用兴趣点凸包及SVM加权反馈实现图像检索.pdf

32 11 Vol. 32 No. 11 2009 11 CHINESE JOURNA L OF COMPU TERS Nov. 2009 SVM 苏小红 丁 进 马培军 ( 150001) , , , , , . , Gabor , . , , . , 20% , 10% . ; ; ; ; ; TP391 DOI : 10. 3724/ SP. J. 1016. 2009.02221 Image Retrieval by Convex Hulls of Interest Points and SVMBased Weighted eedback SU XiaoHong DIN G Jin MA PeiJun ( S ch ool of Comp ute r Sc ienc e and Te chnology , H ar bin Inst itute of Te chnology , H ar bin 150001) Abstract T o solve the problem of image retrieval method based on annular color histogram, a new image characteristics extraction method based on convex hulls of interest points is presented. Firstly , the interest points on an image are detected by w avelet transform. T hen, convex hulls of interest points are calculated recursively and these points are assigned to some buckets by a spe cific algorithm to form a color histogram for every bucket. The similarity of tw o images is calcu lated by the similarity betw een histograms of tw o buckets. Combined w ith spatial distribution feature and Gabor texture feature based on convex hulls of interest points, the system of image retrieval is built. Experiments on image database show that this method w orks w ell w hen isolated points exist in the interest points set and so provide more accurate retrieval performance compa ring w ith other retrieval method based on interest points. Further more, a novel relevance feed back method is presented. It improves the query point movement relevance feedback method by setting w eights based on support vector machine cluster results. T he experi