CN104469369A - Method for improving scalable video coding (SVC) performance by using decoder side information (DSI) - Google Patents

Abstract

The invention provides a method for improving scalable video coding (SVC) performance by using decoder side information (DSI). The method includes the steps that decoding reference frames are added to an SVC-DSI encoder to generate a decoder side information creation (DSIC) module, and extra decoding reference frames are provided for a base layer and an enhancement layer in the encoding process, wherein encoding of frames I and frames P refers to SVC standards in the encoding process, but the DSIC module is used for providing a decoding reference frame for encoding of frames B; the processing procedure of the DSIC module on the base layer is independent of the processing procedure of the DSIC module on the enhancement layer, and information of the enhancement layer cannot be used by the base layer or a lower enhancement layer. By means of the method, the effect of improving video frequency distortion can be achieved, and the quality of the decoding reference frames will directly influence rate-distortion performance of videos. The decoding reference frames are generated by adopting the motion compensated frame interpolation (MCFI) technology. Compared with the prior art, the method has the advantages that video frequency distortion performance is promoted better, and the quality of encoded videos is better.

Description

A kind of method utilizing decoding end information to improve SVC performance

Technical field

The present invention relates to a kind of method improving video quality, particularly relate to a kind of method utilizing the extraneous information of decoding end to improve video quality.

Background technology

Now, the application as digital videos such as multimedia messages, visual telephone, video conference, HDTV (High-Definition Television) is more and more extensive.In such applications, network and terminal capabilities different, and likely to change along with time variations.SVC coding provides the scalability of time, space and quality layers, better can adapt to the environment of various change.Meanwhile, there is another kind of popular coding standard distributed coding.But distributed coding standard uses alternating binary coding prediction, not by means of initial video frame information, also there is no practical encoder at present.

At present, also have some researchs by prediction and distributed coding strategy and H.264/AVC Video coding combine, improve the distortion performance of entirety.But research is not applied under scalable video background.The present invention is directed to the problems referred to above, propose a kind of new method, utilize the extraneous information of decoding end, improve coding efficiency.

Summary of the invention

In order to solve problem in prior art, the invention provides a kind of decoding end information that utilizes to improve the method for scalable video performance.Producing video decode reference frame by the decoded information generation module increased in encoder-side, when encoding to the frame of Primary layer and enhancement layer, utilizing the information of reference frame to improve coding efficiency, reduced rate distortion.

The present invention is achieved through the following technical solutions:

A kind of decoding end information DSI that utilizes is to improve the method for scalable video SVC performance, described method produces DSIC module by increasing decoded reference frame in SVC-DSI encoder, extra decoded reference frame (Decoder SI) is provided for it in the cataloged procedure of Primary layer and enhancement layer, wherein, the coded reference SVC standard of I frame and P frame in cataloged procedure, but then utilize described DSIC module to provide decoded reference frame to the coding of B frame; The processing procedure of DISC module is separate at Primary layer and enhancement layer, and the information of enhancement layer can not be used by Primary layer or lower enhancement layer.Can be reached the effect improving video frequency distortion by method provided by the invention, the quality of decoded reference frame directly can affect the distortion performance of video, and method of the present invention adopts MCTI technology to generate decoded reference frame.

The invention has the beneficial effects as follows: what the present invention proposed utilizes decoding end information to improve the method for scalable video performance, and compared with prior art, video frequency distortion performance has larger lifting, the video quality after coding is better.

Accompanying drawing explanation

Fig. 1 is the frame construction drawing of SVC-DSI encoder of the present invention;

Fig. 2 is the cataloged procedure of B frame in the present invention;

Fig. 3 is the flow chart of Decoded Reference frame generating method in the present invention.

Embodiment

Illustrate below in conjunction with accompanying drawing and embodiment the present invention is further described.

