CN1204757C - Stereo video stream coder/decoder and stereo video coding/decoding system - Google Patents

Abstract

The invention relates to a stereoscopic video data compression technology. When the present invention encodes the binocular video streams collected by the stereo camera system, one of them is encoded according to the standards compatible with the MPEG series, and the other image is respectively used for parallax compensation prediction, joint parallax and motion compensation prediction, and at the decoding end. Coded transmission is performed by frame estimation and interpolation recovery method. Among them, the disparity estimation adopts the multi-level segmentation block estimation based on the Markov model, and the method of frame estimation and interpolation is to use the reference frame image restored at the decoding end and the corresponding disparity and motion vector, and use the method based on the frame estimation probability model Estimate and interpolate. The decoding end is divided into two stages of decoding. One stage is to decode only the main video stream to obtain a single video displayed on a common display device. The other stage is to decode all dual video streams. The restored stereoscopic video signal is synthesized by an autostereoscopic display to display a stereoscopic image. .

Description

A Stereoscopic Video Stream Encoder/Decoder and Its Stereoscopic Video Encoding/Decoding System

technical field

The invention relates to moving image processing technology, in particular to a method and device for encoding/decoding stereoscopic video data.

Background technique

When human beings watch the surrounding world, they can not only see the width and height of objects, but also know their depth, and can judge the distance between objects or between the viewer and objects. The main reason for this three-dimensional visual characteristic is that people usually always watch objects with both eyes at the same time, and due to the distance between the visual axes of the two eyes (about 65mm), when the left eye and the right eye look at an object at a certain distance, the received The visual images received are different, so the brain synthesizes the information of the two images through the movement and adjustment of the eyeballs to produce a three-dimensional effect. When viewing an object with only the left eye and the right eye, the resulting image displacement sensation is called parallax.

The three-dimensional visual characteristics derived from the binocular structure provide us with a direct and simple method to obtain the relative depth sense of the real world from the left and right images, and this relative depth information is used in such as remote communication (telemedicine, teleconferencing) ), telerobotics (remote control, autonomous aviation, surveillance), entertainment (interactive HDTV, stereoscopic movies) and virtual reality. But the obvious cost of introducing relative depth information to increase realism is that the amount of data transmitted and stored is more than double that of single-vision systems. In order to meet the increase in data volume, the solutions are nothing more than increasing channel bandwidth, improving channel utilization with efficient protocols, and reducing source code rates with efficient compression techniques. However, due to the uneconomical increase in memory capacity and network bandwidth, effective image compression techniques must be used.

The stereoscopic video coding methods in the prior art essentially use the correlation between binocular video streams to improve the coding efficiency of the two video signals as a whole. There are generally two types of methods. The first type is a stereoscopic video stream encoding method based on the MPEG video coding standard. The basic principle is to encode one video stream separately, while the other video stream is encoded using disparity estimation and compensation technology. . Most of these methods use hybrid coding methods, such as standard-compliant mixed-definition coding (where one stream has relatively poor definition), psychometric-based bit allocation, multi-resolution-based stereo coding, and frame estimation. Interpolation and reconstruction of the right B-frame shooting, the B-frame of the right video stream is not transmitted, but is interpolated and restored at the decoding end) and other methods. The problems of this kind of method are: the efficiency of disparity estimation and compensation needs to be improved; while using the binocular disparity information, the effective use of the motion information of the right channel is ignored, and the overall coding efficiency still has a large room for improvement; the use of frame estimation Although the interpolation technology has a high compression rate for stereo coding, the existing frame estimation interpolation technology is relatively simple, and the quality of the reconstructed image is not ideal; in general, there is still a lack of a mature and perfect stereo coding system.

The second category is the object-based stereo encoding method, whose basic principle is to segment and extract objects in the scene and encode them in combination with motion and binocular depth information. But when there are multiple objects appearing in the scene, the encoding effect of this kind of method is not good. At the same time, due to the complexity of its calculation, the real-time performance is also poor, which is still far from the requirements of real-time system applications.

Contents of the invention

The object of the present invention is to provide a stereoscopic video stream encoder/decoder, which has the advantages of high encoding compression rate, fast decoding rate and compatibility with single video encoding mode.

