This paper introduces a novel framework for large-scale image retrieval. This work is distinguished by three major contributions. The first is a new global image descriptor called Block HSV Histogram (BHSVH), which captures both the color statistics and color space distribution information. The second is that we propose an efficient filtering process that uses a hashing algorithm to index the original features. Using a hashing algorithm is indeed a promising method for large-scale retrieval. The third is that we propose a new distance measure for re-ranking retrieved images. We conducted extensive experiments to show that the proposed image retrieval framework can scale to very large-scale image database.