Video Error Concealment Using Fidelity Tracking

Akio YONEYAMA  Yasuhiro TAKISHIMA  Yasuyuki NAKAJIMA  Yoshinori HATORI  

IEICE TRANSACTIONS on Information and Systems   Vol.E91-D   No.1   pp.70-77
Publication Date: 2008/01/01
Online ISSN: 1745-1361
DOI: 10.1093/ietisy/e91-d.1.70
Print ISSN: 0916-8532
Type of Manuscript: PAPER
Category: Image Processing and Video Processing
error concealment,  MPEG,  H.264/AVC,  error resilience,  

Full Text: PDF(1.2MB)>>
Buy this Article

We propose a method to prevent the degradation of decoded MPEG pictures caused by video transmission over error-prone networks. In this paper, we focus on the error concealment that is processed at the decoder without using any backchannels. Though there have been various approaches to this problem, they generally focus on minimizing the degradation measured frame by frame. Although this frame-level approach is effective in evaluating individual frame quality, in the sense of human perception, the most noticeable feature is the spatio-temporal discontinuity of the image feature in the decoded video image. We propose a novel error concealment algorithm comprising the combination of i) A spatio-temporal error recovery function with low processing cost, ii) A MB-based image fidelity tracking scheme, and iii) An adaptive post-filter using the fidelity information. It is demonstrated by experimental results that the proposed algorithm can significantly reduce the subjective degradation of corrupted MPEG video quality with about 30 % of additional decoding processing power.