From pixels to information granules:
Granular Computing and Logic Processing
Witold Pedrycz
Dept. of Electrical & Computer Engineering
University of Alberta, Edmonton, Canada
&
Systems Research Institute
Abstract:
Human endeavors in image interpretation and understanding ultimately involve information granules. Content-based image retrieval (CBIR) and effective mechanisms of relevance feedback use concepts of information granules.
In this talk, we present the fundamentals of Granular Computing and demonstrate their relationships to image understanding and interpretation. The calculus of information granules exploits logic constructs and its processing flexibility is realized in terms of fuzzy neurocomputing. We show how the constructs of fuzzy logic neurons and fuzzy neural networks augment the logic processing through learning abilities. Subsequently, the use of the logic networks is illustrated to the problem of content-based image retrieval.