A corner detection-based object representation technique for 2-D images

A corner detection-based object representation technique that operates on a thinned binary image is presented. The proposed method consists of two computational phases. The first phase involves the use of a corner detection method which detects the corners in the image. The second phase is a curve-following process that establishes the topological relationships between the corners for a complete object representation. The final output of the method is a list of corners, each identified by its position, the number of its edges, and the neighboring corners connected to it. Experimental results on a computer-generated line-drawing and an image of a real object are presented. The method gives reasonably good results for the images used in the test experiments.<>