KR100378795B1 - Flexible direct mode coding method for b pictures in moving picture - Google Patents

Abstract

The present invention relates to a technique for more effectively implementing bit saving using a direct skip mode of a B picture in a video coding system. The present invention includes a first process of determining a direct mode as an optimal prediction mode in a target macroblock of a B picture; A second step of examining a prediction mode used in a macroblock at the same position of the next P picture P2 based on the B picture; As a result of the above investigation, when the mode of the P2 picture is the skip mode or the 16 × 16 mode, the direct mode is regarded as a skip, and the remaining data is not transmitted. When the mode of the P2 picture is another mode, the residual coding is performed through the residual coding. This is accomplished by a third process that allows data to be sent.

Description

FLEXIBLE DIRECT MODE CODING METHOD FOR B PICTURES IN MOVING PICTURE}

The present invention relates to a technique for defining a direct skip mode of a B picture based on object motion in a moving picture coding system. In particular, the present invention relates to a complete constraint by presenting a luminance constraint and a chrominance constraint applied to an encoding process of a next P picture. The present invention relates to a flexible direct mode coding method of B pictures in a guaranteed video.

Typically, H.26L uses one variable length code (VLC). Therefore, the direct skip mode cannot be used in the code table for the B picture. In other words, a macroblock having a direct mode must unconditionally transmit a coded block pattern (CBP) and a quantized coefficient value regardless of the existence of residual data. Here, the remaining data means a result of subtracting the macroblock predicted from the original macroblock.

For example, if the current macroblock is using the direct mode, and the remaining data value "0" which is the difference between the original macroblock and the macroblock predicted by the direct mode is "0", the encoder is a macroblock layer. In the macroblock type, code number 0 (1 bit) indicating the direct mode and cbp = 0 in the block layer must be included in the bit stream.

As described above, in the H.26L B picture coding method according to the prior art, a macroblock having a direct mode unconditionally transmits CBP and quantization coefficient values irrespective of the existence of residual data. There was a problem of increasing the bit rate.

Accordingly, an object of the present invention is to provide a method of selectively determining a direct skip mode based on a prediction mode of a next P picture instead of newly defining a direct skip mode in a B picture code table of H.26L. .

It is still another object of the present invention, when the macroblock of the B picture is determined to be in the direct mode, when the next P picture mode is the skip mode or the 16 × 16 mode, the next skip is not regarded as a direct skip and the remaining data is not transmitted. When the picture is in another mode, the present invention provides a method for transmitting the remaining data.

1 is a diagram illustrating a skip mode of a next picture P2.

2A and 2B are explanatory diagrams showing an example in which a target macroblock has a different distribution in a B picture when P2 is in a skip mode;

3 (a)-(h) are exemplary diagrams showing object motion between macroblocks at the same positions of P1, B, and P2 pictures.

4 is a motion vector explanatory diagram in the direct mode.

*** Description of the symbols for the main parts of the drawings ***

P1, P2: P picture B: B picture

According to an aspect of the present invention, there is provided a method of flexible direct mode coding of a B picture, comprising: a first process of determining a direct mode as an optimal prediction mode in a target macroblock of a B picture; A second process of examining a prediction mode used in a macroblock at the same position of a neighboring next P picture P2 based on a B picture; As a result of the above investigation, based on the case in which the mode of the P2 picture is the skip mode or the 16 × 16 mode, the direct mode is divided into a case of skip and a case of skip. It consists of a third process for transmitting the remaining data, described in detail with reference to Figures 1 to 4 attached to the operation of the present invention as follows.

In the present invention, the direct skip mode that does not exist in the B picture code table in the encoder and the decoder is selectively determined from the prediction mode used in the next P picture so that bit saving is possible as much as possible. Therefore, the current target macroblock of the B picture is intra prediction, forward prediction, backward prediction, bi-direction prediction, direct prediction. Assume that the direct mode is determined as an optimal prediction mode having the smallest SAD (Sum of Absolute Difference) value.

Thus, assuming that the direct mode is determined to be the best prediction mode in any macroblock of the current B picture, whether or not to perform residual coding for the selected direct mode is located at the same position of the next P picture. The prediction mode used in the macroblock is viewed and determined. If the prediction mode used is skip or 16 × 16, then the macroblock of picture B uses direct skip; if another prediction mode is used for the next P picture, the picture B uses the flexible direct mode with residual coding. This will be described in detail as follows.

First, a case in which the macroblock of the next P picture is the skip mode will be described.

Suppose that a macroblock at the same position as the target macroblock in the next P picture used the skip mode. In this case, the skip mode of the next P picture is a prediction mode used in a P picture defined when 16 × 16 block size is used and zero motion and the remaining data are “0”.

