A New and Efficient Descriptor for Region Based Shape Retrieval

In this paper, a new and efficient descriptor for region-based shape representation and retrieval is presented. The proposed descriptor called sub-area shape matrix which is obtained based on the circular raster sampling technique and the axis of least inertia. An element of the sub-area shape matrix is obtained as the sum of pixel values on the region of the shape image enclosed by two neighboring circles and two successive diameters. Similarity distance between two shapes is captured as the chi² distance between their shape matrixes. The proposed method of shape representation and retrieval is shown to be invariant to image translation, rotation and scaling and robust to minor deformations on the MPEG-7 core experiment shape database.