Complementing a region descriptor for shape retrieval

In this paper, we propose to complement a region descriptor for shape retrieval with a contour descriptor; the descriptor thus obtained is a composite descriptor. We use a combination of a region descriptor and contour descriptor to obtain the proposed descriptor. A contour extraction technique is also proposed; it is required for extracting the contour of the image prior to applying a contour descriptor. Normalization of descriptors is required before disparate descriptors can be combined into a single descriptor. Hence, we test the performance of two common methods for normalization of descriptors and adopt the better performing method. Experiments are performed to test the effectiveness of the proposed descriptor for retrieval of 2d images. Sets of composite descriptors, obtained by assigning different weights to the region component and the contour component, are also evaluated. Item S8 within the MPEG-7 Still Images Content Set is used for performing experiments; this dataset consists of 3621 still images. Experimental results show that the proposed descriptor is effective.<br />