KR20010069019A - Method of video coding and decoding for video codec - Google Patents

Abstract

PURPOSE: A video coding and decoding method is provided to decrease the number of bit needed in a video compression coding by estimating motions of blocks included by a main region(or a background region) after dividing a full screen of animation into the main region and the background region. CONSTITUTION: In a coding process, an estimation vector(MV(Bk)) of a block to be coded is found out in an input image(Frame)(301). The input image(Frame) is divided into a main region(Fm) and a background region(Fb), and an overall motion vector(MV(Fm))(or MV(Fb)) of the main region(Fm)(or the background region(Fb)) is found out(302). Next, by coding the difference value(delta(Bk)) of the overall motion vector(MV(Fm) or MV(Fb)) and a motion vector(MV(Bk)) of a pertinent block(Bk), the difference value(delta(Bk)) of the overall motion vector(MV(Fm) or MV(Fb)) is coded instead of the motion vector(MV(Bk)) of a pertinent block(Bk)(303).

Description

Image encoding and decoding method {METHOD OF VIDEO CODING AND DECODING FOR VIDEO CODEC}

The present invention relates to a compression encoding method and a decoding method of a moving picture in a system that processes and encodes and decodes an image by a block unit. Then, the total motion estimation is performed using the area important for semantic transmission as the main area and the area not as the background area, and then the motion estimation of the blocks belonging to the main area (and / or the background area) is performed based on this. The present invention relates to a video encoding and decoding method capable of reducing the number of bits required for video compression encoding.

More specifically, the present invention provides a background image and a main image area in a technique of compressing a video by block motion prediction and compensation including standards such as H.26x, MPEG-1, MPEG-2, MPEG-4, etc. A method of encoding and decoding an image which can enhance the performance of image compression encoding by allocating the entire motion vector to each region by dividing it into the (main region) and then predicting the motion vector of each block based on the total motion vector. In particular, the present invention relates to a video encoding and decoding method for improving the performance of a system in which a background area and a main area (a face of a user) are relatively clearly separated, such as a video conferencing system (PVS) for a PC.

Currently, in H.261, H.263, MPEG-2, MPEG-4, etc., which are set as international video coding standards, the screen is divided into small blocks such as 16 × 16 or 8 × 8 to move each block. After performing the prediction (Motion Estimation) and the compensation (Motion Compensation), the residual signal is encoded through a discrete cosine transform (DCT) or the like.

In this case, motion prediction and compensation for each block may be a representative method used to use temporal correlation in a video.

Therefore, the reason for motion prediction and compensation in standard and other image encoding techniques is to fully utilize the temporal correlation of a video, and to utilize the information available from the previous frame to the maximum in encoding the current frame. to be.

In the video coding standards such as H.263, MPEG-2, and MPEG-4, since motion vectors between blocks are more likely to be similar to motion vectors of adjacent blocks, the motion vectors of adjacent blocks are estimated using the motion vectors. The difference values with the estimated values are encoded. For example, after the motion vector is estimated, the difference values with the estimated values are encoded using a variable length coding technique (VLC) or the like.

Therefore, in a variable length coding technique, the smaller the magnitude of the difference value between the estimated value and the estimated vector value is, the closer the zero vector is to the zero vector, The length of the codeword for indicating the motion vector becomes shorter as the length is arranged to be smaller, and the closer the difference with the neighboring vector value for displaying the motion vector of the block is to the zero vector.

As described above, since the motion vector is encoded by using the difference value of each block between the original image and the moving image, the data amount is increased, and the increase of the data amount is suppressed, and the data amount is reduced, thereby increasing the amount of data in the same time. Research has been carried out to enable the transmission of the faithful.

In particular, in the case of a PC-based video conferencing system, if the amount of data is reduced, the number of frames that can be transmitted within the same time can be increased, and thus a video conferencing system closer to a more complete video will be possible.

The present invention relates to a compression encoding method and a decoding method of a moving picture in a system that processes and encodes and decodes an image by a block unit. Then, the total motion estimation is performed using the area important for semantic transmission as the main area and the area not as the background area, and then the motion estimation of the blocks belonging to the main area (and / or the background area) is performed based on this. An image encoding and decoding method for reducing the number of bits required for image compression encoding is provided.

Particularly, the present invention divides an image into a main area and a background area in consideration of the fact that the face of the user is meaningful information compared to the background in view of the nature of the video conferencing system. A method of performing motion estimation and predicting, encoding and decoding motion vectors of blocks belonging to a main region and / or a background region based on the total motion vectors is provided.

1 is a diagram showing an example of motion vector estimation for explaining the present invention;

2 is a diagram showing an example of division of a main area and a background area for explaining the present invention;

3A is an encoder block diagram of an image encoding algorithm for explaining the present invention.

