Corner detection algorithm based on euclidean distance
Shang Zhen-hong · Jisuanji gongcheng · 2004
Corners are important information carriers in computer vision. A new algorithm was presented here to detect corners on contour in digital image. This algorithm was not going to search another way to approximately calculate the curvature of points on curves,which was defined in continuous domain,but utilized the character of corners in digital nature that the square sum of k Euclidean distance between points pair centered at a corner is locally lowest. Derived from this character,the new algorithm detected corners in a two-pass manner. First pass was to filter the points on a curve that obviously can not be corners by using Freeman chain-code. Second pass was to detect the locations of local minima of the square sum of Euclidean distance. Tests comparing the new algorithm to four famous algorithms were given.