As shown in Figure 1, one utilizes decoding end information (Decoder Side Information, DSI) scalable video (Scalable Video Coding is improved, SVC) method of performance, in the encoder of SVC-DSI, increase a decoded information and produce (Decoder SI Creation, DSIC) module.DSIC module provides extra decoded reference frame for it in the cataloged procedure of Primary layer and enhancement layer.In cataloged procedure, the coding of I frame and P frame and SVC standard class are seemingly, but encode to the reference frame that the coding of B frame then utilizes DSIC module to provide, and the coding of Primary layer and enhancement layer independently carries out.MCFI(Motion Compensated Frame Interpolation is adopted in method of the present invention, movement compensating frame interpolation) technology generates decoded reference frame, utilize decoded reference frame can improve rate distortion in video coding process well, improve the quality of video.

Detailed describing is carried out below from following two aspects to technical scheme of the present invention.

1. the coding of pair video B frame

In SVC-DSI method of the present invention, the I frame in all layers and P frame are all encoded according to SVC standard; But, the coding of the B frame in Primary layer and enhancement layer, carries out processing (see accompanying drawing 2) according to the following steps:

Step 1: create decoded reference frame

The video of every one deck (comprising Primary layer and enhancement layer) all associates a Picture Buffer, and the reference frame of video to be decoded is deposited in the inside.Meanwhile, maintain two lists in buffering area---List0 and List1.In the process of decoding, two reference frames are read successively to the video of every one deck, call MCTI method and create decoded reference frame.

Step 2: insert SI frame

The all corresponding image cache district of the video of every one deck (comprising Primary layer and enhancement layer), is placed on the SI frame of every one deck the second place of buffering area List1, after the reference frame namely used in a upper decoding.

Step 3: coding B frame

Current B frame in every one deck (comprising Primary layer and enhancement layer) utilizes the reference frame of buffer memory in Picture Buffer to encode.Best reference frame and best predictive coding module are selected in the process of utilance distortion optimization, and utilance distortion process selects optimum reference frame from multiple decoded reference frame of equivalent layer.(rate-distortion optimization process is here common method)

Step 4: remove decoded reference frame

After the B frame of every one deck has been encoded, this layer of corresponding decoded reference frame has lost value, removes from the Picture Buffer at place.

After Primary layer completes coding, utilize inter-layer prediction tools, by the residual risk of Primary layer, motion vector, the information such as internal data are used to the prediction doing enhancement layer.

DSIC module can be used for producing the additional reference frame of enhancement layer equally.Utilize the frame decoded before each enhancement layer, produce the decoded reference frame of this layer.The decoded reference frame of enhancement layer can improve the compression efficiency of enhancement layer greatly, because compared with the decoded reference frame of low level, comprises more image quality information in high-rise decoded reference frame.Utilize DSIC module, when after the threshold value that the Picture Buffer of certain enhancement layer reaches certain, the information of reference frame just can be utilized to encode.

2. the generation of decoded reference frame

Rate distortion in video coding process is the main cause causing video quality to decline, if in an encoding process, interframe can be utilized to survey reference information, can reduce video distortion rate.The rate distortion caused after the SVC-DSI method that the present invention proposes can lower Video coding to a great extent, thus improve video quality.In the process, owing to needing the information utilizing reference frame, so the effect of decoded reference frame is most important.Therefore, in SVC-DSI method of the present invention, have employed a kind of MCTI method efficiently to generate decoded reference frame.

As shown in Figure 3, detailed process is divided into several step below to the generative process of decoded reference frame:

Step 1: obtain frame of video

Backward and decode forward frame is obtained respectively from two list List0 and List1 of the screen buffer of each layer.

Step 2: by the frame of video of acquisition in step 1 after low pass filter process, as the input of motion estimation algorithm, produces the initial motion vectors between two reference frames after estimation.

Step 3: utilize bi-directional motion estimation algorithm, carries out refinement and standardization in time, improves the accuracy of motion-vector field.

Step 4: utilize weighing vector median filtering algorithm, spatially carries out refinement and standardization, improves the accuracy of motion-vector field.

