KR19980030414A - Moving picture decoding apparatus based on compulsory one-way motion compensation - Google Patents

Abstract

Disclosed is the amount of memory and the amount of memory required for motion compensation by forcibly restricting and decoding motion compensation to one of the forward and reverse directions for a macroblock encoded by a bidirectional motion compensation method in an MPEG-based video decoding apparatus. The present invention relates to a new moving picture decoding apparatus which reduces the overall complexity and price by reducing ½ or less. The apparatus of the present invention includes all the configurations of a general moving picture decoding apparatus, and extracts picture shape information and macroblock shape information from input picture header information and macroblock header information to determine a motion compensation mode of a currently decoded macroblock. Motion compensation type detector and judgment result, if it is not two-way motion compensation mode, it passes macroblock header information as it is and operates in the original mode, and if it is two-way motion compensation mode, it selects one of two directions and compensates for one-way motion compensation. And a one-way exercise compensator for performing exercise compensation according to the selected exercise compensation mode. Here, the one-way selection criterion of the bidirectional is the next frame in which the difference (TT ') between the positions T' of the previous frame used in the forward motion compensation with respect to the position T in the time axis of the currently decoded B-picture frame is used in the backward motion compensation. If it is less than or equal to the difference (T "-T) between positions T", it selects forward motion compensation, otherwise it selects backward motion compensation.

Description

Moving picture decoding apparatus based on compulsory one-way motion compensation

The present invention relates to a moving picture decoding apparatus, and more particularly, by compensating and decoding motion coded data in one direction of a forward or reverse direction for data encoded by a bidirectional motion compensation method, a memory required for motion compensation and a calculation amount thereof can be reduced. The present invention relates to a moving picture decoding apparatus based on forced one-way motion compensation.

Recently, applications for moving picture services such as digital TV, interactive TV, and video on demand (VOD) are rapidly spreading. In such systems, it is necessary to compress a digital video using a video encoding algorithm such as a Moving Picture Experts Group (MPEG) in order to store a large amount of digital video on a limited storage medium or transmit it through a limited channel. In order to implement a video decoder based on standardization systems such as MPEG-1 and MPEG-2, it is necessary to accommodate the most complicated cases defined in the MPEG video decoding algorithm. In other words, the most complex case in terms of complexity and memory bandwidth in the MPEG video decoding algorithm is a case in which all macroblocks are motion-compensated in B-picture in both directions. Should be implemented. One example is shown in FIGS.

Fig. 1 shows a block diagram of a general MPEG video decoding apparatus, and the encoded coded bitstream data is temporarily stored in the buffer 11. The encoded bitstream data stored in the buffer 11 is input to the parser 12. The parser 12 parses input coded bitstream data according to a corresponding standard to restore various parameters necessary for syntax and decoding of a video sequence in the bitstream. That is, macroblock headers such as a motion vector and a pattern of a coding block are decoded in header information related to encoding, such as the shape, size, and encoding form of an encoded video sequence, and a macroblock layer. The variable length decoder 13 receives variable length codes for the discrete cosine transform (DCT) coefficients from the parser 12 and performs variable length decoding, and the variable length decoded data is quantized by the scan converter 14. It is restored to two-dimensional discrete cosine transform (DCT) coefficient block. The inverse quantization unit 15 receives the inverse quantized discrete cosine transform (DCT) coefficients and inverse quantizes it, and the inverse discrete cosine transform (DCT) unit 16 converts the inverse discrete cosine to restore an error signal block. On the other hand, the motion compensator 17 receives the macroblock header information from the parser 12, and the data at the position designated by the motion vector in the macroblock header information from the previous frame stored in the memory 19. Is extracted to form a prediction block. Here, a more specific operation of the exercise compensator 17 will be described later. The adder 18 combines the prediction block output from the motion compensator 17 and the error block output from the inverse discrete cosine transform (DCT) unit 16 to form a reconstruction block, and stores the reconstruction block in the memory 19. At the same time output to the display unit not shown.

