Symbolic fusion of luminance-hue-chroma features for region segmentation

摘要

Segmentation is a very important pre-processing step towards image analysis or image compression. In this paper, we propose an original region segmentation method applied to color images. The two key elements of this method are: the use of a color space where hue is explicitly defined and processed according to its relevance which can be linked to chroma, the use of symbolic representations and rule-base systems combining color and luminance features in order to define homogeneity between pixels. As an illustration, this approach is applied to an iterative region growing method classically used in grey level image processing. Experimental results on real images are given.