KR0178711B1 - Coding and Decoding System Using Adaptive Post Processing - Google Patents

Abstract

The present invention relates to a system for adaptively post-processing reconstructed data according to the degree of damage of quantization level and reconstructed data in a system for encoding and decoding image data divided into blocks of a predetermined size. A coding system that transmits a degree of damage of the reconstruction signal by calculating a, and adaptively performs predetermined filtering according to the degree of damage transmitted from the coding system, and selects a reconstruction signal or a filtered reconstruction signal according to the quantization level and outputs the result to the display unit. It is an encoding and decoding system for preventing the blocking phenomenon of an image and providing more improved image quality by providing a decoding system.

Description

Coding and Decoding System Using Adaptive Post Processing

1 is a block diagram showing an embodiment of a conventional image data encoding apparatus.

2 is a block diagram showing an embodiment of a conventional image data decoding apparatus.

3 is a graph showing the relationship between buffer occupancy and quantization level.

4 is a block diagram showing an embodiment of an image data encoding apparatus according to the present invention.

5 is a schematic diagram showing a data processing block in FIG.

6 is a block diagram showing an embodiment of an image data decoding apparatus according to the present invention.

7 is a detailed block diagram of a portion of FIG.

8 is a schematic diagram showing a data processing relationship of the first filter in FIG.

9 is a schematic diagram showing a data processing relationship of the second filter in FIG.

* Explanation of symbols for main parts of the drawings

11 DCT unit 13 quantization unit

13,20 variable variable encoding unit 14 buffer

15,25 Inverse quantization unit 16,26 Inverse DCT unit

17,28: frame memory 18: motion prediction unit

19,29: Compensation unit 24,27: Variable length decoding unit

21: data combiner 22: data separator

41: error calculation unit 42: classifier

60: post-processing unit 71, 72: filter

74,75,77: selection unit 73: threshold determination unit

76: position detection unit

The present invention relates to a system for encoding and decoding an image signal, and more particularly, to an encoding and decoding system for providing improved image quality by adaptively processing an image signal according to the degree of motion of an image during encoding and decoding.

In general, DCT (Discrete Cosine Transform) encoding method using motion compensation is used as a method of compressing data to store or transmit image data.

In addition to DCT, there are various methods for encoding video signals. Recently, ICs having processing speeds capable of processing video signals have been developed due to the development of digital signal processing technology. These coding methods are currently being standardized by organizations such as CCITT under ISO and CCIR in Europe.

1 illustrates a DCT encoding system using motion compensation as an example of a conventional encoding apparatus. The encoding apparatus of FIG. 1 includes means for quantizing transform coefficients after performing DCT transform on input image data, means for further compressing the data amount by variable-length encoding the quantized data, and keeping the input / output data amount constant. And a buffer for outputting a predetermined quantization level signal for controlling the quantization means, inverse quantization and inverse transformation of the quantized data, and means for performing motion prediction and motion compensation from the inversely reconstructed picture and the block to be encoded. .

The image data of the N × N block is input through the input terminal 10, and the block image data and the image data fed back calculate the error data in the first adder A1. The error data is converted into data in the frequency domain in the DCT 11, and the energy of the converted conversion coefficient is mainly collected toward the low frequency side. The quantization unit 12 converts the transform coefficients into representative values of a predetermined level through a predetermined quantization process. Then, the first variable length coding unit 13 compresses the data V _{CD by} performing variable length coding using the statistical characteristics of the representative values. The output data of the first variable length coding unit 13 is supplied to the buffer 14, and the buffer 14 outputs a variable quantization level signal so that the input image data and the output image data are kept constant. In other words, the output data in the buffer 14 remains constant, but the input data changes in accordance with the amount of motion or complexity of the image. Therefore, the buffer 14 supplies the quantization level Q to the quantization unit 12 to convert the quantization step size, thereby adjusting the amount of data generated by the first variable length encoding unit 13 to adjust the amount of data generated by the buffer 14. Make sure the input and output data amounts are the same. Figure 3 graphically shows the correlation between quantization level (Q) and buffer occupancy. As shown, the buffer occupancy and the quantization level are proportional. In the case where the buffer occupancy increases, that is, the cumulative data amount of the buffer 14 increases, the quantization level increases, and accordingly, the quantization step size increases, so that the amount of data to be encoded is reduced so that the input / output data amount of the buffer 14 remains constant. .

