A Novel Approach to Shape Based Image Retrieval Integrating Adapted Fourier Descriptors and Freeman Code

Abstract

Summary Accurate recognition of a shape depends on its representation. Therefore an efficient representation of shape information is the basic task of computer vision. We present a novel method of representing a shape. It is an adaptation of Fourier theory based descriptors integrated with Freeman code. To obtain a good boundary description, we choose to define it in an 8-connectivity. Fourier Descriptors have a disadvantage of giving equal weights to all its 8-neighbors. So, the Fourier coefficient formula must be completely recomputed so that horizontal and vertical neighbors are given more weight when compared to the diagonal neighbors. Diagonal neighbors are identified using Freeman code. Under Java retrieval framework, we compare our proposed results with the existing Fourier descriptors and validate the proposed method.