An interactive constraint-based system for selective attention in visual search

Abstract: In this paper, we face the problem of model-based object recognition in a scene. Computer vision techniques usually separate the extraction of visual information from the scene from the reasoning on the symbolic data. We propose to interactively intertwine the two parts: the reasoning task on visual information is based on constraint satisfaction techniques. Objects are modeled by means of constraints and constraint propagation recognizes an object in the scene. To this purpose, we extend the classical Constraint Satisfaction Problem (CSP) approach which is not suitable for coping with undefined information. We thus propose an Interactive CSP model for reasoning on partially defined data, generating new constraints which can be used to guide the search and to incrementally process newly acquired knowledge.

 not available