Step 5: utilize in above-mentioned steps the motion vector information estimating to obtain and reference frame to produce final decoded reference frame.

Primary layer is identical with the generative process of the decoded reference frame in enhancement layer and encoder.The decoded reference frame of final generation is stored in Picture Buffer corresponding to every one deck.

Above content is in conjunction with concrete preferred implementation further description made for the present invention, can not assert that specific embodiment of the invention is confined to these explanations.For general technical staff of the technical field of the invention, without departing from the inventive concept of the premise, some simple deduction or replace can also be made, all should be considered as belonging to protection scope of the present invention.

Claims (7)

1. utilize decoding end information DSI to improve a method for SVC performance, described method produces DSIC module by increasing decoded reference frame in SVC-DSI encoder, in the cataloged procedure of Primary layer and enhancement layer, provide extra decoded reference frame for it; It is characterized in that: the coded reference SVC standard of I frame and P frame in cataloged procedure, but then utilize described DSIC module to provide decoded reference frame to the coding of B frame; Described method comprises the steps:

Step 1: create decoded reference frame: the video of every one deck all associates a Picture Buffer, and the reference frame of video to be decoded is deposited in the inside; Meanwhile, maintain two lists in buffering area---List0 and List1; In the process of decoding, two reference frames are read successively to the video of every one deck, adopt MCTI technology to generate decoded reference frame;

Step 2: insert SI frame: all corresponding image cache district of every one deck, is placed on the SI frame of every one deck the second place of buffering area List1, after the reference frame namely used in a upper decoding;

Step 3: coding B frame: the current B frame in every one deck utilizes the reference frame of buffer memory in Picture Buffer to encode, best reference frame and best predictive coding module are selected in the process of utilance distortion optimization, and utilance distortion process selects optimum reference frame from multiple decoded reference frame of equivalent layer;

Step 4: remove decoded reference frame: after the B frame of every one deck has been encoded, this layer of corresponding decoded reference frame loses value, removes from the Picture Buffer at place;

DSIC module produces the additional reference frame of enhancement layer with same step, utilizes the frame decoded before each enhancement layer, produces the decoded reference frame of this layer.

2. method according to claim 1, is characterized in that: after Primary layer completes coding, utilizes inter-layer prediction tools, and by the residual risk of Primary layer, motion vector, the information such as internal data are used to the prediction doing enhancement layer.

3. method according to claim 1, is characterized in that: when after the threshold value that the Picture Buffer of certain enhancement layer reaches certain, just utilize the information of reference frame to encode.

4. method according to claim 1, is characterized in that: the processing procedure of described DISC module is separate at Primary layer and enhancement layer, and the information of enhancement layer can not be used by Primary layer or lower enhancement layer.

5. method according to claim 1, is characterized in that: the quality of described decoded reference frame directly affects the distortion performance of video, and described method adopts MCTI technology to generate decoded reference frame.

6. method according to claim 5, is characterized in that: described MCTI technology comprises the steps: to generate decoded reference frame

Step 1: obtain frame of video, obtains backward and decode forward frame respectively from two list List0 and List1 of the screen buffer of each layer;

Step 2: by the frame of video of acquisition in step 1 after low pass filter process, as the input of motion estimation algorithm, produces the initial motion vectors between two reference frames after estimation;

Step 3: utilize bi-directional motion estimation algorithm, carries out refinement and standardization in time, improves the accuracy of motion-vector field;

Step 4: utilize weighing vector median filtering algorithm, spatially carries out refinement and standardization, improves the accuracy of motion-vector field;

Step 5: utilize in above-mentioned steps the motion vector information estimating to obtain and reference frame to produce final decoded reference frame.

7. method according to claim 5, is characterized in that: the generative process of Primary layer and enhancement layer decoded reference frame is in the encoder and the decoder identical, and the final decoded reference frame generated is stored in Picture Buffer corresponding to every one deck.