An efficient content based image retrieval using edge orientation co-occurrence matrix

In this paper, we propose a novel and effective technique namely edge orientation co-occurrence matrix for content based image retrieval. Edge orientation co-occurrence matrix (EOCM) represents the spatial correlation of texture orientation and is obtained from gradient of gray leveled image. After smoothing the image, gradient of the image is calculated and then the whole image is divided into 3by3 non-overlapping blocks. Every block is characterized with a number between 1 to 9 based on its gradient orientation, and then probability of the occurrence of each block with its neighboring ones is computed. This probability takes into account as the feature of the image. Experiments on WANG database demonstrate that the use of EOCM significantly improves the retrieval performance.