Corner detection based on Blockwise self-similarity of images

摘要

We propose a novel method for corner detection based on image self-similarity. In our proposed method, first, blockwise self similarity of image is extracted in the same way as for fractal coding of gray-scale images. A corner point is detected by using the property that a fixed point of this self-affine mapping based on blockwise self-similarity is conformed to a corner point of an object contour. In order to demonstrate the efficiency of the proposed method, we applied it to snakes and utilized it to move control points of snakes. It was confirmed that the proposed method detects correct corners from real images. Also, it was confirmed that, using the proposed method, snakes extract object contours which have concavity and convexity and which are difficult to extract by conventional snakes.