Self-similarity based interest points extraction for weak texture

In this paper, a novel point detection method for weak texture is presented. The points with weak texture characteristic are detected by the self- similarity concept. The detection mechanism can be separated into two steps: 1) Detecting the candidate feature points by local self-similarity; 2) Using texture strength conception to refine the candidate point set of weak texture. Experiments demonstrate the effectiveness of the proposed point detection algorithm for weak texture.