Above-mentioned purpose of the present invention is achieved through the following technical solutions:

A stereoscopic video stream encoder comprising:

The main video stream encoding unit is used to encode one video stream in the stereoscopic video stream according to the MPEG protocol to generate the main video stream;

A secondary video stream coding unit, which includes:

The disparity/motion compensation estimation unit is used to disparity the corresponding intra-coded frames and inter-frame predicted frames in another video stream in the stereoscopic video stream by using the intra-coded frames and inter-frame predicted frames in the main video stream compensation estimation, and use the previous intra-frame coding frame and/or inter-frame prediction frame in the secondary video stream to perform motion compensation estimation on the current inter-frame prediction frame in the secondary video stream, and the initial value of the disparity compensation estimation is obtained in the following manner: using Perform motion compensation estimation on the disparity field of the previous intra-coded frame or inter-frame prediction frame through the motion vector obtained by performing motion compensation estimation on the inter-frame prediction frame in the auxiliary video stream, and use the new disparity field as the disparity compensation estimation Initial value:

A compensated prediction coding unit, configured to encode the disparity compensation estimation information of the intra-coded frame in the secondary video stream, the disparity compensation estimation information and the motion compensation estimation information of the inter-frame prediction frame to generate the secondary video code stream;

The multiplexer is used to time-division multiplex the main video code stream and the secondary video code stream to generate a stereoscopic video code stream.

Preferably, in the aforementioned stereoscopic video stream encoder, the disparity/motion estimation unit performs disparity estimation based on a layered Markov probability model and a multi-level block matching method. More preferably, in the hierarchical Markov probability model and the overlapping block matching method, the hierarchical level is set to two levels, and the size of the divided block is divided into two types: 8×8 and 16×16.

Preferably, in the stereoscopic video stream encoder, the transmission channel bandwidth occupied by the secondary video code stream is adjusted by changing the DCT quantization coefficient of the residual image in the parallax compensation estimation information.

A stereoscopic video stream decoder comprising:

A demultiplexer is used to decompose the stereoscopic video code stream into a main video code stream and an auxiliary video code stream;

The main video code stream decoding unit is used to decode the main video code stream according to the MPEG protocol to generate the main video stream;

A secondary video code stream decoding unit, which includes:

The disparity/motion compensation prediction unit is used to reconstruct the intra-coded frame of the secondary video stream according to the intra-frame coded frame and inter-frame predicted frame in the primary video stream and the parallax compensation estimation information and motion compensation estimation information contained in the secondary video code stream and intra picture prediction frame;

The frame estimation and interpolation unit is used for bidirectionally predicting/interpolating frames according to the corresponding main video stream, intra-coded frames and inter-frame predicted frames in the secondary video stream, and the parallax compensation estimation information and motion contained in the secondary video stream Compensate estimated information to reconstruct bidirectionally predicted/interpolated frames in the secondary video stream;

The auxiliary video stream reconstruction unit is used to sequence the intra-coded frames and intra-image prediction frames reconstructed by the disparity/motion compensation prediction unit and the bi-directional prediction/bi-directional interpolation frames reconstructed by the frame estimation and interpolation unit in order to generate auxiliary video streams. video stream.

Preferably, in the above-mentioned stereoscopic video stream decoder, the frame estimation and interpolation unit uses a stereoscopic frame estimation and interpolation method based on the Bayesian minimum cost equation to reconstruct bidirectionally predicted/interpolated frames.

In the stereoscopic video stream encoder of the present invention, because only one of the video streams is encoded with high quality according to the MPEG standard, only a few frames are encoded in the other video stream, and the remaining frames are then completely "skipped". The decoding end performs frame estimation, interpolation and recovery, thus greatly improving the coding efficiency and saving the transmission bandwidth.

The object of the present invention is to provide a stereoscopic video processing system, which has the advantages of high encoding compression rate, fast decoding rate and compatibility with single video encoding mode.

