Share Email Print

Proceedings Paper

An edge-based temporal error concealment for MPEG-coded video
Author(s): Yu-Len Huang; Hsiu-Yi Lien
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

When transmitted over unreliable channels, the compressed video can suffer severe degradation. Some strategies were employed to make an acceptable quality of the decoded image sequence. Error concealment (EC) technique is one of effective approaches to diminish the quality degradation. A number of EC algorithms have been developed to combat the transmission errors for MPEG-coded video. These methods always work well to reconstruct the smooth or regular damaged macroblocks. However, for damaged macroblocks were irregular or high-detail, the reconstruction may follow noticeable blurring consequence or not match well with the surrounding macroblocks. This paper proposes an edgebased temporal EC model to conceal the errors. In the proposed method, both the spatial and the temporal contextual features in compressed video are measured by using an edge detector, i.e. Sobel operator. The edge information surrounding a damaged macroblock is utilized to estimate the lost motion vectors based on the boundary matching technique. Next, the estimated motion vectors are used to reconstruct the damaged macroblock by exploiting the information in reference frames. In comparison with traditional EC algorithms, the proposed method provides a significant improvement on both objective peak signal-to-noise ratio (PSNR) measurement and subjective visual quality of MPEG-coded video.

Paper Details

Date Published: 24 June 2005
PDF: 9 pages
Proc. SPIE 5960, Visual Communications and Image Processing 2005, 59602X (24 June 2005); doi: 10.1117/12.632565
Show Author Affiliations
Yu-Len Huang, Tunghai Univ. (Taiwan)
Hsiu-Yi Lien, Tunghai Univ. (Taiwan)

Published in SPIE Proceedings Vol. 5960:
Visual Communications and Image Processing 2005
Shipeng Li; Fernando Pereira; Heung-Yeung Shum; Andrew G. Tescher, Editor(s)

© SPIE. Terms of Use
Back to Top