A Efficient Content based Image Retrieval System using GMM and Relevance Feedback

Content-Based Image Retrieval (CBIR) systems are required to effectively extract information from ubiquitous image collections. Retrieving images from a large and highly varied image data set based on their visual contents is extremely challenging. CBIR has been studied for decades and many good approaches have been proposed. But they do have some drawbacks. Texture and color are the significant features of CBIR systems. This paper gives a novel method of CBIR, in which images can be retrieved using color-based, texture-based and color and texture-based. Auto color correlogram and correlation for extracting color based images, Gaussian mixture models for extracting texture based images are the algorithms used here. For Relevance Feedback, Query Point Movement technique is used. Thus the proposed method achieves better performance and accuracy in retrieved images along with iteration reduction.