Above-mentioned purpose of the present invention is achieved through the following technical solutions:

A video processing system, including a camera that ingests left and right video streams, a time base corrector that synchronizes the video streams output by the two cameras in time, and two channels of video that have been time-synchronized and processed by the time base corrector Frame sequential multiplexer for stream multiplexing to form stereoscopic video stream, computer system comprising stereoscopic video encoder as claimed in claim 1 and stereoscopic video stream decoder as claimed in claim 5 and general display and stereoscopic displays,

Among them, when only a single video stream needs to be transmitted, the main video stream encoding unit of the stereoscopic video stream encoder encodes the video image and sends the signal to the transmission channel; when two video streams need to be transmitted, the main video stream The encoding unit and the auxiliary video stream encoding unit respectively encode the left and right channel video images and send the signals to the transmission channel. When the received video stream only contains one video stream, the main video stream decoding unit of the stereoscopic video stream decoder Decode the encoded code stream and send the decoded signal to the ordinary display. When the received code stream contains left and right video stream signals, the main video code stream decoding unit and the auxiliary video code stream decoding unit respectively decode the left and right two code streams. The stream is decoded and the decoded signal is sent to the stereoscopic display.

In addition to the advantages of high coding efficiency and low transmission bandwidth requirements, the video system of the present invention is also compatible with existing single-video MPEG serial coding standards. This reduces system upgrade costs while maintaining encoding compatibility, and provides flexible control over stereoscopic display quality.

Description of drawings

Through the following description of the preferred embodiments of the present invention in conjunction with the accompanying drawings, you can further understand the purpose, features and advantages of the present invention, wherein:

FIG. 1 is a schematic diagram of a stereoscopic video stream encoder/decoder according to the present invention.

FIG. 2 is a schematic diagram of a video processing system using a stereoscopic video stream encoder/decoder according to the present invention.

Detailed ways

Preferred embodiments of the present invention are described below in conjunction with the accompanying drawings.

FIG. 1 is a schematic diagram of a stereoscopic video stream encoder/decoder according to the present invention. As shown in Figure 1, the stereoscopic video stream encoder 1 is responsible for encoding the input left and right video streams. For the convenience of description, it is assumed that the left video stream is the main video stream and the right video stream is the secondary video stream, but this assumption should not It should be understood as a limitation of the present invention, but in fact it can also be a contrary assumption. The video code stream encoded by the stereoscopic video stream encoder 1 is transmitted to the stereoscopic video stream decoder 3 through the channel 2 .

Referring to FIG. 1 , a stereoscopic video stream encoder 1 includes an MPEG encoder 4 as a primary video stream encoding unit, a multiplexer 7 , and a secondary video stream encoding unit composed of a disparity/motion compensation estimation unit 5 and a compensated prediction encoding unit 6 .

MPEG digital video coding technology is essentially an image compression method that uses the statistical redundancy of video sequences in the time and space directions. It relies on the correlation between pixels (interpel), including such an assumption: that is, in There is simple correlated translational motion between successive frames. Therefore, the value of a pixel on a particular picture can be predicted based on nearby pixels in the same frame using intra-frame coding techniques, or can be predicted based on pixels in nearby frames using inter-frame coding techniques.

When the shot of a video sequence changes, the temporal correlation between pixels in nearby frames is small or even disappears. At this time, intra-frame coding technology should be used to develop spatial correlation to achieve effective data compression. In the MPEG compression algorithm, the discrete cosine transform (DCT) coding technology is used to effectively develop the spatial correlation between the nearby pixels of the same picture in units of 8×8 pixel picture blocks. The following will be based on the intra-frame coding technology A compressed image frame is called an intra-coded frame, and is abbreviated as I ^M or I ^A , where the superscripts M and A represent the main video stream and the auxiliary video stream, respectively.

If there is a large correlation between pixels in nearby frames, that is, when the content of two consecutive frames is very similar or the same, you can use inter-frame DPCM coding technology based on temporal prediction (inter-frame motion compensation prediction) , hereinafter, the image frames that can be compressed according to the inter-frame coding technology are called inter-frame predictive frames, and are abbreviated as ^PM or ^PA , where the superscripts M and A represent the primary video stream and the secondary video stream, respectively.

The MPEG standard also introduces an image frame called a bidirectional predictive frame, which can be restored by using the past frame and the future frame as a reference frame, but it cannot be used as a reference frame. This type of image frame is called a bidirectional predictive frame below. , and abbreviated as B ^M or B ^A , where the superscripts M and A represent the primary video stream and the secondary video stream, respectively.