In general, since there are many similar parts between the screen and the screen, if the screen has a slight movement, the motion vector is estimated by calculating the motion vector, and the data is compensated using the motion vector. Since the difference signal between adjacent screens is very small, the transmission data can be further compressed. In order to perform such motion compensation, the inverse quantization unit 15 and the inverse DCT unit 16 inversely quantize the quantization coefficients output from the quantization unit 12 and inversely transform the image data of the spatial domain. The inversely transformed image data is reproduction error data corresponding to the error data output from the first adder A1. The second adder A2 adds the error data output from the inverse DCT unit 16 to the block data fed back from the motion compensator 19, and stores the output data so that the frame memory 17 displays the screen. Reconstruct Then, the motion predicting unit 18 searches for block data having a pattern most similar to the block data input from the input terminal 10 from the frame data stored in the frame memory 17, and provides a motion vector MV representing the motion between two blocks. Calculate. This motion vector (MV) is transmitted to the receiving side via the second variable length coding unit 20 for use in the decoding system and simultaneously to the motion compensating unit 19. The motion compensator 19 reads block data corresponding to the motion vector from the frame data of the frame memory 17 according to the motion vector supplied from the motion predictor 18 and supplies it to the first adder A1. Then, as described above, the first adder calculates error data between the block data supplied from the input terminal 10 and the block data of the similar pattern supplied from the motion compensator 19, and the error data is encoded again. The encoded image data V _CD , the quantization level Q, and the motion vector MV are thus combined in the data combiner 21 and transmitted to the receiving side as transmission data V _T.

The image data transmitted as described above is input to a decoding apparatus as shown in FIG. The data separator 22 of the decoding apparatus separates the transmission data V _T into the encoded image data V _CD , the quantization level Q and the motion vector MV. Then, the encoded image data V _CD is decoded through a reverse process of variable length encoding by the first variable length encoding unit 24 through the buffer 23, and output from the first variable length encoding unit 24. The data is inversely quantized by the inverse quantization unit 25 by the conversion coefficient of the frequency band. The inverse DCT unit 26 converts the frequency domain transformation coefficient supplied from the inverse quantization unit 25 into image data of the spatial domain. In this case, the inversely transformed image data is reproduction error data corresponding to the error data calculated from the first adder A1 in the encoding apparatus of FIG. In addition, the motion vector MV calculated and transmitted by the motion predictor 18 of the encoding apparatus is similarly separated by the data separator 22 and decoded by the second variable length encoder 27, and then the motion compensator ( 29), the motion compensator 29 protrudes block data corresponding to the motion vector MV from the frame data stored in the frame memory 28 and supplies it to the adder A. FIG. Then, the inversely transformed error data and the block data supplied from the motion compensation unit 29 are combined in the adder A and transmitted to the display unit.

However, in the conventional encoding and decoding system, when the data rate is low, that is, when the compression rate is high, a blocking phenomenon occurs in which the block boundary of the reconstructed picture is prominent when there is a lot of motion or irregular motion in the picture. In addition, irregular noise due to quantization noise, etc., causes a problem of visual degradation of the restored image.

Accordingly, an object of the present invention is to calculate the degree of corruption of image data that has been encoded and then reconstructed for image data before being encoded, and to output the image data together with the encoded image data. The present invention provides an encoding system for transmitting more efficient encoding data.

Another object of the present invention is to further improve image quality by adaptively post-processing a decoded video signal according to the degree of damage of the reconstructed picture and the quantization level in decoding the video data transmitted from the encoding system as described above. To provide a decoding system for providing.

The object of the present invention is to provide a means for quantizing image data divided into blocks of a predetermined size, a buffer for outputting a quantization level for adjusting the output data amount of the quantization means, and a quantization data for restoring motion compensation. An apparatus for encoding video data having a means for receiving the video data, the means for calculating an error value between two data and receiving the video data input to be encoded and the video data reconstructed by the restoring means, and outputting the error data from the error calculating means. It is achieved by the classification means for outputting and transmitting the data of the corresponding grade by comparing the error value to the predetermined grades.

