Spatial-region classification by Min-Connected algorithm for unsupervised segmentation

Abstract

This work lies within the scope of color image segmentation by spatial-region classification. The spatial distribution of objects in each region of image is difficult to be identified when the region are non-connected clusters. A standard of color identification by conventional methods of segmentation remains weak for capturing the spatial dispersion of the various objects of the same color region. We propose to apply a spatial classification to characterize geographical connected sets that represent the same regions. The originality of this paper lies in our new min-connected algorithm which is derived from a spatial-color compactness model. Our work is a hybrid segmentation that takes into account the distribution of colors in the color space and the spatial location of colors in the image plane. Experimental tests on synthetic and real images show that our technique leads to promising results for segmentation.