In the present invention, the MPEG encoder 4 encodes the left video stream according to the MPEG standard to generate the main video code stream, which is composed of encoded I ^M , ^PM and B ^M frame sequences arranged in a certain order .

As shown in the figure, both the left and right video streams are input to the disparity/motion estimation unit 5 in the secondary video stream coding unit, and disparity and motion estimation are performed in this unit. Specifically, the synchronous or corresponding intra-frame coding frames I ^M and I ^A in the main video stream and the auxiliary video stream are compared with the inter-frame prediction frames P ^M and ^PA to obtain an image frame I A or I ^A in the auxiliary video stream. Parallax estimation of ^PA ; compare the previous intra-frame coding frame ^IA or inter-frame prediction frame ^PA of the secondary video stream with the current inter-frame prediction frame ^PA to obtain motion estimation for the current inter-frame prediction frame. The reason why motion estimation information and parallax estimation information are provided for each ^PA frame is because, in general, the best prediction result can be obtained by mixing motion and parallax compensation, so in the present invention, in order to make the decoder recover To produce higher quality image frames, the disparity/motion estimation unit 5 provides motion estimation information for a ^PA frame (by comparing the previous reference frame I ^A frame or ^PA frame with the current ^PA frame in the same video stream) ) and disparity estimation information (according to the corresponding ^PA and ^PM frames), which can effectively solve the problem of reduced coding efficiency due to temporal occlusion and parallax occlusion.

There are many methods for disparity estimation. In the present invention, the disparity/motion estimation unit 5 performs disparity estimation based on hierarchical Markov probability model and multi-level block matching. The advantage of this method is that a smooth and relatively accurate disparity field can be obtained, which will greatly reduce the entropy of the disparity compensation residual image, thereby further improving the compression rate. In order to be compatible with the block size of the MPEG standard, when using the above-mentioned hierarchical Markov probability model and overlapping block matching method, the hierarchical level is set to two levels, and the block size is divided into two types: 8×8 and 16×16. kind.

Motion Compensated Estimation is a time-based DPCM coding prediction technique, which is widely used in MPEG1 and MPEG2 video coding standards. The concept of motion compensation is based on the estimation of motion between video frames, that is, if all objects in the video shot have a displacement in space, then limited motion parameters (for example, for translational motion of pixels, available motion vector to describe) to describe inter-frame motion. Because the spatial correlation between some motion vectors is usually high, sometimes it can be considered that a motion vector represents the motion of an adjacent pixel block, so a frame of picture can be divided into several pixel blocks (one pixel in MPEG1 and MPEG2 standards The block is 16×16 pixels), and only one motion vector representing each pixel block is estimated, coded and transmitted. Since only the prediction error picture (the difference between the original picture and the motion compensated predicted picture) is coded, temporal redundancy between frames is reduced.

Practical observations show that for time-continuous stereoscopic video images, their disparity fields also have a high degree of time redundancy, so in the present invention, it is better to obtain the initial value of the disparity compensation estimate in the following manner: at first for P Perform motion compensation prediction on frame ^A to obtain the motion vector, and then perform motion compensation prediction on the disparity field of the previous reference frame I ^A (or ^PA ) in the same video stream, so that the new disparity field can be used as the initial value of disparity estimation . This method can greatly reduce the time required for encoding the auxiliary video stream and improve the encoding speed.

The compensation prediction coding unit 6 is connected to the disparity/motion estimation unit 5, and it encodes the disparity estimation compensation information of the ^IA frame obtained by the disparity/motion estimation unit 5 and the disparity estimation compensation information or the motion estimation compensation information of the ^PA frame to generate Secondary video code stream. The coded ^IA frame parallax estimation compensation information bit stream is divided into three parts: a parallax vector stream, a parallax compensation residual image and a quadtree structure, wherein the parallax vector stream adopts differential pulse coding (DPCM) encoding, and the residual Images are coded using discrete cosine transform (DCT) and scalar quantization methods.

