Investigation of Resized Shell Chain Code Behavior in Rotated Shapes

Chain codes are major techniques for shape representation. Shell-based chain code, which is one of the related approaches is invariant to rotation. In the shell chain code, the chain code symbols are determined by using distances between the boundary points and the center of gravity of objects. The similarity between the chain codes can be found by means of text comparison metric such as Levenshtein. Nevertheless, the main drawback of the method is that the cost of text similarity computation increases as the length of the chain code increases. On the other hand, the computation cost could be reduced by resizing chain codes. In this paper, the performance of the resized shell-based chain code has been examined. Four rotated shapes were used for testing the proposed method and their chain codes were compared with Levenshtein similarity metric. The obtained results confirm that there is no significant decrease in the performance of similarity computation with the resized shell chain codes.