An Improved Block-Matching Method for Temporal Error Concealment in H.264

碩士 === 國立東華大學 === 電機工程學系 === 95 === For the development of digital communication technology, data in transmission of compressed video over an error prone network, such as the Internet or wireless network, may lead to information loss due to the congestion or routing disruption. In this case, the los...

Full description

Bibliographic Details
Main Authors: Chin-Jung Chen, 陳芝蓉
Other Authors: Chun-Lung Hsu
Format: Others
Language:zh-TW
Published: 2007
Online Access:http://ndltd.ncl.edu.tw/handle/sgtrn5
Description
Summary:碩士 === 國立東華大學 === 電機工程學系 === 95 === For the development of digital communication technology, data in transmission of compressed video over an error prone network, such as the Internet or wireless network, may lead to information loss due to the congestion or routing disruption. In this case, the loss frame is preferred to be recovered for meaningful application by the mechanism of error concealment in decoder. H.264/AVC is a newest video coding standard which provides approximately 50% bit rate saving for equivalent perceptual quality relative to the performance of prior standards. In this thesis, we proposed a marcoblock (MB) based error recovering mechanism called surrounding block-matching algorithm to deal with error concealment problem in H.264/AVC. The main difference between our proposal and the widely used block-matching or boundary-matching methods is that both block-matching and boundary-matching methods take the boundary pixels of neighboring MBs to compute matching costs. Nevertheless, in our proposed algorithm, several candidate motion vectors are firstly discovered to form a motion vector (MV) set from the neighboring MBs of miss MB. Afterwards, the miss MB is compensated by the pixels of candidate MBs associated with MVs selected from previous step, and the proposed surrounding block-matching method is thus applied to find the best result from the candidate MV set. Through the proposed algorithm, the matching accuracy can be improved since our proposal uses entire pixels instead of boundary pixels from the upper/bottom MBs to indicate the best concealment result. Simulation results show that our proposed algorithm outperforms several previous works in different bit error rate while running on JM10.2 reference software.