Abstract: Aiming at the problem for losing boundary information within the region near the corner point during boundary tracking,the method for judging and searching corner point was studied. On the basis of analyzing the relationship between the radius of extended neighborhood and the location of the corner, the corner point judging rule was proposed and the range in which corner point existed was deduced. The region searching algorithm was proposed to detect the corner point in the deduced range. From the relationship of 5 known boundary points, the algorithm estimated the neighborhood in which the corner point was included, and made the neighborhood contracted to the corner point by selecting appropriate radius. Examinations were performed at test images and side face image. Results show that the boundary points increase form 13 to 21 for test image and from 19 to 37 for side face image, the new method can detect the corner point accurately and can also detect right seeds so that the boundary tracking algorithm can include all contour information.