Research on Block Matching Algorithm with It Application to Decode Error Concealment
|Signal and Information Processing
|error concealment best neighborhood matching motion estimation match criterion search strategy zero vector detection
With the multi-media and wireless network’s increasingly popular, errorconcealment has being a research focus in the field of digital images and video codingtechnology. In recent years, the ideas of using different search strategies to find the bestmatching block for the lost one have been successfully applied to the error concealmenttechnology at the decoder and become the new concerns. How to construct a searchstrategy and search pattern, which is accurate, robust and fast, becomes an urgentquestion. Based on the analysis and summary of the relevant research, the contributesmainly in the following three aspects:Firstly, some traditional spatial error concealment methods are emulated, such asbilinear interpolation, direction interpolation and bezier curve fitting. After that, weanalyze which image type they fit well. Then combined the three methods together andan spatial error concealment method in intra frame which based on bezier curve isproposed. The experiment results show the feasible of the proposed algorithm.Secondly, in order to improve the speed of error concealment in intra frame, amodified best neighborhood matching algorithm is proposed. First of all, inconsideration of the difference between texture images and smooth images, we setdifferent initial thresholds for them. Then, the matching criterion is simplified and thesearching process is changed. Last, the proposed algorithm is emulated. Experimentalresults demonstrate the effectiveness of the algorithm in concealment the smooth imagesand a good quality in texture images are gotten.Finally, based on zero vector detection, this paper do research on a smallcross-triangle temporal error concealment algorithm. First of all, according to the motionvector’s centre-biased and directional centre-biased distribution in the real worldsequences, the reversal idea of diamond search is proposed and the loss blocks aredivided into three types. Then, using the proposed template of small cross-triangle, thebest matching block is searched. Combining the technology of zero vector detection, thelost blocks are concealed in temporal. Experimental results demonstrate that the proposed template is rationality and assure the quality of the real world sequences.