images in Table 1.

Table 1 16 subbands after multi-wavelet transform

Step 502, calculating energy of each sub-band in the low frequency part and the high frequency part, respectively.

In step 503, one sub-band with similar energy in each sub-band of the low-frequency part and the high-frequency part is respectively grouped, and one of the groups is selected to be combined into one description to form a plurality of descriptions. The step specifically includes: grouping energy by low frequency part and high frequency part respectively, grouping two sub-bands with similar energy into one group, extracting one sub-band in each group, and extracting respectively from the low frequency part and the high frequency part. Each sub-band is put together to form a description of the multiple descriptions; a plurality of descriptions are formed as described above.

The method of grouping two sub-bands with low energy in each sub-band of the low-frequency part and the high-frequency part has various implementation manners. For example, taking each sub-band of the high-frequency part as an example, the high-frequency part can be firstly used. The two sub-bands with the largest energy in the band are grouped together, and then the two sub-bands with the largest energy among the remaining sub-bands are grouped together until all combinations of the high-frequency partial energy close sub-bands are obtained.

For the video sequence, the energy calculation of step 502 and the grouping of step 503 are performed for the first image frame in each GP to form a multi-description. For the remaining image frames in the same GP, after performing step 501, when performing steps 502 and 503, if the efficiency is increased, the energy of each sub-band may not be calculated, that is, 502 is not executed, and step 503 is directly used. A method of grouping image sub-frames to form a plurality of descriptions.

The following takes the 16 sub-bands shown in Table 1 after performing step 501 as an example, and exemplifies step 503:

For the low frequency part, by calculating the energy of each sub-band, it is found that the energy of L1L1 and L2L1 are similar, and L1L2 is similar to L2L2. Thus, the fractional frequency can be divided into two groups: one is L1L1, L2L1, and the other is L1L2 and L2L2. One sub-band is extracted from each of the two groups to obtain four parts: L1L1 L1L2; L2L1 L1L2; L2L1 L2L2; L1L1 L2L2.

For the high-frequency part, the energy of each sub-band calculated by the calculation is also divided into three groups according to the principle of similar energy: one group is L1 H1, L1H2, L2H1 and L2H2; - the group is H1L1, H1L2, H2L1 and H2L2; The groups are H1H1, H1H2, H2H1 and H2H2; each group has 4 subbands. Thus, one sub-band is extracted from the three groups to obtain four parts: L1H1 H1L1 H1H1; L1H2 H1L2 H1 H2; L2H1 H2L1 H2H1; L2H2 H2L2 H2H2 ₀

The four portions extracted from the low frequency portion and the four portions extracted from the high frequency portion are respectively combined to obtain four descriptions, each of which includes a low frequency portion of 1 /2 and a high frequency portion of 1/4. The four descriptions obtained after the merger can have multiple combinations. The following four descriptions are one of the cases:

Description 1: L1L1 L 1L2 L1 H1 H1L 1 H1 H1

Description 2: L1L2 L2L1 L1 H2 H1L2 H1 H2

Description three: L2L1 L2L2 L2H1 H2L 1 H2H1

Description 4: L2L2 L 1L1 L2H2 H2L2 H2H2

It can be seen that four of the above four descriptions are repeated, which is four more than the 16 sub-bands obtained in step 501, that is, the frame redundancy is 5/4. In fact, the four low-frequency sub-bands that are sensitive to the visual can be fully reconstructed, and the missing 1 / 4 high-frequency part has little effect on the visual effect.

Step 504, for each description, the low frequency sub-bands missing from the description are added from other descriptions, and the respective descriptions after the addition are obtained.

Step 505: Perform multi-wavelet inverse transform on each of the added descriptions to obtain an image after multi-wavelet transform.

^ W is the low frequency coefficient and the high frequency coefficient obtained in step 501, respectively.

Step 506: Perform image coding on the inverse wavelet transformed image separately, and transmit the image encoded data to the decoding end.

In this step, the image coding may be implemented by using an existing coding technique, for example, Jpeg2000 technology may be used for the image, and each frame image in the video may be implemented by using H.264 technology or the like.

Step 507: The decoding end decodes the received data and performs multi-wavelet transform to obtain an added description without loss.

Each post-addition description is the added description described in step 504, including three high frequency sub-bands and four low frequency sub-bands.

The decoding uses a technique corresponding to the encoding in step 203, such as J pe g 2000 technology for images, and H.264 technology or the like for each frame image of the existing video. Step 508: The decoding end obtains all subbands of the low frequency part and subbands of the high frequency part according to the added description of the missing, and the two combine to obtain a combined signal; and performs multi-wavelet inverse on the combined signal. Transform to get the restored image.

The following steps are described in two cases: 1) the received data is not lost; 2) the received data is lost.

1) The received data is not lost. At this time, the combined signal is all subbands of the high frequency portion and all subbands of the low frequency portion shown in Table 1, and the recovered image is the signal of the original image.

The data received by the decoding end is not lost. After the step 504, the added descriptions are obtained. Since the low frequency part of each description part includes the complete low frequency subband, the low frequency part of one of the description parts can be selected; for each description Part of the high frequency part, each description part contains different high frequency sub-bands, each independently, combining the high frequency sub-bands contained in each added description part; The sub-bands are combined with all the sub-bands of the selected low-frequency portion to obtain a combined signal, which is subjected to multi-wavelet inverse transform to recover the original image signal.

