Natural Scene Segmentation Based on Information Fusion and Homogeneity Property

This paper presents a novel approach to natural scene segmentation. It uses both color and texture features in cooperation to provide comprehensive knowledge about every pixel in the image. A novel scheme for the collection of training samples, based on homogeneity, is proposed. Natural scene segmentation is carried out using a two-stage hierarchical self-organizing map (HSOM). The proposed method confirms that the sample selection based on homogeneity and the selflearning ability and adaptability of the HSOM, coupled with the information fusion mechanism, can lead to good segmentation result, which is validated by experiments on a variety of natural scene images.