Abstract: Aiming at the limitation of traditional boundary tracking algorithms which are mainly used to extract annular closed boundary, an improved boundary tracking algorithm based on eight neighborhood targets is proposed, which focuses on the internal and external boundary calibration and "solated island" processing in image tracking. The proposed algorithm and the Moore neighborhood boundary tracking algorithm are used to test the four images in the MPEG7-CE image library respectively. The results show that the improved algorithm is correct. Finally, the algorithm is applied to the sea ice extraction of the Arctic snow image. Experimental results show that the improved algorithm can effectively extract large range of image boundaries with complex topological relationships.

Key words: binary image, eight neighborhood, boundary tracking, Termination criteria, topology relations