KR100266708B1 - Conditional replenishment coding method for b-picture of mpeg system - Google Patents

Abstract

PURPOSE: A conditional supplementary coding method of a bidirectional predictive picture is provided to improve the coding capacity by applying a conditional supplementary coding method to a bidirectional predictive picture. CONSTITUTION: An intra or a non-intra macro block is determined according to an error dispersive value of an MCP(Motion Compensated Prediction) mode and a dispersive value of a primitive image. An intra coding process is performed when a current macro block is the intra macro block. A predetermined threshold value is compared with an absolute mean error value when the current macro block is the non-intra macro block. A DCT block coding process is performed when the absolute mean error value is larger than the predetermined threshold value. A predictive error is not transmitted when the absolute mean error value is not larger than the predetermined threshold value. A type of the current macro block is determined. A variable length coding for information of the macro block is performed.

Description

CONDITIONAL REPLENISHMENT CODIMG METHOD FOR B-PICTURE OF MPEG SYSTEM

According to the present invention, conditional supplementary coding is applied to a Bi-Directional Predictive Picture (B-Picture) (hereinafter referred to as a B-picture) in a MPEG (MOVING PICTURE EXPERTS GROUP) system with a small prediction error and no transfer error. The present invention relates to a conditional supplementary encoding apparatus and method for a bidirectional predictive picture for improving overall encoding performance.

In general, in order to adaptively encode a frame and a field image, MPEG compensates for movement by causal predictive coding (Pure Predictive Coding) and non-causal predictive coding (Interpolative Coding) to remove the redundancy between the images. Or DCT transform coding and quantization to remove spatial domain redundancy in the prediction error signal, and variable length coding (VLC) of motion vector information in units of 16 x 16 blocks with quantized DCT coefficients, and VBR ( Variable bit rate) is used to control the rate to transmit the bit stream generated at the constant bit rate (CBR).

As described above, MPEG uses three types of pictures, i-picture (Interpolative Picture), P-picture (Predictive Picture), and B-picture, to enable high-compression compression by random access to the encoded bit stream and motion compensation. use.

The I-picture, P-picture, and B-picture are periodically input as shown in FIG. 1, and group I, B, P-picture between two successive I-pictures by GOP (Group of Picture). ) And functions as a unit of random access. Also, a group of pictures between two consecutive reference pictures, i.e., an I or P-picture, is called a SGOP (Sub Group of Picture).

The MPEG shown in FIG. 1 has 11 (N = 12) picture intervals in a GOP, 3 (M = 3) picture intervals in an SGOP, and 11 Ps having (N-1) between I-pictures as reference pictures. Or a B-picture is used, and two consecutive B-pictures are (M-1) individuals.

The I-picture is encoded using only original image data without motion compensation. Since the I-picture is not affected by other pictures, the I-picture serves as an access point for enabling random access to the encoded bit stream. An error propagation may be prevented from being propagated to a subsequent picture.

The P-picture is encoded after motion compensation by causal prediction from previous I-picture or P-picture, and is a cyclic predictive picture. That is, the current P-picture is motion compensated and predicted from the previous P-picture or I-picture, and the current P-picture reconstructed after encoding is a reference for the motion compensation of the next P-picture and B-picture. Used as a picture.

In addition, the B-picture is encoded after motion compensation by non-causal prediction from a previous or subsequent I-picture or a P-picture, and is forward from a previous P-picture or I-picture. Three kinds of motion: motion compensation prediction (MCP), backward motion compensation prediction from a later P-picture or I-picture, and bidirectional motion compensation prediction from a before-and-after P-picture or I-picture Make a reward prediction. In this case, each block of the B-picture determines that the sum of square error (SSE) of the prediction error is the minimum in the three motion compensation predictions as the final motion compensation prediction mode of the block.

Here, the bidirectional motion compensation prediction refers to motion compensation prediction by motion compensation interpolation (MCI).

However, when noise is generated in a P-picture that is used as a reference picture of another P-picture or B-picture, not only the picture quality of the P-picture is degraded, but also the quality of subsequent P-pictures and B-pictures is affected. . That is, since the P-picture is motion compensated and predicted by the previous P-picture or I-picture, an error propagation phenomenon occurs in which noise accumulates by affecting subsequent P-pictures or B-pictures. In addition, since the P-picture is a reference picture for motion compensation prediction of the B-picture, the image quality degradation of the P-picture eventually affects the B-picture, thereby degrading the overall encoding performance.

