The colour and texture - a novel image retrieval technology based on human vision

Abstract

If we want the computer to behave like humans, especially when they are recognising an image, we must study the human vision system carefully. Obviously, the colour and the texture are the most important perceptual characteristics of humans. When someone retrieval an image, he or she can use a variant of colour spaces, which have different advantages. In this paper, we replace the original colour space with HSV in MPEG-7 colour layout descriptor according to the nature of the human vision system. Experiment shows that the modified method increased retrieval efficiency greatly. Further more, we establish a multi-feature space, both homogeneous texture descriptor and colour layout descriptor are used. Since there are differences in human perceptions of colour and texture, in order to successfully retrieval an image which caters to the users, parallel genetic algorithm PGA is employed to adjust the weight of each feature space. The experimental evidence shows that the method made the computer behave like human.