2) The received data is lost. At this time, the combined signal is incomplete, including all sub-bands of the low-frequency part and partial sub-bands of the high-frequency part. At this time, the respective high-frequency sub-bands lost in the combined signal must be supplemented, for example, "0" or "1" ".

The following is a detailed description of 2) in the four described examples listed in step 503.

Since the low-frequency part of each added description includes the complete low-frequency sub-band, all the low-frequency sub-bands are obtained by any added description; if one of the added descriptions is lost, the corresponding high-frequency sub-band is lost. For example, when the description after the addition is lost, the corresponding high frequency sub-bands L1H1, H1L1 and H1H1 are lost. When the description 2 is lost, the corresponding loss is L1 H2, H1L2 and H1H2, and the description after the loss is added. The corresponding loss is L2H1, H2L1, and H2H1. When the description of the added four is lost, the corresponding losses are L2H2, H2L2, and H2H2. That is, which of the added descriptions is lost, corresponding to the loss of the corresponding high frequency sub-band. In this case, all low frequency subbands and no The lost high frequency sub-bands are combined, and the missing high-frequency sub-bands are supplemented. The supplementary processing may be "0" or "1", etc., and the combined signals after the complementary processing are obtained, and the combined signals after the complementary processing are added. Do more wavelet inverse transform to get the recovered image.

In fact, in the extreme case, that is, only one post-addition description is left after the loss, at this time, the high-frequency sub-bands included in the other added descriptions are supplemented to obtain the combined signal after the addition. The multi-wavelet inverse transform is performed on the combined signal after the supplementary processing to obtain the restored image. Since the image obtained by multi-wavelet inverse transformation is only based on all the low-frequency sub-bands that are sensitive to the vision, the original image can be approximated. Thus, even in the case where only one added description is left, the original image can be largely restored.

For the video sequence, the steps of Figure 5 can be performed on the frame-by-frame image.

Referring to FIG. 7, which is a specific process example of a multi-description codec method based on a frame according to an embodiment of the present invention, the method includes the following steps:

Step 701 is the same as step 501.

Step 702, combining the sub-bands of the low-frequency part obtained in step 701 and the sub-bands of the high-frequency part to form a plurality of descriptions.

It is assumed that after performing step 701, 16 sub-bands shown in Table 1 are obtained, and the combination described in this step can be implemented as needed, for example, 12 high-frequency sub-bands can be equally divided into 4 parts of low frequency; For the four parts of the high frequency, each part includes two different low frequency sub-bands, and each of the eight sub-bands included in the four parts combined by the low-frequency sub-band appears twice; The four parts and the four parts of the high frequency are combined to obtain four descriptions.

Step 703: Code each formed description separately, and send the encoded data to the decoding end.

In this step, the coding may be implemented by using an existing coding technique, such as Jpeg2000 technology for images, and H.264 technology for each frame of video.

Step 704: The decoding end decodes the received encoded data to obtain a description that is not lost. The decoding can be implemented by using existing decoding technologies, such as Jpeg2000 technology for images, and H.264 technology for each frame of video.

Step 705: Obtain a subband of the low frequency part and a subband of the high frequency part according to the description without loss, and combine the two to obtain a combined signal; perform multi-wavelet inverse transform on the combined signal to obtain the restored image.

Referring to FIG. 8, which is a specific process example of a multi-description codec method based on a frame according to an embodiment of the present invention, the method includes the following steps:

Steps 801 - 803 are the same as steps 501 - 503.

Steps 804 - 806 are the same as steps 703-705.

Figures 5, 6, 7 and 8 are specific illustrations of Figure 4. The method of FIG. 4 of the embodiment of the present invention can also be applied to the case of generating a double description. At this time, it is assumed that after the step 401, the 16 sub-bands shown in Table 1 are also obtained. After the step 402 is performed, two descriptions are formed, which are:

Description 1: L1L1 L1L2 L1 H1 H1L1 H1H1 L1H2 H1L2 H1H2

Description 2: L2L1 L2L2 L2H1 H2L1 H2H1 L2H2 H2L2 H2H2

When step 403 is performed, the low-frequency sub-bands L2L1 and L2L2 are added from the description 2 to the description one, and the low-frequency sub-bands L1L1 and L1L2 are added from the description one to the description two, so that the added description 1 and the added force port are obtained. Description 2, for:

Description after the addition: L1L1 L1L2 L2L1 L2L2 L1H1 H1L1 H1 H1 L1H2 H1L2 H1H2 , Description 2 after addition: L1L1 L1L2 L2L1 L2L2 L2H1 H2L1 H2H1 L2H2 H2L2 H2H2 ₀ Further, after the addition, the first and second inverse wavelet transforms are performed. , get two images of the same size as the original image.

Then, the two images after the inverse wavelet transform are separately encoded, and the encoded data is transmitted to the decoding end. Then, the processing of steps 704 and 705 can be performed.

The embodiment of the present invention further provides a frame-based codec system. As shown in FIG. 9a, the system includes an encoding end and a decoding end.