The multiplexer 7 is connected with the MPEG encoder 4 and the compensation predictive encoding unit 6, and generates a stereoscopic video stream by time-division multiplexing the main video stream and the auxiliary video stream. In the present invention, in order to improve the encoding efficiency, all bidirectional prediction/interpolation frames ( ^BA frames) in the auxiliary video stream are not subjected to any encoding process, nor are they sent to the multiplexer 7 as a part of the auxiliary video code stream for channel 2 on the transmission.

In the above-mentioned stereoscopic video stream encoder, by changing the DCT quantization coefficient of the residual image after parallax compensation, the additional bandwidth of the transmission channel can be flexibly changed to meet stereoscopic display under various bandwidth requirements.

Referring to Fig. 1 again, the stereoscopic video stream decoder 3 includes an MPEG decoder 9 as a main video stream decoding unit, a demultiplexer 7, and a disparity/motion compensation prediction unit 10, a frame estimation and interpolation unit 11 and a secondary video stream The reconstruction unit 12 constitutes a secondary video stream decoding unit.

As shown in Figure 1, the demultiplexer 8 decomposes the stereoscopic video code stream transmitted on the channel 2 into a main video code stream and an auxiliary video code stream and provides the main video stream to the MPEG decoder 9 and provides the auxiliary video stream to the A disparity/motion compensated prediction unit 10 and a frame estimation and interpolation unit 11 .

The MPEG decoder 9 decodes the main video code stream according to the MPEG protocol to generate the main video stream, which is composed of restored I ^M , PM ^and B ^M frame sequences arranged in a certain order.

The disparity/motion compensation prediction unit 10 is also connected with the MPEG decoder 9, the frame estimation and interpolation unit 11 and the secondary video stream reconstruction unit 12, and it is based on the intraframe coding frame ^IM and frame The inter-prediction frame P ^M and the disparity estimation compensation information and motion estimation compensation information contained in the secondary video stream output by the demultiplexer 8 reconstruct the corresponding intra-frame coding frame I ^A and inter-frame prediction frame ^PA in the secondary video stream, The reconstructed I ^A frame and ^PA frame are output to the frame estimation and interpolation unit 11 and the secondary video stream reconstruction unit 12 .

The frame estimation and interpolation unit 11 is also connected with the MPEG decoder 9 and the auxiliary video stream reconstruction unit 12, which is based on the corresponding bidirectional prediction frame B ^M in the main video stream output by the MPEG decoder 9 and the corresponding intra-frame B M in the auxiliary video stream. Coding frame I ^A and inter-frame prediction frame ^PA (such as the I ^A frame and ^PA frame adjacent to the B ^A ) and the disparity estimation compensation information and motion estimation compensation information contained in the secondary video code stream reconstruct the two-way of the secondary video stream The predicted/interpolated frame, and its reconstructed B ^A frame is output to the auxiliary video stream reconstruction unit 12.

In the secondary video stream reconstruction unit 12, the intra-coded frame I ^A and the inter-frame prediction frame PA reconstructed by the disparity/motion compensation prediction unit 10 and the bidirectional prediction/interpolation frame B ^A reconstructed by the frame estimation and interpolation ^unit follow the acquisition time sequence to generate secondary video streams.

Since most of the secondary video streams are ^BA frames, the reconstruction speed and image quality of ^BA frames are very important in the stereo codec structure. To this end, a frame estimation method is adopted in the present invention, which is based on the Bayesian minimum cost equation stereo frame estimation and interpolation method (SFEI_BLCF). This method utilizes the motion, parallax and image information (indicated by the arrow shown in dotted line in Fig. 2) obtained at the decoding end and the characteristics of the stereoscopic video sequence itself to quickly synthesize B ^{and A} frames, and reconstruct the image in the sense of stereo vision is of acceptable quality. The specific reconstruction steps are as follows:

(1) Since the B ^A frame is interpolated between the I ^A and P ^A frames, the motion vector between the I ^A and P ^A frames is scaled according to the distance from the B ^A frame to the I ^A frame to determine the I ^A The position of the pixel in the frame in the B ^A frame.

(2) For the same pixel, if the difference between the pixel values in the corresponding B ^M , I ^A and ^PA frames is less than the set value, it will be regarded as a visible area, and the weighted average of these pixel values The value is used as the value of the corresponding pixel in the ^BA frame, and the motion vector of the pixel pointing to the ^IA and ^PA frames in the ^BA frame and the disparity vector pointing to the ^BM frame are recorded.

