A Novel Vector-Based Approach to Color Image Retrieval Using a Vector Angular-Based Distance Measure

Abstract

Color is the characteristic which is most used for image indexing and retrieval. Due to its simplicity, the color histogram remains the most commonly used method for this task. However, the lack of good perceptual histogram similarity measures, the global color content of histograms, and the erroneous retrieval results due to gamma nonlinearity, call for improved methods. We present a new scheme which implements a recursive HSV-space segmentation technique to identify perceptually prominent color areas. The average color vector of these extracted areas are then used to build the image indices, requiring very little storage. Our retrieval is performed by implementing a combination distance measure, based on the vector angle between two vectors. Our system provides accurate retrieval results and high retrieval rate. It allows for queries based on single or multiple colors and, in addition, it allows for certain colors to be excluded in the query. This flexibility is due to our distance measure and the multidimensional query space in which the retrieval ranking of the database images is determined. Furthermore, our scheme proves to be very resistant to gamma nonlinearity providing robust retrieval results for a wide range of gamma nonlinearity values, which proves to be of great importance since, in general, the image acquisition source is unknown.