Corner detection with bilateral filtering of freeman codes

Abstract

Shape contours are one of the most widely used features in the shape detection and recognition applications. A contour can be represented by the chain code which is obtained by coding the relative directions of the elements which form that contour. Sharp transitions in a chain code of a shape contour correspond to the notches and corners that the shape has on the 2D plane. Moreover, very useful information about geometric structures and orientations of the shapes can be obtained by examining the distribution and the moments of the chain code. But chain codes may not be robust to the rotation because of the quantization process of digitizing the curves. Besides, obtaining them from real images in an efficient way is another tough problem. In this work, efficient chain codes are extracted from real images by Edge Drawing method and a corner detection algorithm based on bilateral filtering is proposed.