Extracting Regions of Interest from Still Images: Color Saliency and Wavelet-Based Approaches

In this paper we explore four distinct approaches to extracting regions of interest (ROI) from still images. We show the results obtained for each of the proposed approaches, and we demonstrate where each method outperforms the other. The four approaches are: 1) a block-based discrete wavelet transform (DWT) algorithm, 2) a color saliency approach, 3) a wavelet coefficients variance saliency approach, and 4) an approach based on mean-shift clustering of image pixels. The wavelet-based approaches are shown to perform well on natural scene images that usually contain regions of distinct textures. The color saliency approach performs well on images containing objects of high saturation and brightness, and the mean-shift clustering approach partitions the image into regions according to the density distribution of pixel intensities.