Shape representation and matching is a not only important but also challenging part in image retrieval. In our previous work, a shape-based image retrieval system was proposed. The Mountain Climbing Sequence (MCS) is used to represent the shape feature of an object, which is invariant to some transformations such as translation, scaling, rotation, and even reversion, and also has a high tolerance for occlusion and noise. In this paper we improve this system, in terms of both retrieval precision and time efficiency, by adopting the Longest Common Subsequence (LCS) matching mechanism to match two MCSs more effectively and utilizing the Selection algorithm to speed up the retrieval process. Experimental results show that the improved system has far better performance. Especially it copes with the occlusion problem very well; that is, it has higher tolerance for occlusion.
淡江理工學刊 = Tamkang journal of science and engineering 10(3), pp.265-274