A Color-Spatial Feature for Content Based Image Retrieval
Yanting Hu · Bulletin of Science and Technology · 2008
In order to solve the problem that color histogram can not describe the color spatial information of an image, a convex-hull intra-angle histogram feature based on color quantization was proposed. Firstly, the image was quantized in HSV color space. Secondly, for a given color quantization value in HSV color space, the convex-hull that includes all pixels whose color quantization values are equal to the given value was computed. Then a histogram was computed based on the all intra-angle (s)of the convex-hull. Finally, the convex-hull intra-angle histogram of the whole image was built by processing all color quantization values. Comparision study was carried out between other similar approaches proposed in literatures and ours. Experimental results showed that our approach can effectively describe the color spatial information. Furthermore, better retrieval accuracy can be achieved by using our approach.