Object Contour Matching Using the Biorthogonal Wavelet Transform

Object contour matching is essential in content-based indexing and retrieval of digital images. In this paper, an automated biorthogonal-wavelet-transform-based object contour matching method is proposed. First, a wavelet transform modulus maxima (WTMM) image is produced by a one level biorthogonal wavelet transform. This WTMM image contains curvature points of the contour. Then the high curvature points (HCPs) are kept in the WTMM image whereas the less important points are discarded. These HCPs are further utilized to locate the centroid of the object contour. Finally, a feature vector for matching object contour is constructed by the sorted distances between each HCP and the centroid. This feature vector is invariant to the translation, rotation, and scale changes. The experimental results demonstrate that the proposed approach can accurately and efficiently match the object contour under these three changes.