(3) For the same pixel, if the difference between the pixel values in the corresponding B ^M , I ^A and ^PA frames is greater than or equal to the set value, the pixel is regarded as an occlusion point, and in its neighborhood Select one of the motion vectors related to each pixel in the visible area of the pixel as the matching motion vector, and map to the corresponding image frame according to this motion vector to obtain the final pixel value of the point.

As can be seen from the above, in the stereoscopic video stream encoder/decoder of the present invention, the B ^M frame of the main video stream, the I ^A and ^PA frames of the auxiliary video stream are used as reference frames for inter-frame compensation prediction, and all need to be coded and transmitted . However, during decoding, the motion and disparity vector values obtained at the decoding end can be directly used to recover and reconstruct the secondary video stream B ^A frame without matching search, so the present invention has the characteristics of high coding compression rate and fast decoding speed.

Fig. 2 shows a schematic diagram of the video processing system of the present invention. As shown in Figure 2, the video processing system includes two cameras 21a and 21b that capture the left and right video streams respectively, a time base corrector 22 connected to the cameras, and a frame sequential multiplexer connected to the time base corrector 22. A multiplexer 23, a computer system 24, a common display 25 and a stereoscopic display 26, wherein the computer system 24 includes the aforementioned stereoscopic video encoder and stereoscopic video stream decoder.

In the above-mentioned video processing system, when encoding, the left and right video streams output by the two cameras 21a and 21b are time-synchronized by the time base corrector 22 and then output to the frame sequential multiplexer 23 for multiplexing. After the stereoscopic video stream is formed, it is sent to the computer system 24 . When only one video stream is input into the computer system 24 or when only a single video stream needs to be transmitted, the video image is encoded by the main video stream encoding unit of the stereoscopic video stream encoder and the MPEG standard code stream signal is sent to the transmission channel, when needed When transmitting two video streams, the main video stream encoding unit and the auxiliary video stream encoding unit of the stereoscopic video stream encoder encode the left and right video images respectively and send the signals containing the main video code stream and the auxiliary video code stream to the transmission channel.

The decoding is completed by the stereoscopic video stream decoder of the computer system 24. When the received video stream only contains one video stream, the main video stream decoding unit of the stereoscopic video stream decoder decodes the coded stream and sends the decoded signal to For an ordinary display, when the received code stream contains left and right video stream signals, the main video code stream decoding unit and the auxiliary video code stream decoding unit of the stereoscopic video stream decoder decode the left and right two code streams respectively and decode The signal is sent to an autostereoscopic display.

The effect of the present invention is illustrated below with a specific application example. Assuming that the image frame is in CIF format (352×288), the main video stream is encoded according to the syntax standard of MPEG encoding. The image quality of this channel is relatively high (the average peak signal-to-noise ratio PSNR is about 35dB), and the encoding rate is 0.14MbS ~2.55MbS. Only a few frames in the auxiliary video stream are predictively encoded and transmitted, and the remaining frames are completely "skipped". The average coding rate of the stream is 14.8Kbs~108Kbs. Through comparison, it can be seen that the additional bandwidth required to transmit the auxiliary video stream is extremely low, so that the total bit stream of stereoscopic digital TV is only about 1.15 to 1.3 times that of normal single-view digital TV transmission bit stream. Although the image quality of the auxiliary video stream is slightly lower than that of the main video stream (the average peak signal-to-noise ratio (PSNR) is about 30dB), this left and right images with mixed resolutions can fully utilize the characteristics of the human visual system (HumanVisualsystem, HVS) and corresponding stereoscopic displays synthesize stereoscopic images with high visual clarity and sufficient depth perception.

