On the resolution of local ambiguity in visual recognition: Modeling by tree-structured Markov random fields

Abstract

SUMMARYTo understand a role of feedback signals in the visual cortex, we have developed a computational model which performs image interpretation in degraded environments; given a degraded picture, we seek the image that maximizes the posterior distribution where maximization is performed via dynamic programming. This is a powerful theoretical framework provided by Geman and Manbeck (1993). We have developed a simplified version of theirs. The capability of the model to perform image interpretation has been investigated. We demonstrate that damages of feedback signals affect noise removal and boundary-finding, indicating a role of the feedback signals.