An improved clustering based on edge detection method

摘要

Edge is one basic feature of image, and extracting edges could give a strong support to target recognition. This paper proposes an improved edge information clustering algorithm based on evidence theory. Through merging the gradient matrix of image based on confidence, a group of eigenvector are obtained. Further, an improved K-means algorithm is applied to cluster, to eliminate the effects of different clustering results caused by different initial centers. The algorithm enhances the efficiency of edge detection, and the experiment is provided to demonstrate the results.