The Application Of Deformable Block Matching Algorithm

My thesis integrates a kind of new motion estimation and compensation algorithm (Deformable Block Matching Algorithm——DBMA) into existing video compression encoding standard MPEG-2. In addition, I have adjusted and optimized the whole video codec system.At first, my thesis introduces basic characteristics and correlative basic knowledge of MPEG-2, and especially introduces the principle of traditional motion estimation and compensation. Secondly, it analyses the essential attribute of video signal and explain the weakness of translational model adopted by block matching algorithm (BMA), and introduces a nodal-displacement-based deformation model and a nodal-search-based deformable block matching algorithm (DBMA) which is more adaptable to non- translational motion. And then, it integrates DBMA into MPEG-2, and the encoder and decoder are adjusted, due to the motion vector parameters in DBMA, in order to construct an integrated video codec system. Finally, it optimizes whole codec system, including PMVFAST_Plus, Fast_DBMA, Two_Mode, SGE, CS, Edge_Filter and Half_Pel, which make our codec system more efficient.Experimental results show that in the case of integer pixel accuracy, the same processing time of motion estimation and compensation and the same target code ratio, Fast_DBMA+PMVFAST_Plus+TM+SGE+CS+Egde_Filter algorithm is super to traditional full search algorithm based on BMA, whether on subjective quality or objective quality (Peak Signal to Noise Ratio——PSNR of reconstruction image of Fast_DBMA+PMVFAST_Plus+TM+SGE+CS+Egde_Filter algorithm is improved 1.86dB). In the case of half pel accuracy, the same processing time of motion estimation and compensation and the same target code ratio, not only can Fast_DBMA+PMVFAST_Plus+TM+Egde_Filter algorithm break through the objective quality limit of full search algorithm (FS) based on BMA, but also subjective video effect of Fast_DBMA+PMVFAST_Plus+TM+Egde_Filter algorithm is improved.