The encoding end is configured to perform a phase frame on a pair of original images according to the determined phase matrix group Decomposing, forming a plurality of descriptions, wherein the number of rows of each matrix in the phase matrix group 1 is half of the original image, and the number of columns is the same as the original image; each of the formed descriptions is separately encoded, and the encoded data is sent to the decoding end ;

The decoding end is configured to receive the encoded data sent by the encoding end, perform inverse phase frame synthesis inverse transformation on the missing description according to the determined phase matrix group 2, and obtain a restored image; Obtained according to the phase matrix group one.

Optionally, the encoding end includes multiple description forming modules, an encoding module, and a sending module. The plurality of description forming modules are configured to perform phase frame decomposition on a pair of original images according to the determined phase matrix group, and form a plurality of descriptions and send the same to the encoding module, where the number of rows of each matrix in the phase matrix group is original Half of the image, the number of columns is the same as the original image;

The sending module is configured to send the encoded data to the decoding end.

Optionally, the plurality of description forming modules include a phase matrix group-determining module and a plurality of description forming sub-modules;

The phase matrix group determining module is configured to determine, after the phase matrix group is sent to the plurality of description forming submodules, the phase matrix group 1 includes three matrices, represented as T1, Τ2, and Τ3; The plurality of description forming sub-modules are configured to perform phase frame decomposition on a pair of original images according to the received phase matrix group, and form five descriptions and send the same to the encoding module, where the five descriptions are: 1 = Λ Λ , ^ώ 2 = Α ^{ώ /} 2 , ^ύ 3 ⁼ 2 ^ύ 1 , ^4 ^{= Ι} 2^ ^Ι 2 , ^ύ 5 = ^ , where S represents the original image, SI represents the description 1, S2 represents the description 2, S3 represents the description 3, S4 represents Description 4, S5 denotes a description of five, denotes the transposition of ^, represents the transposition of T ₂ , and represents the transposition of Τ ₃ .

Optionally, the decoding end includes a receiving module and a decoding module.

The receiving module is configured to receive the encoded data sent by the encoding end;

The decoding module is configured to decode the encoded data according to the determined phase matrix group 2, obtain a description without loss, and perform phase frame synthesis inversion on the description without loss. In other words, the restored image is obtained; the phase matrix group 2 is obtained according to the phase matrix group one. Optionally, the decoding module includes a decoding submodule and an image recovery submodule;

The decoding sub-module is configured to decode the encoded data, and obtain a description that is not lost, and then transmit the description to the image recovery sub-module; if no description is lost, the decoded description includes description 1, description 2, description 3. Description 4 and Description 5, respectively denoted as n ^ST i ^T , ^ύ 2 ^{= 1} ^ ¹ 2 , ^Λ 3 ^{= 1} 2^ ¹ \ , ύ ₄ =ι ₂ ύΐ ₂ , ₅ =ι ₃ ύΐ ₃ , where § denotes the original image, SI denotes the description 1, S2 denotes the description 2, S3 denotes the description 3, S4 denotes the description 4, S5 denotes the description 5, Tl, Τ2 and Τ3 denote the three matrixes of the phase matrix group one, the transposition is represented , indicating transposition of Τ ₂ , indicating transposition of Τ ₃ ; description of the absence of loss is description 1, description 2, description 3, description 4, and description 5; or, the description that is not lost is description 1, description Second, description three and description four;

The image restoration sub-module is configured to perform inverse phase frame synthesis inverse transformation on the non-lost description according to the determined phase matrix group 2, to obtain a restored image, which is expressed as:

S^TjSJl +T\ ^T S ₂ T ₂ +T ₂ ^T S ₄ T ₂ , where S' represents the restored image, ^ and represents a matrix contained in the phase matrix group 2, and the transposition represented by ^, ^The phase matrix group 2 is obtained according to the phase matrix group one, and the number of rows of each matrix in the phase matrix group one is half of the original image, and the number of columns is the same as the original image.

Optionally, the decoding module includes a decoding submodule and an image recovery submodule;

The decoding sub-module is configured to decode the encoded data, and obtain a description that is not lost, and then transmit the description to the image recovery sub-module; if no description is lost, the decoded description includes description 1, description 2, description 3. Description 4 and Description 5, respectively denoted as n ^ST i ^T , ^ύ 2 ^{= 1} ^ ¹ 2 , ^Λ 3 ^{= 1} 2^ ¹ \ , ύ ₄ =ι ₂ ύΐ ₂ , ₅ =ι ₃ ύΐ ₃ , where § denotes the original image, SI denotes the description 1, S2 denotes the description 2, S3 denotes the description 3, S4 denotes the description 4, S5 denotes the description 5, Tl, Τ2 and Τ3 denote the three matrixes of the phase matrix group one, the transposition is represented , Representing the transposition of T ₂ , indicating the transposition of Τ ₃ ; the description of the absence of loss is description 1, description 2, description 3 or description 4;

The image restoration sub-module is configured to subtract ^{4 χ} results of the three descriptions other than the description 5 that are not lost as the lost description, and the obtained restored image is represented as:

S^ TjSJl + T\ ^T S ₂ T ₂ + T ₂ ^T S ₄ T ₂ , where S' denotes the restored image, τ; and ⁷ ^ denotes two matrices of the phase matrix group two, representing the turn The ^TT represents the transposition of ^τ ; the phase matrix group 2 is obtained according to the phase matrix group one, and the number of rows of each matrix in the phase matrix group one is half of the original image, and the number of columns is the same as the original image.

