A novel regions-of-interest based image retrieval using multiple features

In this paper, a new ROI based color image retrieval is proposed. Firstly, the ROI are extracted in DWT domain by using the human visual characteristic and K-mean clustering. Secondly, the ROI is segmented into subregions, and the ROI features are extracted from the subregions (the dominant color and its percentage is used for color feature, the pixel distribution in the subregion is used for shape feature). Finally, the average similarity between images is computed according to the ROI above features. Experimental results show that our image retrieval is more accurate and efficient in retrieving the user-interested images when there are ROI in the image (especially for the image with simple background)