FIG. 2 is a detailed view of the exercise compensator 17 of the apparatus of FIG. 1, wherein the exercise compensator 21 of the exercise compensator 17 is a macroblock header applied from the parser 12 (see FIG. 1). The motion compensation mode is determined by determining the encoding form of the macroblock data currently decoded by the variable length decoder (see FIG. 1) from the macroblock header information, and outputs the motion vector accordingly. Here, the motion vector is one of three cases including both the forward motion vector Vf, the backward motion vector Vb, and the forward and backward motion vectors according to the encoding form of the currently decoded macroblock data. When the motion compensation controller 21 determines that the current macroblock data is encoded according to the forward motion compensation from the macroblock header information input, the motion compensation controller 21 converts the motion vector in the macroblock header information into the forward motion vector ( Output to the forward motion compensator 22 as Vf). On the other hand, if the motion compensation controller 21 determines that the current macroblock data is encoded according to the backward motion compensation from the macroblock header information input, the motion compensation controller 21 reverses the motion vector in the macroblock header information. It outputs to the backward motion compensator 23 as a motion vector Vb. If it is determined that the current macroblock data is encoded according to the bidirectional motion compensation from the macroblock header information input, the motion compensation controller 21 corresponds to the motion vector for both directions in the macroblock header information. The forward motion compensator 22 to the reverse motion compensator 23 is output. That is, if the frame currently decoded is P (Predictive-coded) -picture, only forward motion compensation is performed, and thus the transmitted motion vector always has only forward motion vector Vf. However, in case of B (Bidirectionally predictive-coded) -picture, each macroblock can be one of three cases such as forward motion compensation, backward motion compensation, and bidirectional motion compensation, and thus the motion vector transmitted varies according to each motion compensation type. The forward motion compensator 22 uses the forward motion vector Vf applied from the motion compensation controller 21 as shown in FIG. 3A to forward motion vector Vf from the previous frame stored in the memory (see FIG. 1) 19. The data at the position indicated by is taken to form the prediction block MC_f. At this time, if the transmitted motion vector has half pixel precision, the prediction block is constructed using half pixel interpolation. Also, as shown in FIG. 3B, the backward motion compensator 23 uses the backward motion vector Vb applied from the motion compensation controller 21 to the memory (see FIG. 1) 19 as in the forward motion compensation of FIG. 3A. The prediction block MC_b is obtained by taking the data of the position indicated by the backward motion vector Vb from the next stored frame. In the case of bidirectional motion compensation, as shown in FIG. 3C, the forward prediction block MC_f is obtained using the forward motion vector Vf and the previous frame, and then the backward prediction block is obtained using the backward motion vector Vb and the restored next frame. After the MC_b is obtained, the forward prediction block MC_f and the backward prediction block MC_b are added through the adder 24 to obtain an average divided by two to obtain a bidirectional prediction block MC_fb. In the switch 25, the forward prediction block MC_f applied from the forward motion compensator 22 and the reverse prediction block MC_b applied from the backward motion compensator 23 and the bidirectional prediction block MC_fb applied through the adder 24 are provided. ) Selects a prediction block corresponding to the exercise compensation mode according to the determination of the exercise compensation controller 21 and outputs the predicted block to the adder 18 (see FIG. 1).

However, in the case of bidirectional motion compensation, two prediction blocks must be extracted from the previous frame and the next frame, thus requiring more than twice as much memory and computation amount as the one-way motion compensation in the forward and reverse directions. Furthermore, in the case of B-pictures, since each macroblock may be encoded by bidirectional motion compensation, in the worst case, all macroblocks in one frame may be encoded by bidirectional motion compensation. Therefore, in order for the MPEG-based video decoding apparatus to work properly, the implementation complexity (memory amount and computation amount) of the system should be designed in a case where all macroblocks in the B-picture are encoded by bidirectional motion compensation. Therefore, the complexity of the entire system must be set so that all macroblocks can be decoded according to bidirectional motion compensation for one frame time (1/30 second). This requires a very complicated design, which increases the price of the entire video decoding apparatus.

Accordingly, an object of the present invention is to forcibly set the forward to reverse one-way motion compensation mode for the B-picture bidirectional motion compensation mode having the highest implementation complexity in the MPEG-based video decoding apparatus, and thus for the complexity and memory amount. The present invention provides a video decoding apparatus based on a compulsory one-way motion compensation to reduce the overall complexity and to implement a low cost video decoding apparatus.

1 is a block diagram showing a general moving picture decoding apparatus;

2 is a detailed view showing a motion compensation unit of the device of FIG.

3 is a view for explaining an exercise compensation method for each exercise compensation mode;

4 is a block diagram showing a moving picture decoding apparatus based on forced one-way motion compensation according to the present invention;

5 is a detailed view showing a motion compensation selection unit of FIG. 4;

Figure 6 is a detailed view showing the one-way motion compensation of the Figure 4 device.