Another embodiment of the present invention provides another frame-based codec system. As shown in FIG. 9b, the system includes an encoding end and a decoding end.

The encoding end is configured to perform multi-wavelet transform on the original image to obtain each sub-band of the low-frequency part and each sub-band of the high-frequency part; combining the sub-bands of the low-frequency part and the sub-bands of the high-frequency part to form a plurality of Decoding the plurality of descriptions, and transmitting the encoded data to the decoding end; the decoding end is configured to receive the encoded data sent by the encoding end, and decode the encoded data to obtain no Lost description; according to the description without loss, the subband of the low frequency part, and the subband of the high frequency part are obtained, and the two are combined to obtain a combined signal; the inverse wavelet transform is performed on the combined signal to obtain the recovered image.

Optionally, the encoding end includes multiple description forming modules, an encoding module, and a sending module. The plurality of description forming modules are configured to perform multi-wavelet transform on the original image to obtain each sub-band of the low-frequency part and the sub-bands of the high-frequency part, and combine the sub-bands of the low-frequency part and the sub-bands of the high-frequency part, Forming multiple descriptions and sending them to the encoding module;

The encoding module is configured to encode the formed multiple descriptions, and send the encoded data to the sending module;

The sending module is configured to send the encoded data to the decoding end.

Optionally, the plurality of description forming modules include a multi-wavelet transform sub-module and a multi-description forming sub- Module.

The multi-wavelet transform sub-module is configured to perform multi-wavelet transform on the original image to obtain each sub-band of the low-frequency part and each sub-band of the high-frequency part, and send the obtained sub-bands of the low-frequency part and the sub-bands of the high-frequency part. Forming a sub-module for multiple descriptions;

The multiple description forming sub-module is configured to respectively combine the obtained sub-bands of the low-frequency portion and the two sub-bands of the high-frequency portion into two groups, and select one of each group to form a description. The plurality of descriptions are formed.

Optionally, the encoding module includes an encoding sub-module, which is used to add the low-frequency sub-bands missing from the descriptions to other descriptions, obtain the added descriptions, and perform multi-wavelet inverse transform on each of the added descriptions. Obtaining an image after multi-wavelet transform; respectively, performing image coding on the inverse wavelet transform image, and the image encoded data is the encoded data.

The experimental results of the embodiment of the present invention will be described below with reference to Figs.

Referring to Fig. 10, for the method of Fig. 2, the compression ratio - PSNR value obtained without losing the description is shown, which shows the image PSNR value recovered without losing the description at different compression levels. It can be seen that the PSNR values are high at different compression levels, that is, the image quality recovered at different compression levels is high.

Referring to FIG. 1 1 , a schematic diagram of a compression ratio-PSNR value obtained when a description is lost by using the method of FIG. 2, which shows the loss of description 1, description 2, description 3 or description 4 at different degrees of compression, respectively. The recovered image PSNR value. In the figure, the broken line with a short thick line indicates the compression ratio - PSNR value of the missing description one, the broken line with the triangle indicates the compression ratio - PSNR value of the missing description 2, and the dotted line of the diamond indicates the compression ratio of the missing description three. - PSNR value, with a solid line with a fork indicating the compression ratio - PSNR value of the missing description four, which are close to each other and almost coincident. It can be seen that the PSNR value is high when a description is lost at different degrees of compression, that is, the recovered image quality is high.

Referring to FIG. 12, a schematic diagram of the compression ratio-PSNR value obtained when the three descriptions of the first four descriptions are lost by the method of FIG. 2, which shows the remaining description in the first four descriptions at different compression levels. First, the remaining four descriptions in the first four descriptions, the remaining four descriptions in the first four descriptions or the remaining P picture values in the first four descriptions. In the figure, the solid line with a diamond shape indicates the compression ratio - PSNR value of the first description in the first four descriptions, and the dotted line with a square indicates the compression ratio - PSNR value of the remaining two descriptions in the first four descriptions. The dotted line of the triangle indicates the compression ratio - PSNR value of the remaining three descriptions in the first four descriptions. The cross-hatched line indicates the compression ratio - PSNR value of the remaining four descriptions in the first four descriptions. The four lines are close to each other. Almost coincident. It can be seen that the PSNR values are high when three of the first four descriptions are lost at different degrees of compression, that is, the recovered image quality is high.

Referring to Fig. 13, the compression ratio - PSNR value obtained by the method of Fig. 5 without loss of description is shown, which shows the image PSNR value recovered when the description is not lost at different degrees of compression. It can be seen that the PSNR values are high at different compression levels, that is, the image quality recovered at different compression levels is high.

Referring to FIG. 14, a schematic diagram of a compression ratio-PSNR value obtained when a description is lost by using the method of FIG. 5, which shows that the descriptions are respectively lost when describing one, two, three or four at different degrees of compression. The resulting image PSNR value. In the figure, the solid line with a diamond shape indicates the compression ratio - PSNR value of the missing description one, and the broken line with a short thick line indicates the compression ratio - PSNR value of the missing description two, and the dotted line indicates the compression of the missing description three. The ratio - PSNR value, with the crossed dashed line, shows the compression ratio - PSNR value of the missing description four, which are close to each other and almost coincident. It can be seen that the PSNR value is high when a description is lost at different degrees of compression, that is, the recovered image quality is high.

