Inexhaustive region segmentation by robust clustering

An inexhaustive region segmentation method using a novel robust clustering algorithm is proposed in the present paper. The term `inexhaustive´ means that this method segments only homogeneous and major regions in the image. Therefore, the pure features of the major regions that are the important clues in a recognition process can be obtained. The finite mixture model is used to represent the distribution of the features. The region segmentation is formulated as parameter estimation of the model. The robust clustering algorithm is used in the estimation. The number of major regions is estimated from changes of the number of outliers as a function of the number of components. Experimental results for the real images are shown