Optimized Method of Multi-Feature for Content-based Image Retrieval
Zhengyan Dai, Su‐Juan Qin · 2015
Content-based image retrieval has been focused on attention during recent years.Traditional methods of CBIR most relied on single feature extraction of images, which could only reflect single characters of the image.We proposed an optimized structure of multi-feature for content-based image retrieval.We extracted the color feature using HSV bin to reflect the holistic feature of the image, and then extract the sift descriptor to produce Bag-of-Words bin to reflect the local feature of the image.Using those two features to get fusion vectors represents the image synthetically.Finally, we use hierarchy clustering to cluster images in the database to get efficient retrieval results.In the experiment, we test the precision of our method comparing with the same type of feature extraction combining method to validate our promotion.