Another object of the present invention is a means for inverse quantization of coded image data divided into blocks having a predetermined size according to the transmitted quantization level, and restoring the inverse quantized image data using a motion compensation method to a state before encoding. A decoding apparatus having means, comprising: means for inputting encoded data output from the above-described encoding apparatus and transmitted, and quantization level and grade data supplied from the input means, and image data supplied from the means for restoring the restored image data. It is achieved by post-processing means for adaptively filtering according to the grade data and selectively outputting the filtering data and the reconstructed image data according to the quantization level.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. 4 shows an embodiment of an encoding apparatus according to the present invention. In the encoding apparatus of FIG. 4, the encoding apparatus of FIG. 1 includes an error calculator 14 that calculates error data between the reconstructed image data and the original image data before being encoded, and classifies the calculated error data according to the degree of error. It further comprises a classifier 42 classified as. In FIG. 4, the same reference numerals are given to the same elements as those in FIG. 1, and the description of the operation of the same elements will be omitted.

The error calculator 41 receives the block data S _i and the reconstructed block data S _o output from the second adder A2 through the input terminal 10. Compute the error of the data. At this time, the data processing unit of the error calculator 41 is a macroblock of a predetermined size, and one macroblock is composed of a plurality of unit blocks. FIG. 5 is a schematic diagram showing the relationship between the unit block and the macro block. In the embodiment, 8 × 8 unit blocks and 32 × 16 macro blocks will be used. The error E between the original image data S _i and the restored image data S _o for the 32 × 16 macroblock is calculated by the following formula (1).

Where E is an error value, MN is the size of the macroblock, and (m, n) is the position of the pixel.

Error values from the formula 1] (E) represents the mean value per pixel for the error of the original image data (Si) and the restored image data (S _o) of the macroblock. The larger the M and N values, the larger the size of the macroblock, the smaller the amount of data to be transmitted, but the accuracy for performing post-processing in the decoder decreases. The accuracy at is improved.

In this way, the error value between the original image and the restored image calculated by the error calculator 41 is supplied to the classifier 42. The classifier 42 classifies it into various grades according to the magnitude | size of an error value, and each grade represents the damage degree of a restored image. At this time, reference values Ref for classifying the error value E into each class are experimentally obtained. For example, when classifying the error value E into four classes, three reference values (Ref1, Ref2, Ref3) are required.

ERef1 is class 0, Ref1≤E≤Ref2 is class 1, Ref2≤E≤Ref3 is class 2, and E≥Ref3 is class 3. At this time, Ref1, Ref2 and Ref3 may be 2, 4, 6, etc. as the magnitude of the experimentally obtained signal. Thus, the smaller the error, the smaller the rating, and the larger the error, the larger the rating. The class data CLASS output from the classifier 42 is combined with the video signal V _CD , the quantization level Q, and the motion vector MV coded by the data combiner 21 and transmitted to the decoding apparatus. .

6 shows a decoding apparatus according to the present invention. The decoding apparatus of FIG. 6 further includes a post processor 60 for adaptively post-processing the reconstructed video signal V _DCD according to the quantization level Q and the class CLASS in the decoding apparatus of FIG. Doing. In FIG. 6, the same reference numerals are given to the same elements as those in FIG. 2, and the description of the operation of the same elements will be omitted.

The post processor 60 receives the quantization level Q and the class CLASS calculated and transmitted by the encoding apparatus, and receives the decoded image data V _DCD by a series of decoding systems. The post processor 60 adaptively filters the decoded image data V _DCD according to the input class CLASS, and filters the filtered image data or the decoded image data V _DCD according to the quantization level Q. Optionally output A detailed block diagram of such a post processing unit 60 is shown in FIG.