3B is a decoder block diagram of an image decoding algorithm for explaining the present invention.

The present invention provides a video encoding and decoding method which processes, encodes and decodes an image by block unit.

Separating the image to be encoded into a main region and a background region which are important for semantic transmission, obtaining a total motion vector by performing total motion estimation on the separated region, and a total motion vector obtained by the total motion estimation. The video encoding method is performed by encoding a motion vector of a block belonging to a corresponding region.

In addition, the present invention provides a video encoding and decoding method for processing, encoding and decoding a video by block unit,

Inputting the entire motion vector information for each region by separating the image to be encoded into a main region and a background region that are important for semantic transmission, and moving the block belonging to the region based on the input total motion vector information. An image decoding method comprising the process of decoding a vector.

The present invention is also characterized by estimating the total motion vector for the main region, and encoding a difference value between the full motion vector and the motion vector of the block for the blocks belonging to the main region as motion vector information of the block. do.

In addition, the present invention is characterized by estimating the total motion vector for the main region, and decoding the sum of the total motion vector and the motion vector of the block for the block belonging to the main region as the motion vector information of the block. do.

In addition, the present invention is characterized by estimating the total motion vector for the background region, and encoding the difference value between the total motion vector and the motion vector of the block for the blocks belonging to the background region as the motion vector information of the block. do.

In addition, the present invention is characterized by estimating the total motion vector for the background region, and decoding the sum of the total motion vector and the motion vector of the block for the blocks belonging to the background region as the motion vector information of the block. do.

In addition, the present invention is characterized in that the video is encoded and decoded with the main area as the face area of the user and the remaining area as the background area when using the characteristics of the PC-based video conferencing system.

That is, in the present invention, after performing the entire motion estimation by dividing the area into the main area and the background area in the PC-based video conferencing system, the blocks belonging to each area are estimated by the motion vectors based on the motion vectors. A video encoding and decoding method of a PC-based videoconferencing system is provided to reduce the number of bits for encoding.

As described above, the motion vectors between blocks in motion prediction and compensation are more likely to have a higher correlation with blocks belonging to the same object.

This is due to the assumption that all the pixels in the block have the same motion vector in the process of estimating the motion vectors of the blocks in the above-described international standards for image compression encoding (see Fig. 1).

That is, the motion prediction and compensation used in the current standard are based on the assumption that the objects in the image are translated.

Referring to FIG. 1, when following the assumption that the moving object 102 in the image 101 performs translational transition as a whole as a mass, the motion vectors MVx, MVy, ... of pixels belonging to the object 102 are taken. Is assumed to have the same motion vector.

Blocks found to belong to the same object due to the above reasons are very likely to have similar characteristics in motion vectors.

This gives similar movement characteristics not only in the case of the assumption that the object is translated, but also in the case of rotation, enlargement or reduction.

In order to fully utilize these characteristics, many techniques using global motion estimation have been studied.

That is, many attempts have been made to reduce the size of the motion vector as a whole by estimating the motion vector of each block based on the motion vector obtained through the overall motion estimation after performing the overall motion estimation.

As shown in FIG. 1, blocks belonging to one object 102 have motion vectors MVx, MVy, ... which have the same or very similar characteristics to each other, and blocks having the motion vectors of the same object. The motion compensation is performed by importing pixels of the corresponding parts in the previous frame (screen).

Therefore, the more blocks belonging to the same object have a higher probability of having the same motion vector.

Therefore, if the same object or blocks belonging to the same area are estimated by the motion as a whole, and the motion estimation between the blocks based on the estimated motion vector as a whole, the probability of the size of the motion vectors is very high.

However, as described above, in the PC-based video conferencing system, since only one user is used in most cases, it is easy to distinguish the user's face area from the background area in the transmitted image, and there is almost no movement of the background area. Most of the user's faces move in the same direction.

In addition, in the videoconferencing system, almost all the importance is concentrated on the user's face, and the background image is not so important.

Therefore, by allocating the entire motion vector to the user's face (main area) and performing motion estimation of blocks belonging to the main area, the number of bits required to encode the motion vector is reduced, and the optimal motion vector is searched. It can also reduce the time it takes.

If necessary, the motion region may be estimated as if it is performed in the main region after allocating the entire motion vector to the background region.

In the present invention, a case in which the entire motion vector is allocated to only the main region will be described as an example.

The extension to the background area is the same as for the main area.

As shown in Fig. 2, one screen to be encoded is called F.

At this time, the screen F is divided into a main area and a background area, and these areas are called Fm and Fb, respectively.

The overall motion estimation is performed on the main region Fm and the background region Fb thus separated, and the motion vectors of the respective regions obtained as the result are called MV (Fm) and MV (Fb), respectively.