* Explanation of symbols for main parts of the drawings

41: buffer 42: parser

43: variable length decoder 44: scan conversion unit

45: inverse quantization unit 46: inverse DCT unit

47: motion compensation selector 48: one-way motion compensation

49: adder 50: memory

The moving picture decoding apparatus based on the compulsory one-way motion compensation of the present invention for achieving the above object is a moving picture decoding apparatus that decodes and inputs coded bitstream data, wherein the picture header information and macroblocks in the coded bitstream data are output. Determines the motion compensation mode of the macroblock data currently decoded by receiving the header information, and outputs the motion vector in the macroblock header information together with the motion compensation mode in the forward or reverse motion compensation mode, and forcibly forwards in the bidirectional motion compensation mode. And a motion compensation selector which selects one of the reverse directions and outputs a motion vector in the macroblock header information together with the motion compensation mode, and a motion compensation using the motion compensation mode and the motion vector applied from the motion compensation selection part. One-way motion compensation The.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

4 is a block diagram showing a moving picture decoding apparatus based on forced one-way motion compensation according to the present invention. As shown, the apparatus of the present invention is configured in the same manner as the conventional apparatus of FIG. 1, and receives motion picture compensation information and macroblock header information from the parser 42. It further comprises a motion compensation selecting section 47 for selecting a mode and a modified motion compensation section 48 for compensating for motion only in one direction in the forward or reverse direction in accordance with the selected exercise compensation mode (dotted line display section).

The operation of the moving picture decoding apparatus based on the compulsory one-way motion compensation of the present invention configured as described above will be described in more detail with reference to FIGS. 5 and 6.

FIG. 5 is a detailed view of the motion compensation selecting unit 47 of the apparatus of FIG. 4. As shown, the motion compensation selection unit 47 of the present invention includes a motion compensation shape detector 51 for detecting a motion compensation shape from input picture header information and macroblock header information; When the detected motion compensation form is a bidirectional motion compensation mode, a motion compensation direction selector 53 is provided for selecting one of the two directions. In addition, the motion vector and the motion compensation direction selector output from the first switch 52 to pass the motion vector as it is or output to the motion compensation direction selector 53 according to the detected motion compensation form. The second switch 54 for selectively outputting the motion vector output from the 53 and the first motion compensation mode output from the motion compensation shape detector 51 and the motion compensation mode selected by the motion compensation direction selector 53. It is configured to have a third switch 55 for selectively outputting.

FIG. 6 is a detailed view of the unidirectional motion compensation unit 48 of the apparatus of FIG. 4. As shown, the one-way motion compensator 48 of the present invention uses the motion vector input according to the motion compensation mode applied from the motion compensation selector 47 (see FIGS. 4 and 5) 47 to forward motion vector Vf. And the forward motion compensation block MC_f by performing forward motion compensation according to the fourth switch 61 outputting one of the reverse motion vectors Vb and the forward motion vector Vf input through the fourth switch 61. The backward motion compensator 63 outputs the backward prediction block MC_b by performing backward motion compensation according to the forward motion compensator 62 to be output and the reverse motion vector Vb input through the fourth switch 61. Equipped. In addition, a fifth switch 64 selectively outputs the forward prediction block MC_f and the backward prediction block MC_b between the forward motion compensator 62 and the backward motion compensator 63.

First, in FIG. 4, the buffer 41 stores the input coded bitstream data once, and outputs the stored coded bitstream data to the parser 42. The parser 42 parses the input bitstream data to conform to the MPEG standard, and parses the picture header information and the macroblock header information into a motion compensation selection unit. 47), and variable length codes for the discrete cosine transform (DCT) coefficients are output to the variable length decoder 43. The variable length decoder 43 variably decodes the input variable length codes, and the scan converter 44 receives the variable length decoded data and restores the quantized two-dimensional discrete cosine transform (DCT) coefficient block. The inverse quantization unit 45 and the inverse discrete cosine transform (DCT) unit 46 perform inverse quantization and inverse discrete cosine transform (IDCT) on the quantized discrete cosine transform (DCT) coefficients in order to restore the error signal block, and adder 49 ) On the other hand, the motion compensation selector 47 receives picture header information and macroblock header information from the parser 42 and executes motion of macroblock data currently decoded by the variable length decoder 43. Recognizes the compensation mode and forcibly switches to the one-way motion compensation mode. Here, a more specific operation of the exercise compensation selecting unit 47 will be described later. The one-way motion compensator 48 extracts the data at the position designated by the motion vector from the previous reconstructed frame stored in the memory 50 according to the motion compensation mode and the motion vector selected by the motion compensation selector 47 to obtain a prediction block. Form. Here, a more specific operation of the one-way motion compensation unit 48 will be described later. The adder 49 adds the prediction block formed by the one-way motion compensator 48 and the error block formed by the inverse discrete cosine transform (DCT) unit 46 to form a reconstruction block, and adds it to the memory 50. At the same time as the storage and output to the display unit not shown.

