Intra-Key-Frame Coding and Side Information Generation Schemes in Distributed Video Coding

摘要

In this thesis investigation has been made to propose improved schemes for intra-key-frame coding and side information (SI) generation in a distributed videoudcoding (DVC) framework. From the DVC developments in last few years it hasudbeen observed that schemes put more thrust on intra-frame coding and betterudquality side information (SI) generation. In fact both are interrelated as SIudgeneration is dependent on decoded key frame quality. Hence superior qualityudkey frames generated through intra-key frame coding will in turn are utilized toudgenerate good quality SI frames. As a result, DVC needs less number of parityudbits to reconstruct the WZ frames at the decoder. Keeping this in mind, we haveudproposed two schemes for intra-key frame coding namely,ud(a) Borrows Wheeler Transform based H.264/AVC (Intra) intra-frame codingud(BWT-H.264/AVC(Intra))ud(b) Dictionary based H.264/AVC (Intra) intra-frame coding using orthogonaludmatching pursuit (DBOMP-H.264/AVC (Intra))udBWT-H.264/AVC (Intra) scheme is a modified version of H.264/AVC (Intra)udscheme where a regularized bit stream is generated prior to compression. Thisudscheme results in higher compression efficiency as well as high quality decodedudkey frames. DBOMP-H.264/AVC (Intra) scheme is based on an adaptiveuddictionary and H.264/AVC (Intra) intra-frame coding. The traditional transformudis replaced with a dictionary trained with K-singular value decomposition (K-SVD)udalgorithm. The dictionary elements are coded using orthogonal matching pursuitud(OMP).udFurther, two side information generation schemes have been suggested namely,ud(a) Multilayer Perceptron based side information generation (MLP - SI)ud(b) Multivariable support vector regression based side information generationud(MSVR-SI)udMLP-SI scheme utilizes a multilayer perceptron (MLP) to estimate SI framesudfrom the decoded key frames block-by-block. The network is trained offline usingudtraining patterns from different frames collected from standard video sequences.udMSVR-SI scheme uses an optimized multi variable support vector regressionud(M-SVR) to generate SI frames from decoded key frames block-by-block. LikeudMLP, the training for M-SVR is made offline with known training patterns apriori.udBoth intra-key-frame coding and SI generation schemes are embedded inudthe Stanford based DVC architecture and studied individually to compareudperformances with their competitive schemes. Visual as well as quantitativeudevaluations have been made to show the efficacy of the schemes. To exploit theudusefulness of intra-frame coding schemes in SI generation, four hybrid schemesudhave been formulated by combining the aforesaid suggested schemes as follows:ud(a) BWT-MLP scheme that uses BWT-H.264/AVC (Intra) intra-frameudcoding scheme and MLP-SI side information generation scheme.ud(b) BWT-MSVR scheme, where we utilize BWT-H.264/AVC (Intra)udfor intra-frame coding followed by MSVR-SI based side informationudgeneration.ud(c) DBOMP-MLP scheme is an outcome of putting DBOMP-H.264/AVCud(Intra) intra-frame coding and MLP-SI side information generationudschemes.ud(d) DBOMP-MSVR scheme deals with DBOMP-H.264/AVC (Intra)udintra-frame coding and MSVR-SI side information generation together.udThe hybrid schemes are also incorporated into the Stanford based DVCudarchitecture and simulation has been carried out on standard video sequences.udThe performance analysis with respect to overall rate distortion, number requestsudper SI frame, temporal evaluation, and decoding time requirement has been madeudto derive an overall conclusion.