Edge Detection for Color Image Based on Feature Points

Traditional edge detection techniques consist of following steps: searching for the points whose gradient value is larger than the intended threshold by global image scanning, removing pseudo edge points and connecting disconnected edge. To get a more excellent performance of edge detection and expand the adaptability of the basic technology, a novel edged detection algorithm is proposed in this paper, which combines feature point detection and edge growing together to improve the detection of edge in image. In this method, edges itself are treated as a special kind of region, and classical region growing method is employed to generate edges. First, feature point detection is adopted to find out points with outstanding characteristics; these points are then used as seeds in order to trace edges in the similar way as region growing. During region growing, the response of edge gradient and the similarity among regions are used to check growing direction. Simulation results show this method can reduce the number of pixels involved in comparison, remove isolated edges and ensure the detected edge is continuous and with a single pixel width.