New similarity measure for illumination invariant content-based image retrieval

Similarity measure is used to study the similarity between patterns and forms the basis of content-based image retrieval systems. We have investigated existing similarity measures, and proposed a new similarity measure for illumination invariant content-based image retrieval that does not consider any prior knowledge about the camera or the illuminant. Normalized cumulative colour histogram is adopted in this paper for image feature modeling, while the new similarity measure compares the query and target images to search among large databases. Our algorithm is tested on the SFU database, and the experimental results prove the efficiency of the proposed technique during successful image retrieval.