Randomize color signature employing locality-sensitive hashing (LSH)

Visual signatures construction of images is an essential step for the CBIR system. When the database size become larger. Most existing algorithms (e.g. k-means, Kd-tree, Mean-shift) to build signature become unfavorable due to the prohibitive time and space requirements. In this paper we propose the randomize color signature based on the LSH technique. The proposed descriptor benefited of the effectiveness of LSH in terms of time and accuracy of clustering. Experiments show the effectiveness of our approach.