An image recognition algorithm using relationships between segmented coarse regions

In this paper, we propose an image recognition algorithm using the relationships between regions obtained by coarse region segmentation processing. The segmentation is achieved by the resistive-fuse network (RFN) model. The proposed algorithm analyzes the relationships (distance and angle) between the regions using the width, height and average color intensity of the regions. Finally, the algorithm outputs the positions and attributes of the regions. Experimental results with persons and cars confirm the validity of the proposed algorithm.