Linguistic description of images based on fuzzy histograms

Abstract

The paper presents a new approach for generating linguistic description of color digital images based on fuzzy histograms. The CIE chromaticity diagram is classified into fuzzy color granules and employed in order to recognize color clusters in an image or collection of images. A fuzzy inference process uses the histograms and fuzzy IF-THEN rules concerning color, location, size, and shape of the color clusters. The fuzzy histograms illustrate participation rate of pixels in the fuzzy color granules, and also within the so-called macropixels of different sizes of the same color. The histograms are derived from the matrix that expresses membership grades of pixels to the fuzzy color granules of the CIE diagram. The linguistic description is applied in image recognition and image retrieval tasks.