An Object Boundary Following Algorithm

Pattern recognition is a popular area in digital image processing. The computer can classify and recognize objects of different pattern through comparing the color composition, texture features or boundary contour。 When the computer identifies the different mode based on the boundary information, it will reduce the complexity of the recognition largely. Due to the boundary’s important role in pattern recognition, this article will discuss the boundary following algorithm which used to extract the object’s contour information.This thesis carries on the detailed analysis and research to some existing boundary following algorithm, and on this basis, put forward a new algorithm tracking in higher order neighbor to extract the boundary of the object. This algorithm can be applied to the actual requirements to extract different fine degree of object contour through changing the neighbor’s order to track. Relative to the existing boundary following algorithm, the improvement of this algorithm is that the algorithm expands eight neighbor into the larger outside area in the track of the border search. When the boundary of the target graphics to be processed has a breakpoint, the search process can skip the breakpoint, and get more correct contour than the boundary following algorithm based on the eight neighbor. In this paper, the algorithm’s correctness and effectiveness are verified in the actual test by using it to deal with some template images in MPEG7-CE_shape-1_part_B standard library. During verification, the results’ Deviation of the algorithm are studied with respect to the classical Moore neighbor tracing algorithm. In addition, the paper analyze the effect of the algorithm in extracting different graphics templates. At last, this paper use the result of the algorithm in simple contour matching, and the availability of the boundary results is verified.