Accordingly, an object of the present invention is to provide a conditional supplementary encoding method for preventing degradation of image quality by improving conditional supplementary encoding of a B-picture in a macroblock layer and a DCT block layer in an MPEG system.

In order to achieve the above object, the present invention provides an intra macroblock or non-intra (Intra /) according to an error variance value of the Motion Compensated Prediction (MCP) mode and an variance value of an original image determined according to a prediction error of a macroblock unit of a B-picture. Non-intra) a macroblock determination process for determining a macroblock, if the current macroblock is an intra macroblock as a result of the determination, an intra encoding process for performing intra-image encoding, and if the current macroblock is a non-intra macroblock as a result of the determination In the transmission decision process for determining whether to transmit by comparing the absolute average error value of the prediction error of the macro block with a set threshold value, if the absolute average error value is larger than the threshold value, the DCT block in the macro block is determined. The DCT block encoding process of encoding, wherein the determination result is the absolute mean error value If it is not greater than the value, the macroblock decoding process that does not transmit a prediction error in the macroblock, the intra coding process, the DCT block encoding process, and the macroblock decoding process are performed, and then the type of the current macroblock is determined. A conditional supplemental encoding method of a B-picture, which is performed by a macroblock type determination and a VLC process for variable length encoding of macroblock information, is provided.

1 is a picture structure of a typical MPEG picture

2 is a block diagram of an MPEG encoder to which the present invention is applied

3 is a flowchart of a conditional supplemental coding method of a B-picture according to the present invention.

4 is a detailed flowchart of the DCT block conditional supplemental encoding step of FIG.

FIG. 5 is a diagram for explaining a threshold of conditional supplemental coding of FIG. 3. FIG.

FIG. 6 shows the bit amount of each picture allocated by the conditional supplemental coding of FIG.

Drawing shown

7 is a diagram illustrating SNR by conditional supplemental coding of FIG. 3.

<Description of the code | symbol about the principal part of drawing>

10, 110: frame memory 20: motion calculation unit

30: subtractor 40: control unit

50: DCT 60: quantizer

70: dequantizer 80: IDCT

90: adder 100: VLC

120: adaptive predictor

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

As shown in FIG. 2, the MPEG encoder to which the present invention is applied includes a frame / field memory 10 for storing an input video signal in units of frames / fields, and a frame / field output from the frame / field memory 10. The motion calculator 20 calculates a motion by inputting the video signal of the previous frame / field and outputs a motion vector, and inputs the motion vector and the previous frame / field output from the motion calculator 20 as inputs. An adaptive predictor 120 adaptively predicting a next frame / field, a predicted frame output from the adaptive predictor 120 from an image signal of a frame / field output from the frame / field memory 10 The subtractor 30 outputs an error value by subtracting a video signal of a field, and the intra / non-intra is determined according to the error value output from the subtractor 30. The control unit 40 controls to perform supplementary coding on the key part, the DCT unit 50 for transforming the error value to DCT (Discrete Cosine Transform) under the control of the control unit 40, and the quantization for quantizing the output of the DCT unit 50. 60, a VLC unit 100 for variable length coding (VLC) the output of the quantizer 60, an inverse quantizer 70 for quantizing the output of the quantizer 60, An inverse DCT unit 80 that inverses the output of the inverse quantizer 70, an error value output from the inverse DCT unit 80, and a predicted frame / field output from the adaptive predictor 120 An adder 90 that adds a video signal of a frame / field by adding a video signal of a frame, and a frame that stores a video signal of a frame / field output from the adder 90 and outputs the video signal of the frame / field to the adaptive predictor 120. / Field memory 110.

MPEG is a general coding technique that standardizes only common parts common to various digital storage media, and coded representation of video sequences, that is, the syntax of encoded bit strings and decoding therefrom as a minimum standard for compatibility. It is a technique to standardize only the method. Therefore, the structure and performance of the MPEG encoder may vary depending on the application field, and there is much room for improvement.

