Image Retrieval Based on Syncretizing of Object Colors and Edge Features
YI Jia-fu · Journal of Sichuan University · 2007
Due to the complexity of image features,computation of characterizing,extracting,indexing,matching for image retrieval is very vast,vector of features is hundreds upon thousands of dimensions at every turn.Applica- tion of CBIR for Intemet faces the difficulty of so-calleddimension calamity.The paper provides a new image re- trieval model which integrates the spacial distributing information of colors and the shape features.The color fea- tures are represented by an amelioration of HSV color model:color sector.To describe the relative position and ori- entation of objects within the image,hue edges in the image,computed using Canny approach,are processed by Hough transform to generate the spatial histogram which is used as shape representation.To address matching of im- age,they define two global distance measures based on the accumulative distance and vector-based distance to com- pare the similarity of two images.Experimental results indicate that the approach provides higher retrieval rate than that of single feature(color),is a better tradeoff between high efficiency and reasonable precision.