In FIG. 5, the motion compensation shape detector 51 of the motion compensation selecting unit 47 is configured in picture header information and macroblock header information applied from the parser 42 (see FIG. 4). Determine the motion compensation mode of the macroblock data currently decoded by the variable length decoder (see FIG. 4) by extracting picture type and macroblock type information, and determining the determined motion compensation mode and Output the motion vector according to the The first switch 52 selectively outputs the motion vector applied from the motion compensation shape detector 51 to the second switch 54 to the motion compensation direction selector 53 under the control of the motion compensation shape detector 51. . That is, if the motion compensation shape detector 51 determines the forward or reverse motion compensation mode, the motion compensation mode detector 51 outputs the motion vector to the second switch 54 through the first switch 52 so that decoding can be performed intact. . On the other hand, if the motion compensation shape detector 51 is a B-picture and the macroblock type is an interpolated macroblock, the motion compensation shape detector 51 determines the bidirectional motion compensation mode to force one-way motion compensation. The motion vector is output to the motion compensation direction selector 53 through the first switch 52 so that decoding can be performed in the mode. On the other hand, the motion compensation direction selector 53 selects one of the motion compensation directions in the forward and reverse directions by using the motion vector input through the first switch 52 and forcibly switches to the one-way motion compensation mode. The criterion for selecting one of the forward and reverse directions in the motion compensation direction selector 53 is, as shown in Fig. 3c, that the position of the currently decoded B-picture frame is T on the time axis, and is used for forward motion compensation. If the position of the frame is T 'and the position of the next frame used for the backward motion compensation is T ″, if (TT') is less than or equal to (T ″ -T), the forward motion compensation is selected. Choose as a reward. The second switch 54 selectively outputs the motion vector applied from the first switch 52 and the motion vector applied from the motion compensation direction selector 53 under the control of the motion compensation shape detector 51. Here, since the second switch 54 is interlocked with the first switch 52, if the motion compensation type detector 51 determines the forward or reverse motion compensation mode, the motion vectors are the first switch 52 and the second switch 54. ) Is passed through as it is and is output to the one-way motion compensator (see FIG. 4) 48. In addition, when it is determined that the motion compensation mode detector 51 is the bidirectional motion compensation mode, the motion vector is one-way motion compensation unit through the first switch 52, the motion compensation direction selector 53, and the second switch 54 (FIG. 4). 48). On the other hand, the third switch 55 selectively selects the first motion compensation mode determined by the motion compensation shape detector 51 and the motion compensation mode selected by the motion compensation direction selector 53 according to the control of the motion compensation shape detector 51. Will print Of course, since the third switch 55 also interlocks with the first switch 52 and the second switch 54, if the motion compensation type detector 51 determines the forward or reverse motion compensation mode, the first motion compensation mode is selected. Output to the one-way motion compensator (see Fig. 4) 48, and if it is determined that the two-way motion compensation mode is selected, the one-way motion compensator (see Fig. 4) by forcibly selecting the motion compensation mode selected by the motion compensation direction selector 53 ( 48).

In FIG. 6, the one-way motion compensator 48 receives the one-way motion compensation mode and the motion vector from the motion compensation selector 47 to form a prediction block by the one-way motion compensation. That is, the fourth switch 61 outputs the input motion vector as the forward motion vector Vf to the reverse motion vector Vb according to the input motion compensation mode. The forward motion compensator 62 uses the forward motion vector Vf applied from the fourth switch 61 to display data of the position indicated by the forward motion vector Vf from the previous frame stored in the memory (see FIG. 4) 50. To obtain the forward prediction block MC_f. The backward motion compensator 63, like the forward motion compensator 62, is a backward prediction block from the next frame stored in the memory (see FIG. 4) 50 using the backward motion vector Vb applied from the fourth switch 61. Obtain (MC_b). The fifth switch 64 selects the forward predicting block MC_f applied from the forward motion compensator 62 and the reverse predicting block MC_b applied from the backward motion compensator 63 according to the input motion compensation mode, thereby adding an adder ( 4).

Thus, by compensating and decoding the data in macroblock units encoded by bidirectional motion compensation in B-picture in one of the forward and reverse directions, the amount of memory and computation of the entire system is less than ½ of the conventional MPEG video decoding apparatus. It can be reduced to implement a low-cost video decoding apparatus. Here, the quality of the B-picture may be degraded by decoding the macroblock encoded by the bidirectional motion compensation into one-way motion compensation, but the effect of the B-picture on the overall picture quality is more than that of other types of pictures. Its small size does not significantly affect the overall picture quality.

