A fast image retrieval system based on color-space and color-texture features

Abstract

This paper presents two image features, multi-orientation color complexity (MOCC) and color-space relation (CSR). MOCC refers to the color complexity. CSR concerns the spatial relations of similar color pixels in an image. By combining both features, an image retrieval system was developed. The experimental results revealed that such a system can perform expressively at accurate recognition rate. To further speed up this system, a clustering based filer was applied to quickly sieve out the most unqualified database images.