Research On Distributed Video Coding Based On Wavelet Domain

In addition to the common video application business scenarios such as traditional streaming video on demand and broadcasting,a large number of emerging video applications(such as wireless video surveillance,wireless digital cameras,and mobile video phones)are rapidly developing,and traditional video coding is no longer applicable.The emergence of Distributed Video Coding(DVC)has attracted widespread attention.The coding system only needs to use video redundancy to achieve coding performance similar to conventional video coding.It migrates the more complex video inter-frame prediction from the video encoding end to the decoding end,with better fault tolerance,simple encoding,and relatively complicated decoding,which makes it very suitable for wireless video terminals with limited computing power,limited memory and power.Provides a new solution for the above emerging video services.This thesis mainly studies the side information refinement and related noise modeling techniques of distributed video coding.The main research contents are as follows:(1)This article analyzes the theoretical to application of distributed video coding technology from the perspective of basic principles,coding framework,key technologies and development trends.The distributed video coding framework based on wavelet transform is adopted(2)By analyzing the characteristics of various side information(SI)generation algorithms and multi-resolution motion estimation,this thesis introduces non-local means(MRM)based on the Multi-resolution Motion Refine(MRMR)algorithm.Nonlocal Means(NLM)method to refine high-frequency subbands.At the decoding end,a Multiresolution Side Information Refine(MRSIR)algorithm is used to enable the decoder to optimize side information from the airspace.Based on the current Motion refinement for frame low-resolution reconstruction.Improved encoding efficiency for higher resolution data,thereby improving SI quality.(3)Aiming at the problem that the Laplacian distribution cannot accurately describe the correlation noise modeling,this thesis proposes a progressive correlation noise optimization(PCNO).The correlation noise residuals are updated through the reconstruction coefficients obtained from the decoded bit plane,and then the bit plane is re-decoded,the reconstruction coefficients are classified in the refinement process,and the optimal reconstruction coefficients are estimated according to the classification results.Then the correlation noise residuals are refined to improve the accuracy of the correlation noise residuals and the RD performance of the system.(4)This thesis analyzes the proposed distributed video coding architecture in terms of peak signal to noise ratio(PSNR),coding rate and rate distortion(RD)performance,and compares it with traditional Video coding H.264 / AVC,H.264-NoMotion,MRMR and MRSIR-LM schemes were compared.The experimental results show that this coding framework is better than mrsir-lm and mrmr.The most important point is that it can greatly improve the subjective effect of the video.