However, any method for improving performance should be able to encode representation of an existing syntax, and the syntax should not be modified to improve performance. Therefore, the conditional supplemental coding scheme of the B-picture performed by the controller 40 also belongs to the category using the existing syntax.

The syntax for representing the bit stream encoded by the encoder of the MPEG is composed of six layers each having different functions. In the conditional supplemental coding method according to the present invention, MB prediction error in the macroblock (MB) layer is used. From the size we simply decide whether to replenish the macroblocks and DCT blocks.

As shown in FIG. 3, the conditional supplemental encoding method of the B-picture according to the present invention includes a macroblock determination process (200, 210, 220, 230), an intra encoding process (240, 250, 260), a transmission determination process (270), a DCT block encoding process (280, 290, 300). , Macroblock decoding process 310, and macroblock type determination and VLC processes 320 and 330.

The macroblock determination process (200, 210, 220, 230) is an intra macroblock or non-intra (Intra / Intra / Intra / block) according to the error dispersion value of the Motion Compensated Prediction (MCP) mode determined according to the prediction error of the macroblock unit of the B-picture and the dispersion value of the original image Non-intra) to determine the macroblock, the bidirectional motion vector estimation step of estimating the bi-directional motion vector in units of the macroblock of the B-picture, the prediction error of the MCP mode according to the estimated bidirectional motion vector MCP mode determination step 210 to determine the Motion Compensated Prediction (MCP) mode by calculating the SSE (Sum of Square Error), error variance calculation step of calculating the variance value of the error and the original image of the determined MCP mode (220) and compares the calculated error variance with the variance of the original image, and if the calculated error variance is greater than the variance of the original image, the current macroblock is checked. La is determined by the macro block, and carried out by macroblock decision step 230 to determine if the calculated error variance value is greater than the variance of the original image, the current macroblock in non-intra macroblock.

The intra encoding process 240, 250, and 260 performs intra-image encoding when the current macro block is an intra macro block as a result of the determination. In the DCT step, DCT processing an input video signal when the current macro block is an intra macro block as a result of the determination. 240, quantization step 250 of quantizing the DCT processed signal, and VLC step 260 of variable length coding of the quantized DCT coefficients.

In the transmission determining step 270, if the current macroblock is a non-intra macroblock, the transmission result is determined by comparing an absolute average error value MAE of the prediction error of the macroblock with a set threshold TH. The process of judging.

Here, the threshold value TH is adaptively adjusted according to the quantization level of the corresponding macroblock based on the reference threshold value TH0 for the current B-picture.

The reference threshold TH0 increases in proportion to the average MAEavg-b of the absolute average error per frame MAE for the previous encoded B-picture, and in the case of the first B-picture, the encoded previous It increases in proportion to the average (MAEavg-p) of the absolute mean error value (MAE) per frame for the predicted gain of the P-picture.

The DCT block encoding process (280, 290, 300) is a process of encoding a DCT block in a macro block if the absolute mean error value is greater than the threshold value TH as a result of the determination, and the absolute mean error value MAE is determined by the determination result. A DCT step 280 for DCT processing a DCT block in a macroblock if greater than a threshold value TH, a quantization step 290 for quantizing the DCT processed signal, and the quantized DCT coefficients of each DCT block in the macroblock. In the DCT block conditional supplementary encoding step 300 of determining whether or not to transmit each DCT block in the macroblock according to the number Snz of non-zero DCT coefficients, performing encoding and decoding. Is performed by.

That is, the DCT block conditional supplementary encoding step 300 includes the number of non-zero DCT coefficients (Snz) among the quantized DCT coefficients of each DCT block in the macroblock (Snz) and the absolute mean of the corresponding DCT blocks. The six DCT blocks in the macroblock are encoded and decoded by comparing the error value MAEb with the set threshold TH and the set first and second DCT thresholds N1 and N2.