Referring to FIG. 15, a schematic diagram of the compression ratio-PSNR value obtained when the three descriptions are lost by using the method of FIG. 5, which shows that the description is left at different degrees of compression, the remaining description is two, and the remaining description is three. Or the PSNR value of the image recovered separately at the time of four is left. In the figure, the dotted line with a diamond shape indicates the compression ratio - PSNR value of the remaining description one, and the dotted line with a square indicates the compression ratio - PSNR value of the remaining description two, and the dotted line of the triangle indicates the compression of the remaining description three. The ratio - PSNR value, with the dotted dashed line, shows the compression ratio - PSNR value of the remaining four, which are very close together and almost coincident. It can be seen that the PSNR values are high when three descriptions are lost at different compression levels, that is, the recovered image quality is high.

Referring to FIG. 16, which is a schematic diagram of the number of frames-PSNR values obtained by the method of FIG. 1 and the method of FIG. 4 for a video sequence, the figure shows that for different frame numbers, the phase frame method is used to generate 3 5, the multi-wavelet frame method is used to generate the corresponding PSNR values in the description of 4 and 2. The method of FIG. 1 is also a multi-description codec method based on a phase frame, and the method of FIG. 4 is a multi-description codec method based on a multi-wavelet frame. Here, the method of FIG. 1 is simply referred to as a phase frame method, and the method of FIG. 4 is used. Referred to as the multi-wavelet frame method. In the figure, the solid line with a square indicates the compression ratio - PSNR value when the code side of the phase frame method is used to generate 5 description, and the solid line of the diamond shape indicates the compression ratio when the code side of the phase frame method is used to generate 3 descriptions - The PSNR value is represented by the dotted line of the triangle, and the compression ratio - PSNR value when the code side is generated by the multi-wavelet frame method is represented by the dashed line. The dotted line of the short thick line is used to represent the compression when the code side generates the 2 description by the multi-wavelet frame method. Ratio - PSNR value, these four lines are very close, and some parts are overlapped. It can be seen that for a video sequence, the PSNR value is high in the case of different frames, that is, the recovered image quality is high.

Referring to FIG. 17, a compression ratio-PSNR value obtained by using the method of FIG. 1 and the method of FIG. 4 respectively when no packet is lost. In the figure, the solid line with a diamond shape is used to represent the compression ratio - PSNR value obtained by the multi-wavelet frame method, and the squared dashed line is used to represent the compression ratio - PSNR value obtained by the phase frame method. It can be seen that the PSNR value obtained by the method of FIG. 1 is higher than the PSNR value obtained by the method of FIG. 4, that is, the image quality ratio recovered by the method of FIG. 1 is recovered by the method of FIG. The image quality is high.

Referring to Fig. 18, a schematic diagram of the compression ratio - PSNR value obtained by using the method of Fig. 1 and the method of Fig. 4, respectively, for the loss of description.

Referring to Fig. 19, a schematic diagram of the compression ratio - PSNR value obtained by using the method of Fig. 1 and the method of Fig. 4, respectively, for the loss of the description.

Referring to FIG. 20, a schematic diagram of the compression ratio-PSNR value obtained by using the method of FIG. 1 and the method of FIG. 4 respectively for the description of the missing three. Referring to FIG. 21, a schematic diagram of the compression ratio-PSNR value obtained by using the method of FIG. 1 and the method of FIG. 4 respectively for the description of the missing four.

Referring to Fig. 22, a compression ratio - PSNR value obtained by subtracting the three descriptions using the method of Fig. 1 and the method of Fig. 4, respectively.

In Fig. 18 to Fig. 22, the corresponding compression ratio - PSNR value obtained by using the multi-wavelet frame method is represented by a solid line with a diamond shape, and the corresponding compression ratio obtained by the phase frame method is represented by a solid line of a square - PSNR value.

Referring to Figure 23, for the video sequence, the two methods are lost using the method of Figure 1, respectively, and the method of Figure 4 is lost with a description of the compression ratio - PSNR value. The figure shows the compression ratio - PSNR value in the case where the description of the code side of the phase frame method is 5 and the data received by the decoding side is lost, and the solid line indicates that the code side is generated by the multi-wavelet frame method. 4 Describes the data received by the decoder and loses the compression ratio - PSNR value in the case of a description. The two lines are close together and some are overlapped.

As can be seen from the comparison of FIG. 17-23, the method of FIG. 1 and the method of FIG. 4 are also a multi-description codec method of the phase frame and a multi-description codec method based on the multi-wavelet frame, both of which have Advantages and disadvantages, when the compression ratio is not very large, the multi-description codec method based on phase frame is better than the multi-description codec method based on multi-wavelet frame; in the case of large compression, multi-description based on multi-wavelet frame The codec method is superior to the multi-description codec based on the phase frame.

See Figure 24 for a plot of the compression ratio - PSNR value for the case where no double packet is generated when the dual description is generated using the method of Figure 4. The figure shows the image PSNR values recovered separately without loss of description at different levels of compression. It can be seen that the PSNR values are high at different compression levels, that is, the image quality recovered at different compression levels is high.