In FIG. 1, the macro block of the P2 picture is the same as the macroblock of the P1 picture, and since the target macroblock of the B picture has a high probability of having the same characteristics as the macroblocks of the P1 picture and the P2 picture in time, the target macroblock is directly skipped. It is very likely to be used. Therefore, the target macroblock is defined to use direct skip when the macroblock of the next P picture P2 at the same position as the target macroblock is declared skip.

However, there may be a case where the target macroblock has a different characteristic from the macroblocks of P1 and P2, but FIG. 2 shows an example.

When the B picture has a different distribution from the adjacent P pictures P1 and P2 as shown in FIG. 2, the target macroblock of the B picture should use a direct mode that uses residual coding instead of direct skipping so as to be closer to the target macroblock in the decoder. Can be reconfigured. However, this is a violation of the direct skip definition of the present invention. This is because the target macroblock of the B picture should not select the direct skip mode, even though the P2 picture uses the skip mode. Therefore, when decoding each macroblock of a P2 picture as a method for guaranteeing a direct skip definition of the present invention, a constraint is placed on the case shown in FIG.

As a method for detecting the case as shown in FIG. 2, a motion detection technique is used. For example, the P1 picture and the B picture (or B and P2) may be compared to recognize that an object motion occurs between two macroblocks. As a method for detecting the object motion, the absolute value difference between the pixels and the pixels of the two macroblocks is respectively determined, and when any one of these values is detected to be greater than the threshold (eg, 20), the object motion is generated.

Therefore, in FIG. 2, motion exists between two macroblocks of a P1 picture and a B picture, and a motion exists between the B picture and the P2 picture. Therefore, when encoding each macroblock of the P2 picture, if a motion such as that shown in FIG. 2 is detected in the motion detection process, a mandatory restriction is applied to predict a 16 × 8 mode or less.

This is to ensure that the direct skip definition of the present invention is ensured by disabling the P2 picture even though it can be originally used in the skip mode. In the encoder, however, the P1 picture uses the already decoded macroblock, the P2 picture is the original macroblock to be coded, and the target macroblock of the B picture is the original target macroblock stored in the buffer for coding after the P2 picture. Use

On the other hand, even if the macroblock of the P2 picture does not use skip, Figure 3 shows another technique for determining direct skip from the object motion between P1 and the macroblock of the P2 picture at the same position as the target macroblock of the B picture. It demonstrates with reference.

3 (a) shows a case in which all three macroblocks have similar characteristics, and (b) (c) shows a case where the target macroblock of the B picture and each macroblock of the P2 or P1 picture have similar characteristics. In this case, recalling that the prediction mode of the target macroblock is selected as a direct mode, in the case of another prediction mode, particularly (a), although the two macroblocks at the same position as the target macroblock have similar characteristics, forward prediction, The fact that the direct prediction mode is selected rather than backward prediction or bidirectional prediction means that the result predicted by the direct mode is most similarly predicted to the original target macroblock as compared to other prediction modes. Similarly, in case of (b), the macroblock of the P2 picture has similar characteristics to the target macroblock, but in direct mode instead of backward prediction, in case of (c), the macroblock of the P1 picture has similar characteristics to the target macroblock. Has been selected but in direct mode rather than forward prediction and the meaning is the same.

Therefore, in such a case, it is possible to reduce the bit rate by using the direct skip mode instead of the direct mode for residual coding. Of course, some degradation in image quality may occur compared to residual coding, but it is negligible compared to the amount of bits saved by the direct skip mode. In the present invention, if a macroblock having similar characteristics to a target macroblock of a B picture exists in a P1 or P2 picture, the present invention intends to use the direct skip mode.

On the contrary, in the remaining cases (d to h) of FIG. 3, it can be seen that an object motion exists between the P1 picture and the B picture, and the B picture and the P2 picture in common. As such, the existence of object motion means that there is no similar property between two macroblocks. Thus, there is no basis for determining how similarly the result predicted by the direct mode is predicted to the original target macroblock. Therefore, in the case of (d) to (h), direct prediction must be performed through the remaining coding.

In the present invention, in order to define the direct skip mode based on the object motion of FIG. 3, the target macro when the macroblock of the P2 picture satisfies (a) to (c) and uses the prediction mode at 16 × 16 simultaneously. The block was predicted in direct skip mode. In this case, as shown in FIG. 4, the 16 × 16 mode has a zero or non-zero motion vector, from which a direct forward motion vector and a direct backward motion vector are derived.