The DCT block conditional supplemental coding step 400, 410, 420, 430, 440, 450, 460 is a non-zero DCT coefficient among DCT coefficients quantized in a corresponding DCT block in the macroblock as shown in FIG. 4. Searching whether the number of times Snz is greater than the set first DCT threshold value N1 (400), and the first DCT threshold value where the number Snz of non-zero DCT coefficients is set as a result of the search. If greater than the value N1, searching for whether the absolute mean error value MAEb of the corresponding DCT block is greater than a value corresponding to 1.5 times the threshold value TH; If the value MAEb is not greater than 1.5 times the threshold value TH, the number of non-zero DCT coefficients Snz is greater than the set second DCT threshold value N2. In step 420, the search results indicate that the absolute mean error value MAEb of the corresponding DCT block is 1.5 times the threshold value TH. Variable length encoding the DCT coefficients of the DCT block when the number Snz of the non-zero DCT coefficients is greater than a corresponding value or greater than the set second DCT threshold value N2 (430). ), If the number Snz of the non-zero DCT coefficients is not greater than the first DCT threshold value N1 or greater than the second DCT threshold value N2, the search result. In step 440 of not encoding the DCT coefficients of the DCT block, and searching for encoding and decoding for all DCT blocks in the macroblock, it is determined whether to encode and decode all DCT blocks in the macroblock. It is performed by step 450 to perform the above steps repeatedly until processing.

That is, the DCT block conditional supplementary coding steps 400, 410, 420, 430, 440, 450, and 460 are larger than the first DCT threshold value N1 of the non-zero DCT coefficients Snz among the quantized DCT coefficients. The absolute mean error value MAEb of the corresponding DCT block is greater than 1.5 times the threshold value TH, or the absolute mean error value MAEb of the DCT block is 1.5 times the threshold value TH. If the number of non-zero DCT coefficients Snz is greater than the set second DCT threshold value N2, the DCT coefficients of the DCT block are variable length coded (430). ), The number of non-zero DCT coefficients Snz is not greater than the set first DCT threshold value N1, or the absolute average error value MAEb of the corresponding DCT block is equal to the threshold value (Snz). Not greater than 1.5 times TH) and the number of non-zero DCT coefficients Snz is greater than the set second DCT threshold value N2. If not, the DCT coefficients of the DCT block are not encoded (440).

Here, the first DCT threshold value N1 is set smaller than the second DCT threshold value N2.

The macroblock decoding process 310 is a process of not transmitting a prediction error in the macroblock unless the absolute average error value is greater than the threshold value TH as a result of the determination.

The macroblock type determination and the VLC process 320 and 330 determine the type of the current macroblock after performing the intra coding process 240, 250 and 260, the DCT block coding process 280, 290 and 300, and the macroblock decoding process 310. Variable length coding is performed on the macroblock information.

The processing of the conditional supplemental coding method of the B-picture according to the present invention performed as described above will be described in detail.

Intra macroblock or non-intra is performed according to the macroblock determination process (200, 210, 220, 230) according to the error variance value of the Motion Compensated Prediction (MCP) mode and the variance value of the original image according to the prediction error of the macroblock unit of the B-picture. / Non-intra) macroblock, which will be described in detail below.

After performing the bidirectional motion vector estimating step 200 for estimating the bidirectional motion vector in units of the macroblock of the B-picture, the SSE of the prediction error for the forward, reverse, and bidirectional MCP modes according to the estimated bidirectional motion vector. Calculate (Sum of Square Error). The MCP mode determination step 210 is performed by selecting and determining the MCP mode having the smallest SSE among the three MCP modes of the forward, reverse, and bidirectional as the MCP mode of the macroblock. That is, the MCP mode of the corresponding macroblock is selected as the MCP mode having the smallest prediction error.

After determining the MCP mode as described above, an error variance value calculating step 220 for calculating the error variance value and the variance value of the original image of the determined MCP mode is performed, and the calculated error variance value and the variance value of the original image are performed. By comparison, if the calculated error variance is greater than the variance of the original image, the current macroblock is determined as an intra macroblock. If the calculated error variance is not greater than the variance of the original image, the current macroblock is a non-intra macroblock. Determining the macroblock determination step 230 is performed.

That is, if the calculated error variance value is larger than the variance value of the original image, the current macroblock is determined as an intra macro block to perform the intra coding process (240, 250, 260), and the calculated error variance value is the variance value of the original image. If it is not larger, the current macroblock is determined as a non-intra macroblock so that the transmission decision process 270 may be performed.