Claims (7)

1. A stereoscopic video stream encoder, characterized in that, comprising: The main video stream encoding unit is used to encode one video stream in the stereoscopic video stream according to the MPEG protocol to generate the main video stream; A secondary video stream coding unit, which includes: The disparity/motion compensation estimation unit is used to perform disparity compensation on the corresponding intra-coded frames and inter-frame predicted frames in another video stream in the stereoscopic video stream by using the intra-frame coded frames and inter-frame predicted frames in the main video stream estimate, and use the previous intra-coded frame and/or inter-frame predicted frame in the secondary video stream to perform motion compensation estimation on the current inter-frame predicted frame in the secondary video stream, wherein the initial value of the parallax compensation estimate is obtained in the following manner: Perform motion-compensated estimation on the disparity field of the previous intra-coded frame or inter-predicted frame using the motion vector obtained by performing motion-compensated estimation on the inter-predicted frame in the secondary video stream, and use the new disparity field as the disparity-compensated estimate initial value; A compensated prediction coding unit, configured to encode the disparity compensation estimation information of the intra-coded frame in the secondary video stream, the disparity compensation estimation information and the motion compensation estimation information of the inter-frame prediction frame to generate the secondary video code stream; The multiplexer is used to time-division multiplex the main video code stream and the secondary video code stream to generate a stereoscopic video code stream. 2. The stereoscopic video stream encoder according to claim 1, wherein the disparity/motion estimation unit performs disparity estimation based on a hierarchical Markov probability model and a multi-level block matching method. 3. stereoscopic video stream encoder as claimed in claim 2, is characterized in that, in described layered Markov probability model and overlapping block matching mode, hierarchical level is set as two levels, and segmented block size is divided into two stages. There are two types of 8×8 and 16×16. 4. The stereoscopic video stream encoder according to any one of claims 1-3, wherein it adjusts the transmission occupied by the auxiliary video code stream by changing the DCT quantization coefficient of the residual image in the parallax compensation estimation information channel bandwidth. 5. A stereoscopic video stream decoder, characterized in that, comprising: A demultiplexer is used to decompose the stereoscopic video code stream into a main video code stream and an auxiliary video code stream; The main video code stream decoding unit is used to decode the main video code stream according to the MPEG protocol to generate the main video stream; A secondary video code stream decoding unit, which includes: The disparity/motion compensation prediction unit is used to reconstruct the intra-coded frame of the secondary video stream according to the intra-frame coded frame and inter-frame predicted frame in the primary video stream and the parallax compensation estimation information and motion compensation estimation information contained in the secondary video code stream and intra picture prediction frame; The frame estimation and interpolation unit is used for bidirectionally predicting/interpolating frames according to the corresponding main video stream, intra-coded frames and inter-frame predicted frames in the secondary video stream, and the parallax compensation estimation information and motion contained in the secondary video stream Compensate estimated information to reconstruct bidirectionally predicted/interpolated frames in the secondary video stream; The secondary video stream reconstruction unit is used to sort the intra-coded frames and intra-image prediction frames reconstructed by the disparity/motion compensation prediction unit and the bi-directional prediction/bi-directional interpolation frames reconstructed without frame estimation and interpolation in order of time to generate secondary video stream. 6. The stereoscopic video stream decoder according to claim 5, wherein the frame estimation and interpolation unit uses a stereoscopic frame estimation and interpolation method based on the Bayesian minimum cost equation to reconstruct bidirectionally predicted/interpolated frames. 7. A video processing system, characterized in that it includes a camera that takes in the left and right video streams, a time base corrector that synchronizes the video streams output by the two cameras in time, and a time base corrector that passes through the time base corrector. The processed two-way video stream is multiplexed to form a frame sequential multiplexer for a stereoscopic video stream, comprising the stereoscopic video encoder as claimed in claim 1 and the stereoscopic video stream decoder as described in claim 5 computer systems and regular and stereoscopic displays, Among them, when only a single video stream needs to be transmitted, the main video stream encoding unit of the stereoscopic video stream encoder encodes the video image and sends the signal to the transmission channel; when two video streams need to be transmitted, the main video stream The encoding unit and the auxiliary video stream encoding unit respectively encode the left and right channel video images and send the signal to the transmission channel. When the received video stream only contains one video stream, it is decoded by the main video stream of the stereoscopic video stream decoder The unit decodes the encoded code stream and sends the decoded signal to the ordinary display. When the received code stream contains left and right video stream signals, the main video code stream decoding unit and the auxiliary video code stream decoding unit encode the left and right two channels respectively. The code stream is decoded and the decoded signal is sent to the stereoscopic display.

Images