Skip to Main Content
This paper presents a method for extracting texture and color hybrid features and constructing an adaptive weight operator, which can be used for content-based image retrieval (CBIR). This method extracts texture feature effectively based on Brushlet transform, quantifies in the HSV space, and extracts color feature by color histogram. K-mean clustering is introduced to count overall characteristics of images and to classify images. Moreover, in considering the distinction between images, an adaptive feature weight operator is constructed. Experiments show that average precision of our method has increased by 12% compared with hybrid features of traditional weight-based.