The post-processing unit of FIG. 7 includes a first filter 71 which performs predetermined filtering on the decoded image data V _DCD to suppress the blocking phenomenon, and a reconstructed image data V having a quantization level Q of a predetermined size or more. _DCD) a first selector for selectively outputting the reconstructed image data (V _DCD) or low-pass filtered image data (V _F2) according to the second filter 72 and, rating data (CLASS) for lowpass filtering ( 74), the second selecting unit 75 for selecting and outputting the output signal of the first filter 71 when the reconstructed image data is the boundary of the block, and the reconstructed image data V _DCD according to the quantization level Q. The third selector 77 selectively outputs the output image data VS2 of the second selector 75. The post processor 60 receives the reconstructed image data V _DCD and the quantization level Q and the class data CLASS transmitted from the encoding apparatus, respectively. The reconstructed image data V _DCD is input to one input terminal of the first filter 71, the second filter 72, and the first selecting unit 74 and one input terminal of the third selecting unit 77, respectively. The first filter 71 performs median filtering on the boundary of the block of the reconstructed image data to suppress the blocking phenomenon. The pixel filtered by the spatial median filter used as the first filter 71 is illustrated in FIG. 8. FIG. 8 is for the M × N macroblock, where X is pixels located at the boundary of the block, and 0 represents pixels located in other areas. As shown in the drawing, the median filter selects one pixel and the pixels located at the top, bottom, left, and right of the pixel (dotted line in Fig. 8) for the pixel located at the boundary of the block, and the data of each of these five pixels is selected. By comparing the sizes, the data size of the pixel having the median value is set as the representative value of the five pixels. Therefore, it is possible to reduce the level difference between adjacent pixels generally present in the block boundary. As such, when the level difference of the block boundary is reduced, the blocking phenomenon is suppressed.

The second filter 72 is a low pass filter for attenuating high frequency components in the reconstructed image data. In the reconstructed image data, when the quantization level Q is larger than the predetermined reference value Qr and the degree of damage is relatively small, three-dimensional low pass filtering is performed as shown in FIG. 9. One pixel P of the macroblock M _{2 from} which the three-dimensional low pass filtering is filtered, the pixels a, b, c, and d located at the top, bottom, left, and right sides of the pixel, and the front and back frames M Filtered output image data is obtained by performing an operation as shown in Equation 2 with respect to the pixels e and f existing at the same positions of ₁ and M ₃ ).

Where V _{LF (P)} is the LPF output for the P pixel, a to f are the size of the sample data of each pixel located above, below, left, right, before and after the P sample data being sampled, and P is the current sampling. This is the size of the sample data of the current pixel.

The 3D low pass filtered image data is supplied to another input terminal of the first selector 74. In addition, the first selector 74 receives a threshold value TH of a predetermined size from the threshold value determiner 73. The threshold determination unit 73 receives grade data CLASS transmitted from the encoding apparatus and generates different threshold values TH according to the grade data. Therefore, the greater the damage degree of the restored image, the larger the threshold TH.

Then, the first selector 74 selects and outputs the restored data V _DCD or the output data V _F2 of the second filter in accordance with the threshold TH supplied from the threshold value determiner 73. That is, as shown in Equation 3 below, the difference between the restoration data V _DCD and the filtering data V _F2 is calculated, and if the difference value is not smaller than the threshold value TH, the filtering data V _F2 is selected and the difference value is selected. If it is smaller than this threshold TH, the restoration data V _DCD is selected.

The output signal of the first selector 74 thus obtained is supplied to the second selector 75. Then, the second selector 75 selects the output signal V _F1 of the first filter 71 or the output signal V _F2 of the second filter 72 according to the output signal of the position detector 76. . Here, the position detector 76 receives the restoration data V _DCD to detect the boundary position of the block. That is, the boundary position data detected by the position detector 76 is supplied to the second selector 75 to select the output signal of the first filter 71 that filters the pixels located at the boundary of the block. Therefore, when the position detection unit 76 detects the position data corresponding to the block boundary, the second selector 75 selects and outputs the output signal V _F1 of the first filter 71, and otherwise, The output signal V _F2 of the second filter 72 is selected and output. The output signal of the second selector 75 thus obtained is supplied to the third selector 77. Then, the third selector 77 selects and outputs the image data V _DCD reconstructed or the output signal VS2 of the second selector according to the quantization level Q transmitted from the encoding apparatus. The third selector 77 outputs reconstruction data V _DCD when the input quantization level Q is smaller than a predetermined threshold TQ obtained experimentally, and the quantization level Q is the threshold value TQ. If larger, the output signal VS2 of the second selector 75 is output. That is, when the quantization level Q is relatively small (Q≤TQ), since the probability of error is relatively small in the reconstruction data, the reconstruction data is transmitted to the display unit as it is, and when the quantization level Q is relatively large (Q≥TQ). Since the probability of error in the reconstruction data is relatively large, the reconstruction data is filtered and transmitted.