Referring to FIG. 25, a schematic diagram of the compression ratio-PSNR value in the case of losing a description when generating the double description by the method of FIG. The figure shows the image PSNR values recovered separately when a description is lost at different degrees of compression. It can be seen that the PSNR value is high when a description is lost at different degrees of compression, that is, the recovered image quality is high. In addition to the case where the above description refers to five descriptions, four descriptions, and one description, the embodiment of the present invention can be applied to a plurality of cases described as other numbers, such as three descriptions. , 6 descriptions or more described cases. Not here - enumeration.

The embodiment of the present invention performs phase frame decomposition on a pair of original images of the determined phase matrix group to form a plurality of descriptions, and encodes the multiple descriptions; or, performs multi-wavelet transform on the original image to obtain each subband of the low frequency portion. Each sub-band of the high-frequency portion; each sub-band of the low-frequency portion and each sub-band of the high-frequency portion are combined to form a plurality of descriptions, and the plurality of descriptions are encoded. In the scheme of the embodiment of the present invention, if a certain frame data is lost during transmission, the image with high quality can still be recovered according to other frames. Moreover, the embodiment of the present invention provides a perfect multi-description codec technology based on multi-wavelet frame.

The above described specific embodiments of the present invention are further described in detail, and are intended to be illustrative of the embodiments of the present invention. The scope of the protection, any modifications, equivalents, improvements, etc., made within the spirit and scope of the invention are intended to be included within the scope of the invention.

Claims

Claim

A frame-based multi-description codec method, the method comprising: performing phase frame decomposition on a pair of original images of a determined phase matrix group to form a plurality of descriptions; respectively coding each formed description And sending the encoded data to the decoding end; the number of rows of each matrix in the phase matrix group 1 is half of the original image, the number of columns is the same as the original image, and the phase matrix group one is expressed as:

2. The method according to claim 1, wherein after the encoding the data is sent to the decoding end, the method comprises:

Decoding, the decoding end decodes the encoded data to obtain a description that is not lost;

The decoding end performs inverse phase frame synthesis inverse transformation on the missing phase description according to the determined phase matrix group 2 to obtain a restored image; the phase matrix group 2 is obtained according to the phase matrix group one, and each of the phase matrix groups 2 The number of rows of the matrix is half of the original image, and the number of columns is the same as the original image; the phase matrix group two includes two matrices, which are sum Γ ₂ ', and the phase matrix group two is obtained according to the phase matrix group one.

3. The method according to claim 2, wherein the plurality of descriptions are five descriptions, and the five are described as: ^{1 = 1 1} , ^{2 = ιύ 2} , ^{3 = 7} 2 ^{ύ 7} 1 ,

Where S represents the original image, SI represents the description 1, S2 represents the description 2, S3 represents the description 3, S4 represents the description 4, S5 represents the description 5, represents the transposition of ^, represents the transposition of T ₂ , represents Τ ₃ Transposed.

The method according to claim 3, wherein the description that is not lost is the five descriptions, or the descriptions that are not lost are description one, description two, description three, and description four, The recovered image is expressed as:

S' = T; ^T ST; + T; ^T S ₂ TT; T; +T ^T S ₄ T , where S' denotes the recovered image, Τ; ^τ denotes transpose, indicating 'transpose.

The method according to claim 3, wherein after the obtaining the description that is not lost, the method comprises: determining, according to the description that the loss is not, the description of the loss is description 1, description 2, description 3 or Description four;

According to the determined phase matrix group 2, the phase frame synthesis inverse transform is performed on the description without loss, and the method for obtaining the restored image includes: subtracting ^{4 χ} from the result of 3 descriptions other than the description 5 The lost description, the resulting recovered image is expressed as:

The method according to claim 3, wherein after the obtaining the description that is not lost, the method comprises: determining, according to the description that is not lost, that the description of the loss is two or three, and the method includes Fifth description;

According to the determined phase matrix group 2, the phase frame synthesis inverse transform is performed on the description without loss, and the method for obtaining the restored image includes: respectively performing inter-row or inter-column mean interpolation processing on the description adjacent to the lost description As the missing descriptions, the resulting recovered image is represented as:

The method according to claim 3, wherein after the obtaining the description that is not lost, the method comprises: determining, according to the description that is not lost, that the description of the loss is two or three, and The fifth description is not included;

According to the determined phase matrix group 2, the phase frame synthesis inverse is performed on the description without loss.

The method of transforming and obtaining the restored image is as follows:

o,

Subtract ^{4x from} no o, there is a description of the loss, divide the subtraction result by the number of missing descriptions, divide the o,

The results are respectively recorded as missing o, descriptions, or inter-row or inter-column mean interpolation of the description adjacent to the missing description, as the missing descriptions, and the resulting restored image is represented as:

2丄

S' = T; ^T ST; + T; ^T S ₂ TT; T; +T ^T S ₄ T , where S' denotes the recovered image, τ; ^τ denotes a transpose, indicating a transposition of ^Τ 2 '.