In the intra encoding process 240, 250, and 260, the intra macro block is determined as the intra macro block to perform intra-image encoding. That is, when the current macroblock is an intra macroblock, the input video signal is DCT processed (240), the quantized to quantize the DCT processed signal (250), and the variable length coding (Variable Length Coding) (260).

On the other hand, when the current macroblock is a non-intra macroblock, the transmission decision process 270 is performed, which is a threshold value TH that is set to an absolute mean error value MAE for the prediction error of the macroblock. This is a process of determining whether to transmit the data.

In a macroblock having a small motion compensation prediction error of a B-picture, an overhead bit of the total bit amount of the macroblock increases, which will be described in detail as follows.

When comparing the average bit amount of DCT coefficients per macroblock generated according to the MAE for the prediction error of the macroblock with the quantization level fixed, the bit amount for the DCT coefficient is increased as the MAE increases and the quantization level increases. It increases rapidly as it gets larger. Also, the average number per block of non-quantized '0' DCT coefficients increases with the MAE of the DCT block. For example, if the quantization level is 12 and the MAE is less than or equal to 10, non-zero DCT coefficients occur on average less than 2 per block. On the other hand, the amount of overhead bits for the macroblock information except for the DCT coefficients is almost constant regardless of the magnitude of the MAE or quantization level. That is, in the macroblock of which the MAE of the prediction error is small among the encoded macroblocks that must be transmitted, the proportion of the overhead bits in the total bit amount of the macroblocks increases. Therefore, if any one of the quantized DCT coefficients of the macroblock is not '0', the macroblock is determined as an encoded macroblock. Therefore, transmission of a macroblock with a small MAE is a kind of loss of a transmission bit amount.

Accordingly, as shown in FIG. 5, the absolute average error value MAE of the prediction error of the macroblock is compared with the set threshold TH to determine whether the transmission is performed.

That is, if the absolute mean error value is not greater than the threshold value TH, the macroblock decoding process 310 is performed to not transmit a prediction error in the macroblock. In other words, as shown in FIG. 5, if the MAE for each macroblock is not greater than the set threshold TH, as described above, a loss of the transmission bit amount may occur, so that the transmission is not performed without encoding (310).

Macroblocks forcibly 'uncoded' by conditional supplemental coding in the macroblock layer have the same macroblock's MCP mode and motion vector as the macroblock determined to be 'uncoded' through DCT transformation and quantization. Compared with those of the 'skip' macroblock or 'MC + unsigned' macroblock is determined, and the result is encoded and transmitted.

When the DCT coefficient of the macroblock is not transmitted in the B-picture, the macroblock type is determined as a 'MC + Not Coded' macroblock or a 'Skipped' macroblock. An unsigned macro block means when all DCT coefficients quantized in six DCT blocks in the macro block are '0'. Among the 'unsigned' macroblocks, if the MCP mode and the motion vector (MV) value are the same as the previous macroblock, it is determined as a skipped macroblock, otherwise it is determined as a 'MC + unsigned' macroblock. The skipped macroblock is a macroblock that can be restored using the MCP mode and the motion vector of the previous macroblock even if no information on the macroblock is transmitted. However, the 'MC + unsigned' macroblock is applied to the macroblock type information and the MCP mode. It is a macroblock that can be restored only by transmitting motion information about a mobile station.

The uncoded macroblock of B-picture in MPEG increases with better prediction efficiency, and the prediction efficiency increases with better image quality of the reconstructed reference picture. In addition, as the number of macroblocks determined as uncoded macroblocks increases, the amount of encoded bits generated in a B-picture decreases, and the amount of bits saved can be allocated to subsequent I-pictures and P-pictures.

That is, as shown in FIG. 6, the amount of encoded bits b1 and b2 generated in a B-picture is reduced due to an increase in an uncoded macroblock, and the reduced bit amount is allocated to a subsequent P-picture to be assigned to a P-picture. A large amount of bits can be increased by the reduced amount of bits b1 + b2 = p1. As more bits are allocated to a picture to be encoded, encoding performance is improved, and thus, an increase in uncoded macroblocks has a significant effect on encoding performance.

Therefore, as shown in FIG. 7, as the encoding performance of the P-picture is improved, the encoding performance of the subsequent B-picture is improved, so that not only the B-picture but also the SNR (Signal per Noise Ratio) of the P-picture is increased. .

