Research Of The Key Techniques Of Video Coding/Decoding Based On Fuzzy Metric

H.264/AVC is the most advanced video coding standard in recent years, which covers awide range of application from low bits communication to HDTV. It has developed a lot of newcoding methods and strategies. The new changes improve the coding efficiency considerably, butunfortunately increase the computation complexity to a large extent, which means large amount oftime and resource occupation for H.264/AVC coding system. Therefore it is necessary to developefficient fast algorithms, in order to decrease the complexity of the system while keepingnecessary performance. More new methods are also needed to improve the coding efficiency.Besides, error concealment is especially important for error resilience decoder to recover thecorrupt image.This dissertation has conducted the research focusing on some key technologies in videoencoding and error concealment based on H.264/AVC, the work of the dissertation is composed ofencoding algorithm and error concealment research listed as follows:(1) Proposed a class of fuzzy metric to measure the image distortion based on fuzzy integraland the character of video coding. The fuzzy metric can be used in video processing.Experimental results demonstrated its ability to measure the image distortion.(2) Proposed a new method of prediction mode decision in intra frame encoding with thepresent fuzzy metric. Experimental results show that the encoder with proposed fuzzymetric can reduce the bit rate by average 6.52% and achieve about 0.1dB PSNR gain.(3) Based on the analysis of motion activity and scene change detection information ofmacroblocks in inter frame, we proposed a methodology to simplify the computation ofencoder while maintaining the coding efficiency. According to the results of analysis, thealgorithm skips the prediction modes that benefit little to the coding efficiency. Simulationresults demonstrate its valuable efficiency of considerably reducing the encodingcomplexity while incurring little loss in quality and increase in bit rate.(4) To efficiently recover the lost motion vectors of corrupt macroblocks, we select theproposed fuzzy metric as distortion metric in temporal error concealment, and thenintegrate the proposed algorithm into H.264/AVC decoder. Experimental results fromdecoding several types of video sequences have shown that our scheme can considerablyimprove the visual quality of reconstructed images and achieve the PSNR gain at differentframe. (5) Proposed a scene change detection based error concealment methodology for H.264/AVC.According to the results of scene change detection, the fuzzy metric based temporalconcealment with boundary match criteria and spatial error concealment scheme isadaptively adopted. Simulation results demonstrate the robustness of proposedmethodology and its considerable improvement in visual quality of reconstructed pictures.(6) Proposed an efficient temporal error concealment algorithm that composes of 4×4 and16×16 motion vectors recovery. For some video objects with fast movement, the motionvectors within a small area are more correlative than large area. Among the 4×4 and16×16 motion vectors, the one with the minimal fuzzy metric is selected as the finalmotion vector. Simulation results show that the subjective and objective performances ofthe proposed technique are superior to conventional temporal concealment techniques.