Color Image Segmentation by Compacigram Analysis

This work lies within the scope of color image segmentation by pixel classification. Classes of pixels are generally constructed by mode detection applied to the color histogram, but this method tends to fail when the color distributions of the different objects highly overlap in the color space. A new characteristic function of the distribution of the colors within the image is proposed here. The compacigram takes into account both the distribution of colors in the color space and the spatial location of colors in the image plane. Experimental tests on two reference images show that the mode detection technique based on a convexity analysis of the compacigram leads to promising results for segmentation purposes.