Recognition of Shapes for Object Retrieval in Image Databases by Discrete Curve Evolution

This paper presents the recognition of shapes for object retrieval in image databases using skeleton-based and contour-based representation by discrete curve evaluation and two consecutive primitive edges. Humans tend to use high-level concepts in everyday life. Object segmentation and recognition is the primary step of computer vision to achieve image retrieval of high-level image analysis. Contour-based and skeleton-based representations are important for object recognition in different areas. In comparing the contour-based approaches with the skeleton-based approaches for object representation, the contour-based is more sensitive to noise than a skeleton-based approach based on a good skeleton pruning method, but a rough shape classification can be performed since the obtained skeletons do not represent any shape details. In this paper, we proposed a novel method to integrate the contour-based approaches with skeleton-based approaches for object representation. The contour-based and skeleton-based representations are based on the proposal of two consecutive primitive edges method and discrete curve evolution method respectively. Experimental results demonstrate that the performance of the proposed algorithm is superior to Torsello and Hancock’s method in terms of retrieval accuracy.