Here, the threshold value TH is adaptively adjusted according to the quantization level of the corresponding macroblock based on the reference threshold value TH0 for the current B-picture, and the reference threshold value TH0 is encoded. Applied to all macroblocks of a B-picture, it increases in proportion to the average (MAEavg-b) of the absolute mean error value (MAE) per frame for the previous coded B-picture, and the first B-picture. If is increased in proportion to the average (MAEavg-p) of the absolute average error value (MAE) per frame for the prediction gain of the encoded previous P-picture.

The threshold TH for conditional supplemental coding has a great influence on the overall coding performance. As the threshold TH increases, the number of encoded bits of the B-picture decreases because the number of 'uncoded' macroblocks increases. Since the bit amount saved in the B-picture can be allocated to the next I-picture and P-picture, the encoding performance of the I / P picture increases in proportion to the increase of the threshold value TH. However, when the threshold value TH is greater than or equal to a certain threshold, the performance of the B-picture does not apply the conditional supplemental coding method, so that the threshold value TH is lowered than when the threshold value TH is '0' and the block phenomenon is intensified. Extremely, if all macroblocks of a B-picture are 'uncoded' macroblocks, the I / P picture will get the best coding performance, but the coding performance of the B-picture will be equal to the prediction gain.

The reference threshold TH0 is determined from an average MAE (MAEavg-b) and an average threshold (Tavg-b) per frame for the immediately preceding B-picture encoded. In this case, the reference threshold value TH0 is increased in proportion to the average MAE per frame of the B-picture.

However, since the first B-picture of the encoding process does not have a previous B-picture, the reference threshold (TH0) is calculated by the following equation 1 from the average per frame MAE (MAEavg-p) for the prediction gain of the previously encoded P-picture. ) Is determined.

TH0 = MAEavg-p-(M-1)

Here, M is a picture interval between two reference pictures, that is, two I or P-pictures, and (M-1) is MAEavg-p for the prediction error of the P-picture as M increases, so that Factors are taken into account.

In addition, if there is a previous B-picture already encoded, it is assumed that the average MAE of the B-picture is almost equal to the MAE of the current B-picture to be encoded, and is equal to the MAE (MAEavg-b) of the previous B-picture. The reference threshold TH0 is determined by the following equation 2 according to the average threshold Tav-b.

Here, Kcrc is a value that has a decisive influence on the adjustment of the threshold TH. According to the experimental results based on the time when the coding gain of the B-picture is the highest, when M = 2, Kcrc = 1.25 and M ≠ 2 When Kcrc = 1.35, the overall coding gain is maximum.

Meanwhile, the threshold TH of each macroblock is adaptively adjusted according to the quantization level of the macroblock based on the reference threshold TH0 for the current B-picture. That is, as the quantization level increases, the number of non-zero DCT coefficients and the amount of generated bits decrease. Therefore, in order to reduce the number of inefficient 'coded' macroblocks, the threshold TH must be adjusted to be proportional to the quantization level. Assuming that the average quantization level per frame of the current B-picture to be encoded is the same as that of the previous B-picture, the threshold value TH of each macroblock is adjusted as shown in Equation 3 below.

Here, Qmb is the quantization level of the macroblock determined according to the bit rate control of MPEG, and Qavg is the average quantization level per frame of the previous B-picture. Therefore, for the first B-picture, the average quantization level of the P-picture coded just before is used.

Next, if the absolute mean error value is greater than the threshold value TH, the DCT block encoding process 280, 290, 300 is performed to encode the DCT block in the macroblock.

That is, when the absolute mean error value MAE is greater than the threshold value TH, the DCT block in the macro block is DCT-processed (280), and the DCT-processed signal is quantized (290). Thereafter, among the quantized DCT coefficients of each DCT block in the macroblock, whether or not to transmit each DCT block in the macroblock is determined according to the number Snz of non-zero DCT coefficients. A DCT block conditional supplementary encoding step 300 for performing encoding is performed.

The six DCT blocks of the macroblock whose MAE is larger than the threshold TH are each DCT and quantized, and then each DCT block is transmitted according to the number of non-zero DCT coefficients (Snz), that is, encoded. It is determined as a DCT block and an uncoded DCT block which is not encoded.