The method according to any one of claims 3 to 7, wherein the T1, Τ2 and

Τ3 are:

a,l-a, , 0, 0, 0, ···, 0, 0

0, 0, a,l-a, 0, 0, ···, 0, 0

7: =| 0, 0, 0, 0, α, 1-α, ···, 0, 0

0, 0, 0, 0, 0, 0, ···, α,Ι-α,

--\,2-α, 0, 0, 0, 0, ···, 0, 0

0, 0, α-\,2-α, 0, 0, ···, 0, 0

Τ 0, 0, 0, 0, α-1,2-α,···, 0, 0

0, 0, 0, 0, 0, 0, ■•■ a- 2-α 丄,丄, 0, 0, 0, 0, ..., 0, 0

twenty two

0, 0, 丄, 丄, 0, 0, ..., 0, 0

twenty two

τ =

¹ 3 0, 0, 0, 0, -,-,···, 0, 0

twenty two

0 < a < 1, a≠- where ² , the number of lines in Tl, T2 and T3 is half of the original image, the number of columns and the original image Like the same.

9. A frame-based multi-description codec method, the method comprising: performing multi-wavelet transform on an original image to obtain sub-bands of each low-frequency portion and sub-bands of a high-frequency portion; The bands are combined with the respective sub-bands of the high-frequency portion to form a plurality of descriptions; the plurality of descriptions are separately encoded, and the encoded data is transmitted to the decoding end.

10. The method according to claim 9, wherein the method of performing multi-wavelet transform on the original image to obtain each of the low frequency sub-bands and the high-frequency sub-bands comprises: performing a low-pass filter on the original image Low-pass filtering, obtaining a low-frequency coefficient, k, the low-frequency coefficient is a vector of 2 X 1; 釆 high-pass filtering the original image with a high-pass filter to obtain a high-frequency coefficient,

, the high-frequency...coefficient, for, 2 1 'toward a quantity;

According to the low frequency coefficient and the high frequency coefficient, each subband of the low frequency and each subband of the high frequency are obtained.

11. The method according to claim 10, wherein the multi-wavelet inverse transform as: k, wherein _1 ^ ⁵ ^ 1 ^ and low frequency coefficients and the high frequency coefficients are obtained.

12. The method of claim 10, wherein

_( 0.8279 0.5117 θ.1954 -0.1208 —!

The low pass filter is: -0.1208 0.1954J 0.5117. .8279 / ·;

_(-0.5\Π 0.8279 -0.1208 -0.1954 —!

The high pass filter is: " ^(Z) -0.1954 - 0.120 ₈ J ⁺ t 0.8279 -0.5117 / ·.

The method according to claim 9, wherein the combining the sub-bands of the low-frequency portion and the sub-bands of the high-frequency portion to form a plurality of descriptions includes:

The obtained sub-bands of the low-frequency portion and the two sub-bands of the high-frequency portion are respectively formed into a group, and each of the groups is selected to be combined into one description to form a plurality of descriptions.

14. The method according to claim 13, wherein said separately describing said plurality of descriptions Line coding, the method of transmitting the encoded data to the decoding end includes:

For each description, the low frequency sub-bands missing from the description are added from other descriptions, and the added descriptions are obtained, and the multi-wavelet inverse transform is performed on each of the added descriptions to obtain an image after multi-wavelet transform; The image after wavelet inverse transform is separately image-encoded, and the image-encoded data is transmitted to the decoding end.

The method according to claim 14, wherein after the encoding the data is sent to the decoding end, the method includes:

The decoding end respectively decodes the encoded data to obtain a description without loss; according to the description of the loss, the subband of the low frequency part and the subband of the high frequency part are obtained, and the two are combined to obtain a combined signal. And supplementing the subband of the low frequency part lost in the combined signal or/and the subband of the missing high frequency part; performing multi-wavelet inverse transformation on the combined signal after the supplementary processing to obtain the restored image.

16. A framework-based multiple description codec system, the system comprising an encoding end and a decoding end;

The encoding end is configured to perform phase frame decomposition on a pair of original images according to the determined phase matrix group to form a plurality of descriptions, wherein the number of rows of each matrix in the phase matrix group 1 is half of the original image, the number of columns, and the original The images are the same; each of the formed descriptions is separately encoded, and the encoded data is sent to the decoding end;

17. A framework-based multiple description codec system, the system comprising an encoding end and a decoding end;

The encoding end is configured to perform multi-wavelet transform on the original image to obtain each sub-band of the low-frequency part and each sub-band of the high-frequency part; combining the sub-bands of the low-frequency part and the sub-bands of the high-frequency part to form a plurality of Decoding; encoding the plurality of descriptions, and transmitting the encoded data to the decoding end;

The decoding end is configured to receive the encoded data sent by the encoding end, decode the encoded data, and obtain a description that is not lost; according to the description that is not lost, obtain a subband of the low frequency part, and the high The sub-bands of the frequency portion combine the two to obtain a combined signal; and perform multi-wavelet inverse transform on the combined signal to obtain a restored image.

18. A framework-based multiple description encoding apparatus, the apparatus comprising a plurality of description forming modules, an encoding module, and a transmitting module;

The plurality of description forming modules are configured to perform phase frame decomposition on a pair of original images according to the determined phase matrix group, and form a plurality of descriptions and send the same to the encoding module, where the number of rows of each matrix in the phase matrix group is original Half of the image, the number of columns is the same as the original image;

The sending module is configured to send the encoded data to the decoding end.

