Communication theoretic image restoration for binary-valued imagery |
| |
Authors: | Neifeld M A Xuan R Marcellin M W |
| |
Affiliation: | Department of Electrical and Computer Engineering, Optical Sciences Center, University of Arizona Tucson, Arizona 85721, USA. neifeld@ece.arizona.edu |
| |
Abstract: | We present a new image-restoration algorithm for binary-valued imagery. A trellis-based search method is described that exploits the finite alphabet of the target imagery. This algorithm seeks the maximum-likelihood solution to the image-restoration problem and is motivated by the Viterbi algorithm for traditional binary data detection in the presence of intersymbol interference and noise. We describe a blockwise method to restore two-dimensional imagery on a row-by-row basis and in which a priori knowledge of image pixel correlation structure can be included through a modification to the trellis transition probabilities. The performance of the new Viterbi-based algorithm is shown to be superior to Wiener filtering in terms of both bit error rate and visual quality. Algorithmic choices related to trellis state configuration, complexity reduction, and transition probability selection are investigated, and various trade-offs are discussed. |
| |
Keywords: | |
本文献已被 PubMed 等数据库收录! |
|