As described above, the present invention relates to a moving picture decoding apparatus based on a compulsory one-way motion compensation, and compulsorily for a macroblock decoded by a B-picture bidirectional motion compensation method that determines the complexity when implementing an MPEG-based moving picture decoding apparatus. By limiting to one of the forward and reverse directions, the amount of memory and calculation required for the exercise compensation can be reduced to less than ½, which reduces the complexity and reduces the overall price.

Claims (7)

A moving picture decoding apparatus for decoding and outputting input coded bitstream data, comprising: receiving picture header information and macroblock header information in the coded bitstream data and determining a motion compensation mode of the currently decoded macroblock data to determine whether the motion compensation mode is forward or not. In the backward motion compensation mode, the motion vector in the macroblock header information is output together with the motion compensation mode.In the two-way motion compensation mode, one of the forward and reverse directions is forcibly selected to display the motion vector in the macroblock header information together with the motion compensation mode. And a motion compensation selector for outputting a motion vector, and a motion compensation mode configured to perform motion compensation using the motion compensation mode and the motion vector applied from the motion compensation selector. The method of claim 1, wherein the motion compensation selection unit receives picture header information and macro block header information to detect picture shape and macro block shape information to determine a motion compensation mode of the currently decoded macroblock data, and to thereby compensate for motion. Motion compensation shape detector for outputting mode and motion vector If the motion compensation shape detector determines that the motion compensation mode is a bidirectional motion compensation mode, the motion compensation direction selects one of the motion compensation direction among the forward motion compensation mode and the backward motion compensation mode according to a predetermined criterion. If the selector and the motion compensation shape detector determine the forward or reverse motion compensation mode, the motion compensation mode and the motion vector are output to the one-way motion compensation unit as they are, and if the motion compensation shape detector is determined to be the bidirectional motion compensation mode, the motion compensation is performed. Exercise compensation mode and exercise in one direction selected from the direction selector And an output unit for outputting a vector to the one-way motion compensator. 3. The first switch of claim 2, wherein the output unit selectively outputs a motion vector applied from the motion compensation shape detector to a second switch or the motion compensation direction selector according to a result of the motion compensation mode determination of the motion compensation shape detector. A second switch for selectively outputting the motion vector input through the first switch and the motion vector input through the motion compensation direction selector to the one-way motion compensator according to the motion compensation mode determination result of the motion compensation type detector; And a third switch for selectively outputting the first motion compensation mode determined by the motion compensation shape detector and the motion compensation mode selected by the motion compensation direction selector to the one-way motion compensator according to the result of the motion compensation mode judgment of the motion compensation shape detector. Based on compulsory one-way compensation The picture decoding apparatus. The method of claim 3, wherein the motion compensation shape detector detects the motion vector in the macroblock header information as it is when the macroblock data currently decoded by the detected picture shape and macroblock shape information is determined to be forward or backward motion compensation mode. Compulsory one-way motion compensation by passing through the first switch and the second switch in order to output to the one-way motion compensation unit, and outputs the determined first motion compensation mode to the one-way motion compensation unit through the third switch as it is. Based moving picture decoding apparatus. The method of claim 3, wherein the motion compensation shape detector detects the currently decoded macroblock data as a bidirectional motion compensation mode when the detected picture shape is a B-picture and the macroblock shape is interpolated. Outputs a motion vector to the motion compensation direction selector through the first switch and compensates the one-way motion through each of the second and third switches together with a motion compensation mode of one direction forcedly selected by the motion compensation direction selector. A moving picture decoding apparatus based on compulsory one-way motion compensation, characterized in that outputted as a negative. 3. The motion compensation direction selector according to claim 2, wherein the motion compensation direction selector has a difference (TT ') between the position (T') of the previous frame used in the forward motion compensation with respect to the position (T) in the time axis of the currently decoded B-picture frame. Forced one-way motion, characterized in that the forward motion compensation mode is selected if it is less than or equal to the difference (T ″ -T) between the position T ″ of the next frame used in the backward motion compensation, otherwise the reverse motion compensation mode is selected. A video decoding apparatus based on compensation. The fourth switch of claim 1, wherein the one-way motion compensator receives a motion vector and a motion compensation mode from the motion compensation selector, and outputs a motion vector input according to the motion compensation mode as a forward motion vector or a reverse motion vector. Using the forward motion vector input through the fourth switch to obtain a forward prediction block by taking the data of the position indicated by the forward motion vector from the previous frame reconstructed using the forward motion vector input through the fourth switch And a fifth switch for selectively outputting the forward predictive block and the backward predictive block according to the input motion compensation mode by taking the data of the position indicated by the reverse motion vector from the next frame restored. Compulsory one-way motion compensation characterized by having Based video decoding apparatus.