The apparatus according to claim 18, wherein the plurality of description forming modules comprise a phase matrix group-determining module and a plurality of description forming sub-modules;

The phase matrix group determining module is configured to determine, after the phase matrix group is sent to the plurality of description forming submodules;

The plurality of description forming sub-modules are configured to perform phase frame decomposition on a pair of original images according to the received phase matrix group, form five descriptions, and send the description to the encoding module, where the five descriptions are: = ^Τ ^ ,

, ^ύ 3 ⁼ 2 ^ύ , = ^丄 1 ^丄 1 , 5 = , where S represents the original image, SI represents the description 1, S2 represents the description 2, S3 represents the description 3, S4 represents the description 4, S5 represents the description 5, T' represents The transposition of ^ indicates the transposition of T ₂ and indicates the transposition of Τ ₃ .

20. A frame-based multiple description decoding apparatus, the apparatus comprising a receiving module and a decoding module;

The decoding module is configured to decode the encoded data according to the determined phase matrix group 2, obtain a description without loss, perform inverse phase frame synthesis inverse transformation on the description without loss, and obtain a restored image; The phase matrix group 2 is obtained according to the phase matrix group 1. The number of rows of each matrix in the phase matrix group 1 is half of the original image, and the number of columns is the same as the original image, and the phase moment is The group one is expressed as:

The apparatus according to claim 20, wherein the decoding module comprises a decoding submodule and an image restoration submodule;

The decoding sub-module is configured to decode the encoded data, and obtain a description that is not lost, and then transmit the description to the image recovery sub-module; if no description is lost, the decoded description includes description 1, description 2, description Third, description four and description five, respectively denoted as two ^T i ^ST i ^T , ^S 2 =

= ¹ , =丄ι 丄1 , ^5 = ^ , where S denotes the original image, SI denotes the description 1, S2 denotes the description 2, S3 denotes the description 3, S4 denotes the description 4, S5 denotes the description 5, Tl, Τ2 and Τ3 Representing the three matrices included in the phase matrix group one, indicating the transposition of ^, the transposition indicating Τ ₂ , and the transposition indicating ^; the descriptions without loss are description 1, description 2, description 3, description 4, and description 5; or, the descriptions that are not lost are description 1, description 2, description 3, and description 4;

S' = T; ^T ST; + T; ^T S ₂ TT; T; +T ^T S ₄ T , where S' denotes the restored image, ^ and 2 matrixes representing the phase matrix group 2, The transposition, 7 represents the transposition of ⁷ ^; the phase matrix group 2 is obtained according to the phase matrix group one, and the number of rows of each matrix in the phase matrix group one is half of the original image, and the number of columns is the same as the original image.

22. The apparatus of claim 20, wherein the decoding module comprises a decoding submodule Block and image recovery sub-module;

= ¹ , =丄ι 丄1 , ^5 = ^ , where S denotes the original image, SI denotes the description 1, S2 denotes the description 2, S3 denotes the description 3, S4 denotes the description 4, S5 denotes the description 5, Tl, Τ2 and Τ3 Representing the three matrices included in the phase matrix group one, indicating the transposition of ^, the transposition indicating Τ ₂ , and the transposition indicating D ₃ ; the description of no loss is description 1, description 2, description 3 or description 4;

S^ TjSJl + T\ ^T S ₂ T ₂ + T ₂ ^T S ₄ T ₂ , where S' denotes the restored image, ^ and 2 matrixes representing the phase matrix group 2, representing the transpose, 7 The transposition of ⁷ ^ is represented; the phase matrix group 2 is obtained according to the phase matrix group 1, and the number of rows of each matrix in the phase matrix group one is half of the original image, and the number of columns is the same as the original image.

23. A framework-based multiple description encoding apparatus, the apparatus comprising a plurality of description forming modules, an encoding module, and a transmitting module;

The encoding module is configured to respectively code the formed multiple descriptions, and send the encoded data to the sending module;

The sending module is configured to send the encoded data to the decoding end.

The device according to claim 23, wherein the plurality of description forming modules comprise a multi-wavelet transform sub-module and a multi-description forming sub-module; The multi-wavelet transform sub-module is configured to perform multi-wavelet transform on the original image to obtain each sub-band of the low-frequency part and each sub-band of the high-frequency part, and send the obtained sub-bands of the low-frequency part and the sub-bands of the high-frequency part. Forming a sub-module for multiple descriptions;

The apparatus according to claim 23, wherein the encoding module includes an encoding sub-module for adding a low-frequency sub-band missing from each description from other descriptions, and obtaining each of the added descriptions, Each of the added descriptions is subjected to multi-wavelet inverse transform to obtain an image after multi-wavelet transform; the image after multi-wavelet inverse transform is separately image-encoded, and the image-encoded data is the encoded data.

26. A framework-based multiple description decoding apparatus, the apparatus comprising a receiving module and a decoding module;

The decoding module is configured to decode the encoded data to obtain a description without loss; according to the description that is not lost, obtain a subband of a low frequency part, and a subband of a high frequency part, Combining, obtaining a combined signal; performing multi-wavelet inverse transform on the combined signal to obtain a restored image.