An active edge detection approach

In this paper, an active edge detection approach is proposed. It at first locates the discontinuity neighborhoods in the image with a simple error analysis algorithm. To suppress the noise and interpolate the missing points included in the edge neighborhoods, an improved relaxation labeling algorithm is then applied. The sizes of the spatial extent of the discontinuity neighborhoods can be estimated from the resultant images. Based on these measures, the optimal difference operators are than applied to the corresponding neighborhoods in the original image to extract the edge points exactly. As the new method can concentrate the major calculation on where the edges may be and extract the edgepoints adaptively, it works with high precision and low computational cost. Experimental results show the good quality and high speed of the proposed method.<>