Image retrieval method combining global description and local description
Renbo Xia · Jisuanji gongcheng yu sheji · 2012
With respect to two typical problems in CBIR such that Global feature is lack of spatial localization and local description should deal with the problem of image segmentation,an image retrieval method that combines the global color feature and local Gabor wavelet feature is presented.The global feature description,namely MPEG-7 dominant color feature,is extracted from the entire image.The image is partitioned into 5 overlapping blocks,and then Gabor wavelet transforms are carried out on these blocks to extract the Gabor texture feature and color moments as the local feature description.An improved Hausdroff distance metric is proposed to compute the similarity using the local description,which can overcome low Retrial rate caused by image translation and rotation.Finally,the total similarity is obtained by fusing the global similarity and local similarity.The experimental results on the Corel database demonstrate the efficiency of the proposed CBIR method.