A new corner detection algorithm for chain code representation

Abstract

A new algorithm is proposed to detect a corner of a thinned binary image that is represented by an eight-connected contour chain code. The algorithm is based on chain-coded image, deriving the slope between each code, analyze the series of chain code, and finally decide the existence of corner at the current pixel location. The work assumes that the pre-processing processes on the image, namely thinning and digitization, have been done. Two weighted parameters identified as significant factors in determining the accuracy of the corner detection algorithm are discussed. The parameters are the length of segment and threshold value. Computational phases to derive values of rows and columns given a series of chain code are also given in detail. The algorithm can be used to interpret line drawing that represents three-dimensional object.