A DCT block that is forcibly determined to be an 'uncoded' block by conditional supplemental coding in the DCT block layer is encoded in the same manner as a block that is determined to be 'uncoded' through DCT transformation and quantization.

Therefore, even if the conditional supplemental coding method according to the present invention is applied to MPEG, the syntax of the MPEG does not need to be modified at all because there is no change in the representation of the encoded bit string.

As described above, the DCT block conditional supplementary encoding step 300 performed in the DCT block layer includes a number Snz of non-zero DCT coefficients among the quantized DCT coefficients of each DCT block in the macroblock. Encoding and uncoding six DCT blocks in a macroblock by comparing the absolute mean error value MAEb of the DCT block with the set threshold TH and the set first and second DCT thresholds N1 and N2. This will be described in detail with reference to FIG. 4.

First, the number of non-zero DCT coefficients Snz of the quantized DCT coefficients in the corresponding DCT block in the macroblock is searched for greater than the set first DCT threshold value N1 (400). If the number Snz of non-zero DCT coefficients is greater than the set first DCT threshold value N1, the absolute mean error value MAEb of the corresponding DCT block is equal to the threshold value. TH is searched for greater than 1.5 times that of TH) (410). If the absolute mean error value MAEb of the corresponding DCT block is not greater than 1.5 times the threshold value TH, the number of non-zero DCT coefficients Snz In operation 420, the controller searches whether the second DCT threshold value N2 is greater than the set second DCT threshold value N2.

The search results (400, 410, 420) of the quantized DCT coefficients, the number of non-zero DCT coefficients Snz is greater than the set first DCT threshold value N1 and the absolute mean error value of the corresponding DCT block ( MAEb) is greater than 1.5 times the threshold value TH, or the absolute mean error value MAEb of the DCT block is not greater than 1.5 times the threshold value TH and If the number of non-zero DCT coefficients Snz is greater than the set second DCT threshold value N2, the DCT coefficients of the DCT block are variable length coded (430).

In addition, the number of search results 400, 410, and 420 of the non-zero DCT coefficients Snz is not greater than the first DCT threshold value N1 or greater than the second DCT threshold value N2. If it is not large, the DCT coefficients of the DCT block are not encoded (440).

As described above, after encoding and uncoding are determined and processed for one DCT block, conditional supplemental encoding of the DCT block is performed on the remaining DCT blocks in the macroblock as described above.

That is, each of the six DCT blocks in the macroblock determined as the unsigned macroblock is subjected to conditional supplemental coding of the DCT block to determine the uncoding of each DCT block.

On the other hand, the first DCT threshold (N1) and the second DCT threshold (N2) for determining whether to transmit each DCT block is determined by an experiment, because the macro block is used as a target for the B-picture Decreases the encoding performance. Therefore, the experiment is set to an appropriate value to improve the coding performance of the entire picture while improving the coding performance of the B-picture. In this case, the first DCT threshold value N1 is smaller than the second DCT threshold value N2.

As described above, after performing the DCT block encoding process (280, 290, 300), the intra encoding process (240, 250, 260), and the macroblock decoding process (310) including the conditional supplemental encoding step 300 in units of DCT blocks, the macroblock type determination is performed. And VLC processes 320 and 330 to determine the type of the current macroblock and variable length encoding the macroblock information.

That is, as described above, it is determined whether the current macroblock is one of the macroblock types of 'MC + uncoded', 'capsulated' intra coding, and 'MC + coding' (320). (330).

As described above, the conditional supplemental coding method of the B-picture according to the present invention has an effect of improving the overall encoding performance by applying conditional supplemental coding to the B-picture.

