A Contour-Based Method on Image Retrieval

Abstract

Compared to color and texture, contour is closer to human vision. So contour-based image retrieval has been a hot topic. However, existing methods can't obtain object contour effectively, which makes contour description and matching more difficult. To address these problems, this paper uses KPAC (Kernel-based Parametric Active Contour) to obtain the object contour, and proposes a novel method called LICS (Local Important Contour Sequence) for feature description and matching. Experimental results demonstrate that this approach achieves state-of-the-art performance in image retrieval.