The block to be encoded is referred to as Bk, and the motion vector of the block to be encoded is referred to as MV (Bk).

Then, it is determined whether the block Bk currently encoded belongs to the main region Fm, and if the block Bk belongs to the main region Fm (Bk ∈ Fm), the motion vector for the block Bk is determined. Instead of directly encoding the motion vector MV (Bk) for the block, Δ (Bk) = MV (Bk)-MV (Fm), where the main region full motion vector (MV (Fm)) The difference value of the motion vector MV (Bk) of the block Bk is encoded.

At this time, Δ (Bk) has a high probability of showing a distribution smaller than the values of MV (Bk).

The reason for this has been explained earlier.

In addition, the decoder obtains a motion vector MV (Bk) for the block Bk by the following process.

That is, as MV (Bk) = Δ (Bk) + MV (Fm), the motion vector (Bk) of the corresponding block Bk is obtained by summing the total main motion vector MV (Fm) and the difference value Δ (Bk). MV (Bk)) can be obtained.

In this way, the main region and the background region have already been separated in the previous stage of motion estimation, and the map information (divided information of the main region and the background region) necessary for this is encoded in the partitioning step. No additional information is needed to indicate whether to use or not.

3A and 3B are shown as block diagrams of the processes to be performed as a whole.

First, the encoding process will be described with reference to FIG. 3A.

The motion vector MV (Bk) of the block Bk to be encoded in the input image Frame is obtained (301).

In addition, the main region Fm and the background region Fb are separated from the input image Frame, and the entire motion vector MV (Fm) of the main region Fm (and / or the background region Fb) (and / or MV (Fb)) is obtained (302).

Then, with respect to the block Bk, the main region Fm (and / or the background region Fb) of the entire motion vector MV (Fm) (and / or MV (Fb)) and the corresponding block Bk. By encoding the difference value Δ (Bk) with the motion vector MV (Bk), instead of the motion vector MV (Bk) of the corresponding block Bk belonging to the main region (and / or background region), the Δ (Bk) (303)

The decoding process corresponds to an inverse process of encoding, which is shown in FIG. 3B.

The block is divided into a main region and a background region by inputting the motion vector information Δ (Bk) of the entire motion vector MV (Fm) and / or MV (Fb) and the block Bk belonging to the region. Decodes the motion vector MV (Bk) of (Bk) (304).

That is, for the main region Fm, the motion vector MV (Bk) of the block Bk belonging to the main region is decoded by Δ (Bk) + MV (Fm), and the previous figure is also shown for the background region Fb. If the encoding process is performed as in 3a, the motion vector of the corresponding block belonging to the background region is decoded by Δ (Bk) + MV (Fb).

The present invention can be used in all kinds of coding standards and algorithms for motion prediction and compensation.

In view of the fact that the blocks belonging to one object have similar motion characteristics, the present invention has proposed an algorithm capable of encoding these motion vectors more efficiently.

In particular, the present invention utilizes the fact that the user is one person in most cases in the PC-based videoconferencing system, and therefore the main area is easily estimated as the user's face, so that the entire motion vector is also applied to the main area and the background area if necessary. By allocating, the number of bits used to encode a motion vector can be reduced.

Therefore, the present invention can improve the compression coding performance by using the characteristics of these systems in the videoconferencing system currently used.

Claims (5)

A video encoding and decoding method for processing, encoding, and decoding a video by block unit, Dividing an image to be encoded into a main region that is important for semantic transmission and a background region that is not important to the encoding; obtaining a total motion vector of a corresponding region by performing a total motion estimation on a required region among the separated regions; And encoding a motion vector of a block belonging to the corresponding region based on the entire motion vector obtained by the method. 2. The method of claim 1, wherein a total motion vector of the corresponding region is estimated for the main region, the background region, or the main region and the background region, and the total motion vector and the motion vector of the block belong to the block belonging to the region. And encoding a difference value of? As motion vector information of the block. The method according to claim 1 or 2, wherein when the image to be encoded is an image processed by a PC-based video conferencing system, the face region of the user is extracted and the user face region is the main region, and the rest is the background region. The video encoding method, characterized in that for performing encoding. A video encoding and decoding method for processing, encoding, and decoding a video by block unit, Separating the image to be encoded into a main region that is important for semantic transmission and a background region that is not important, and inputting the entire motion vector information for the required region among the separated regions, and based on the inputted whole motion vector information, And decoding a motion vector of the belonging block. 5. The method of claim 4, wherein a total motion vector of the corresponding region is estimated for the main region, the background region, or the main region and the background region, and the total motion vector and the motion vector of the block belong to the block belonging to the corresponding region. And decoding the sum of the values as motion vector information of the block.