If the macroblock of the P2 picture is the same as (d) to (h) of FIG. 3, in order to guarantee the direct skip definition based on the object motion of the present invention, when the P2 picture is encoded, the motion detection method of FIG. After detecting the cases d) to (h), a mandatory restriction was made to predict the 16 × 8 mode or less so that the target macroblock of the B picture is not declared as a direct skip.

Meanwhile, the luminance constraint for selecting the direct skip mode and the chrominance constraint for preserving color information will be described as follows.

The video used in H.26L is divided into a luminance component (Y) that provides all video information including motion and a chrominance component (U, V) that provides color information.

As described above, in the present invention, the direct skip mode is selectively determined according to the prediction mode used in the macroblock of the P2 picture for bit saving. That is, if the macroblock of the P2 picture uses the skip mode or the 16 × 16 prediction mode, the target macroblock of the B picture causes prediction to be performed in the direct skip mode.

However, as shown in FIG. 2, when object motion is detected between the P1 picture and the B picture, and object motion is detected between the B picture and the P2 picture, even if the macroblock of the P2 picture is in the skip mode, 16x8 or less may be forced. It was limited to determine the prediction mode of. Also, when the macroblock of the P2 picture uses the 16 × 16 mode, when all object motions are detected as shown in FIGS. 3D to 3H, the prediction is forced to 16 × 8 or less mode as in the skip mode. Limits were made to determine the mode.

Since the selection of the direct skip mode based on the object motion uses the luminance component of the video, the constraints for guaranteeing the direct skip definition become the luminance constraint.

If there is a sudden color change between image frames, and a direct macro skipping condition is used in the luminance component and the target macroblock of the B picture uses the direct skip mode, there is a problem that the color component is not properly reconstructed. Therefore, if the chrominance of the macroblocks of the P1 and P2 pictures at the same position as the target macroblock of the B picture is examined, and a pixel changes more than the threshold (for example, 60), the macro of the P2 picture is the same as the luminance constraint. The block is forced to determine the prediction mode to a mode of 16 × 8 or less.

As described in detail above, the present invention does not newly define the direct skip mode in the B. picture code table of H.26L, but instead defines the direct skip mode based on the prediction mode of the next P picture. There is an effect that can maximize the coding efficiency of H.26L that can not.

In addition, by providing a luminance constraint and a chrominance constraint to define the direct skip mode of the B picture based on the object motion, there is an effect of ensuring an acceptable range of image quality.

Claims (8)

A first step of determining a direct mode as an optimal prediction mode in the target macroblock of the B picture; A second step of examining a prediction mode used in a macroblock at the same position of the next P picture P2 based on the B picture; As a result of the above investigation, when the mode of the P2 picture is the skip mode or the 16 × 16 mode, the direct mode is set to skip and the remaining data is not transmitted. When the P2 picture uses another mode, the residual coding is performed through the residual coding. A flexible direct mode coding method of a B picture in a video, characterized by comprising a third step of transmitting data. The method of claim 1, wherein when encoding a next P picture, one of these values is obtained by obtaining an absolute difference between pixels of pixels of a P1 picture and a B picture, or two macroblocks in the same position on the B picture and the P2 picture, respectively. Even if a value above the threshold is detected, it is determined that the object motion has occurred, and the step of forcibly limiting the prediction to the 16x8 mode or less even if the macroblock of the P2 picture is skipped mode further comprises B. Flexible direct mode coding method for pictures. The method of claim 2, wherein when obtaining an absolute difference between pixels of pixels of two macroblocks, the P1 picture uses the already decoded macroblock, the P2 picture uses the original macroblock to be coded, and the target of the B picture. A macroblock is a flexible direct mode coding method of a B picture in a video, characterized by using an original target macroblock stored in a buffer for coding after a P2 picture. 3. The method of claim 2, wherein the threshold is twenty. The method of claim 1, wherein the third process is that although the mode of the P2 picture is not the skip mode, a macroblock having characteristics similar to the target macroblock of the B picture is present in the P1 picture or the P2 picture and simultaneously predicted by 16 × 16. The method of claim 1, wherein the target macroblock further comprises the step of predicting the direct skip when the mode is used. The method of claim 1, wherein the third process is to forcibly determine the prediction mode to a mode of 16 × 8 or less if both object motions are detected in the P1 picture and the P2 picture when the macroblock of the P2 picture uses the 16 × 16 mode. A method of flexible direct mode coding of a B picture in a video, further comprising the step of limiting. The method of claim 1, wherein the third process examines the chrominance of the macroblocks of the P1 and P2 pictures at the same position as the target macroblock of the B picture, and forcibly changes the macroblock of the P2 picture if a pixel changes more than a threshold. And limiting the determination of the prediction mode to a mode of 16 × 8 or less. 8. The method of claim 7, wherein the threshold is 60.