Fuzzy Descriptors based on Color, Coarseness, Directionality and Contrast for Image Retrieval

Abstract

In this paper the concept of fuzzy descriptor as a level-two fuzzy set to represent some visual features is proposed. In particular, a fuzzy descriptor based on the dominance of color and texture features is defined and applied to image retrieval. For this purpose, the color and texture are modelled using fuzzy sets, taking into account the imprecision related to these visual features. In addition, the dominance is defined on the basis of a nondecreasing fuzzy quantifier, and the degree of dominance is calculated by means of quantified sentences evaluation. Finally, comparison measures between fuzzy descriptors are presented, and the proposed descriptors and measures are illustrated in image retrieval examples.