A Modified Feature Points Extraction Algorithm and it's Adaptability Evaluation

Extracting integral feature points is important to object recognition and 3D reconstruction in image processing field. This paper modifies a kind of feature extracting algorithm based on image edge. First, image edge contour is extracted. Second, geometric gravity center of edge outline is calculated. Third, argument-polar radius curve of polar coordinate system whose pole is geometric gravity is formed, and maximum and minimum value points are searched, they are feature points. Comparing the original algorithm, the modified algorithm can not only extract protruding points, but also concave points. In order to evaluate the quality of the algorithm, the paper appraises the adaptability of the algorithm in the case of scale transformation, rotation transformation and affine transformation. The result indicates that the algorithm is simple and its adaptability can reach 79.1%. In actual application, the algorithm has important value to auto-input of two-dimension curve based on feature points and three-dimension reconstruction.