Obtaining the relevant colors of an image through stability-based fuzzy color histograms

Abstract

In a previous work, we introduced a fuzzy-based color image segmentation method for detecting the "relevant colors" of an image, founded on color histogram analysis. The present paper outlines the whole process and introduces the construction of a new fuzzy paradigm-based color histogram, which uses color components' stability degrees to modify a set of histogram positions according to the uncertainty of the pixel color components. The processing of such histogram provides a group of fuzzy sets that fit the number and domain of image's relevant colors better than the processing of the traditional color histogram. Moreover, the image segmentation results are more similar to the human visual perception of the scene.