Co-occurrence of Maximum and Minimum Local Sign and Center-symmetric Edges for image retrieval

Content-based image retrieval has become an important task these days because of existence of large image databases. An efficient feature extraction method is prominent for image indexing and retrieval. Texture is an informative feature of image and many texture feature descriptors have been proposed for image retrieval. In this paper, a new texture feature descriptor based on maximum and minimum local edges is proposed. Local edges are obtained by comparing center pixel with its surrounding pixels and in the proposed feature descriptor, both sign and center-symmetric edges are used. Finally rather than using a histogram for calculation of frequency of each pattern, co-occurrence matrix is used. For evaluation of proposed work, ORL face and Brodatz texture databases are adopted and experiments show effectiveness of the proposed feature descriptor over other well-known methods like LBP, LTP, DLEP etc.

Keywords:

Content-based image retrieval , feature extraction , texture , local edge , sign edges , center-symmetric edges , gray level co-occurrence matrix , ORL face database , Brodatz database