As described above, the encoding and decoding system according to the present invention calculates the degree of damage between the original signal and the reconstruction signal in the encoding system and transmits it to the decoding system, and the decoding system reconstructs the data according to the degree of damage of the reconstruction signal and the quantization level. By adaptively filtering, it provides better image quality while suppressing blocking phenomenon and compressing transmission data effectively.

Claims (11)

Image data comprising means for quantizing image data divided into blocks of a predetermined size, a buffer for outputting a quantization level for adjusting the output data amount of the quantization means, and means for restoring quantization data for performing motion compensation An encoding apparatus comprising: means for receiving an image data input to be encoded and image data reconstructed by the decompression means, and calculating an error value between two data; And classifying means for comparing and outputting error values output from the error calculating means with predetermined grades to output and transmit data of the corresponding grades. 2. The apparatus of claim 1, wherein the error calculating means calculates an average value per pixel of the image data block based on the error values of the input image data and the reconstructed image data. The apparatus of claim 1, wherein the error calculating means calculates and processes image data of a macroblock composed of a plurality of unit blocks. 2. The apparatus of claim 1, wherein the classification means comprises: means for dividing the minimum value and the maximum value of the error value output from the error calculation means at predetermined intervals; Means for assigning representative value codes to the separated intervals, respectively; And means for outputting a representative value code of a section including an input error value as a grade. And a means for inverse quantization of image data divided and encoded into blocks having a predetermined size according to the transmitted quantization level, and means for restoring the dequantized image data to a state before encoding using a motion compensation method. Means for inputting encoded data output from the encoder of claim 1 and transmitted; Receive the quantization level and grade data supplied from the input means and the image data reconstructed by the reconstruction means and adaptively filter the reconstructed image data according to the grade data, and selectively filter the filtered and reconstructed image data according to the quantization level. And post-processing means for outputting the video data decoding apparatus. 6. The apparatus of claim 5, wherein the post processing means comprises: means for comparing the quantization level with a predetermined threshold value; And first selecting means for transmitting the reconstructed image data to a display means when the quantization level is less than a threshold value, and for transmitting the filtered data to a display means when the quantization level is not less than a threshold value. An apparatus for encoding video data. The apparatus of claim 5 or 6, wherein the filtering unit comprises: a first filter for low-pass filtering the image data located at the boundary of each block divided for the input reconstructed image data; A second filter for selectively performing low pass filtering according to the grade data supplied through the input means; And second selecting means for selecting the output signal of the first filter or the output signal of the second filter and outputting the output signal of the first filter to the first selecting means according to the positions of blocks of the image data input to the first and second filters. Decoding apparatus for video data, characterized in that. 8. The method of claim 7, wherein the first filter selects a sample data having a median data size among any one sample data located at the boundary of the block and the sample data located at the top, bottom, left, and right of the sample data. And a filter for selecting and outputting the image data. 8. The method of claim 7, wherein the second filter is predetermined for any one sample data positioned outside the boundary of the block in the reconstructed image data of the current frame, and the sample data adjacent to each other in three-dimensional space with respect to the sample data. A low pass filter for performing an operation; Means for receiving the grade data and calculating a predetermined threshold value indicating a degree of damage of the reconstructed image data; Means for comparing an error value between the reconstructed image data and an output data of a low pass filter and the threshold value; And selecting means for selecting output data of the low pass filter if the error value is greater than a threshold value, and selecting and outputting the reconstructed image data if the error value is not greater than a threshold value. Device. 10. The apparatus of claim 9, wherein the low pass filter performs an operation such as the following expression on predetermined sample data. Where P is the size of any sample data that is currently filtered, a to f is the size of the sample data located above, below, left, right, before and after the p sample data, Is the output of the filter. 8. The apparatus of claim 7, wherein the second selecting means comprises: means for detecting whether the input data is located at the boundary of the block by receiving the restored image data applied to the input terminal of the post-processing means; The output signal of the first filter and the second filter is supplied, and if the data of the block boundary is detected by the position detecting means, the output signal of the first filter is selected, and if the data other than the block boundary is detected, the output of the second filter is detected. And a means for selecting a signal.