Fast image retrieval using error diffusion block truncation coding and unsupervised clustering

Content Based Image Retrieval (CBIR) employing compression technique and unsupervised clustering focus on the faster retrieval of desired images with high amount of accuracy. In this paper color images are indexed using the features extracted from Error Diffusion Block Truncation Coding (EDBTC). A new framework of CBIR with unsupervised clustering is used here in which the amount of time required for comparing the target and query image is significantly reduced. Experimental result shows that the proposed method not only achieves a good quality of image compression but achieves a significant reduction in the amount of time required for image retrieval. The proposed method was able to obtain an average of 4 to 5 sec difference in performance time as compared to the old retrieval method without clustering. The speed of retrieval varies linearly with the number of images in the database i.e. an increase in database images showed significant increase in the system performance.