Claims (10)

Determined as an intra macroblock or an intra / non-intra macroblock according to an error variance value of a motion compensated prediction mode (MCP) mode determined according to a prediction error of a macroblock unit of a B-picture and a variance value of an original image Macroblock determination process; An intra encoding process for performing intra-image encoding when the current macro block is an intra macro block as a result of the determination; Determining whether the current macroblock is a non-intra macroblock and determining whether to transmit the result by comparing an absolute average error value of the prediction error of the macroblock with a set threshold; A DCT block encoding process of encoding a DCT block in a macro block if the absolute mean error value is greater than the threshold as a result of the determination; A macroblock decoding process that does not transmit a prediction error in the macroblock if the absolute average error value is not greater than the threshold as a result of the determination; And After performing the intra encoding process, the DCT block encoding process, and the macroblock decoding process, the macroblock type determination and the VLC process determine the type of the current macroblock and variable length encode the macroblock information. A conditional supplemental coding method of a B-picture. The method of claim 1, wherein the determining of the macroblocks is performed. A bidirectional motion vector estimating step of estimating a bidirectional motion vector in units of macroblocks of a B-picture; An MCP mode determination step of determining a Motion Compensated Prediction (MCP) mode by calculating a sum of square error (SSE) of a prediction error for an MCP mode according to the estimated bidirectional motion vector; An error variance calculation step of calculating an error variance value of the determined MCP mode and a variance value of an original image; And Comparing the calculated error variance value with the variance value of the original image, if the calculated error variance value is larger than the variance value of the original image, the current macroblock is determined as an intra macroblock, and the calculated error variance value is the variance of the original image. And performing a macroblock determination step of determining the current macroblock as a non-intra macroblock if it is not larger than the value. The method of claim 1, wherein the intra encoding process A DCT step of DCT processing an input video signal when the current macroblock is an intra macroblock as a result of the determination; A quantization step of quantizing the DCT processed signal; And And a VLC step of performing variable length coding on the quantized DCT coefficients. The method of claim 1, wherein the threshold is A conditional supplemental encoding method of a B-picture, characterized in that the adaptive adjustment according to the quantization level of the corresponding macroblock based on the reference threshold value for the current B-picture. The method of claim 4, wherein the reference threshold is A conditional supplementary encoding method of a B-picture, characterized in that it increases in proportion to the average of the absolute mean error value per frame for the previous coded B-picture. The method of claim 5, wherein the reference threshold is A conditional supplemental encoding method of a B-picture, characterized in that it increases in proportion to the average of the absolute average error value per frame with respect to the predicted gain of the previous P-picture. The method of claim 1, wherein the DCT block encoding process A DCT step of DCT processing a DCT block in a macro block if the absolute average error value is greater than the threshold value as a result of the determination; A quantization step of quantizing the DCT processed signal; And DCT block conditional supplemental encoding that performs encoding and decoding by determining whether to transmit each DCT block in the macroblock according to the number of DCT coefficients other than '0' in the quantized DCT coefficients of each DCT block in the macroblock. A conditional supplemental encoding method of a B-picture, characterized in that performed by step. 8. The method of claim 7, wherein the DCT block conditional supplementary encoding step is Among the quantized DCT coefficients of each DCT block in the macroblock, the number of non-zero DCT coefficients and the absolute mean error value of the corresponding DCT block are compared with the set threshold and the set first and second DCT thresholds. A conditional supplemental encoding method of a B-picture, characterized by encoding and uncoding six DCT blocks in a block. 9. The method of claim 8, wherein the DCT block conditional supplemental encoding step is Searching whether the number of non-zero DCT coefficients among the quantized DCT coefficients in the corresponding DCT block in the macroblock is greater than a set first DCT threshold value; Searching whether the absolute mean error value of the corresponding DCT block is greater than 1.5 times the threshold value if the number of DCT coefficients other than '0' is greater than the set first DCT threshold value; Searching whether the number of DCT coefficients other than '0' is greater than a set second DCT threshold value if the absolute mean error value of the corresponding DCT block is not greater than 1.5 times the threshold value; The DCT of the DCT block if the absolute mean error value of the DCT block is greater than 1.5 times the threshold or the number of non-zero DCT coefficients is greater than the set second DCT threshold. Variable length coding the coefficients; Not encoding the DCT coefficients of the DCT block if the number of DCT coefficients other than '0' is not greater than a first DCT threshold value or greater than the second DCT threshold value as a result of the search; And Performing the above steps until all DCT blocks in the macroblock have been determined and processed for encoding and decoding for all DCT blocks in the macroblock, until the encoding and decoding are determined and processed for all DCT blocks in the macroblock. Conditional supplemental coding method of a B-picture, characterized in that. 10. The method of claim 9, wherein the first DCT threshold is set smaller than the second DCT threshold.

Images