Corner detection by means of adaptively estimated curvature function

A corner detection algorithm is proposed for planar shapes based on a new curvature function. The main novelty of the function is that its calculation parameters are not constant, but adapt to the curvature at each point of the contour. The process consists in estimating the maximum length of the contour, yielding no significant discontinuities at each of its points in order to calculate the incremental adaptive chain code of the contour at those points. This improved chain code provides a precise curvature index that enables a better characterisation of the contour to be made. Thus, all corners can be efficiently detected even if working with poor signal to noise ratios