"A Novel Approach to Unequal Error Protection for Image/Video Delivery with Source Detector"

Wenjun Zeng, Yap-Peng Tan, Bede Liu and Lih-Huah Yiin

presented at SPIE Voice, Video and Data Communications, Photonics East'96, Boston, Nov. 1996.

ABSTRACT

One way to efficiently combat channel errors is to employ unequal error protection (UEP) for information of different importance. The traditional approach to UEP through channel coding does not fully take advantages of the image signal properties. The high correlation nature of the image signal makes it possible to detect and correct many channel errors directly on the decoded image. Based on this observation, we propose in this paper a new approach which provides UEP through channel coding in a way seemingly contrast to traditional ones. Specifically, some good source detection/correction schemes are incorporated to detect/correct, directly on the decoded image, the noticeable damage which is often caused by channel errors affecting the $important$ (in the view of traditional UEP schemes) $data$. With the help of source detectors, the problem of how to prioritize the transmitted data for channel protection is revisited and reformulated. We show that by intelligently combining unequal channel protection and source detection, both subjectively and objectively better quality of reconstructed images can be obtained. A general framework of the new approach will be presented. Three case studies, including one for transmission of vector quantized images over noisy channels, one for a simplified fax coding system and the other one suggested by an intelligent block dropping approach, are examined.

Keywords

unequal error protection, joint source-channel coding, data prioritization, error resilience, source detector


Paper

·  Postscript version of the paper (1.4 M)


Last update: Oct. 2, 1996.

wzeng@ee.princeton.edu