| Summary: | 碩士 === 國立雲林科技大學 === 電子與光電工程研究所碩士班 === 101 === Nowadays, the three-dimensional multimedia is becoming to be used gradually,thus,the H.264/MPEG-4 AVC extension application is developed,it call Multiview Video Coding( MVC ),because some transmission environments are unfriendly now,which usually happened in the wireless internet or the place there are much interence with noise and the transmission packets will lose because of the internet environment is congested,the packets which cause the received bitstream can’t decode correctly are lost and people will feel umcomfortable if they are looking for.So this thesis especially focuses on the frame which lose Macroblock( MB ),we proposed a H.264/AVC Multiview extension applied decoder with Error Concealment algorithm for Lost MBs.Due to MVC not only refers frames of same view but also refers frames of different view to compress, and it’s prediction structure call Hierarchical B prediction sturecture,thus it contains I frames、P frames and B frames by reference from intra frame,frames of different views or temporal frame.The thesis proposed MB-based error concealment for P frame and B frame and the experiment result shows that the effect by we proposed is better than other algorithms subjectively and objectively.
The thesis proposed four MB-based error concealment including Pseudo Direct Mode( PDM )、Pseudo Inter 8×8 Mode( PI8×8M )、Earlily Scratch and Predict Intra mode( ESPI ) and Surrounding DVs Detection( SDD ),in the VGA of resolution and XGA of resolution,we simulate four types of lost MBs of situation,they are 3%,5%,10% and 20% BER,the frames through our proposed error concealment compare to its through traditional method.In the 3% BER ,we averagely gained 0.7296 dB in the view 0, 3.2526 dB in the view 1 and 5.3167 dB in the view 2.In the 5% BER ,we averagely gained 0.766 dB in the view 0, 2.8769 dB in the view 1 and 5.0927 dB in the view 2. In the 10% BER,we averagely gained 0.6638 dB in the view 0, 1.8501 dB in the view 1 and 3.3886 dB in the view 2. In the 20% BER ,we averagely gained 0.2071 dB in the view 0, 1.1975 dB in the view 1 and 2.1876 dB in the view 2.Furthermore,the thesis also compares to other algorithm which mainly uses OBMA concept. In the 3% BER,we averagely gained 0.1564 dB in the view 0, 0.4576 dB in the view 1 and 1.0946 dB in the view 2. In the 5% BER,the view 0 of VGA,we averagely gained 0.1657 dB, 0.7767 dB in the view 1 and 1.65 dB in the view 2. In the 10% BER,we averagely gained 0.2324 dB, 0.4371 dB in the view 1 and 1.0509 dB in the view 2. In the 20% BER,,we averagely gained 0.1378 dB in the view 0, 0.2426 dB in the view 1 and 0.6069 dB in the view 2.
|
|---|
