Comprehensive Multi-feature Image Retrieval Research

With the wide application of image retrieval technology in many fields.How to achieve accurate and effective retrieval is always a difficult problem in image retrieval.In order to improve the precision of image retrieval system and ranking value,this paper is based on the theory of the content of image retrieval technology,in view of the image color,texture and shape feature extraction method for research,design a kind of effective comprehensive characteristics of image retrieval technology and retrieval system.The main research contents of this paper are as follows:(1)In the aspect of extracting the color features of the image,a novel block color volume kernel feature method is proposed.For color quantization,on the basis of overall volume using color histogram of color information and spatial information is described,considering the image at the same time the main goal of the parts are located in the middle area of the image,and in order to solve the similarity calculation in the process of generating linear inseparable problem,put forward the overlapping nine partition strategy and kernel function is introduced,finally to better express color features.(2)In the aspect of extracting the texture features of the image,the traditional LBP operator has poor noise immunity and different structural pattern recognition.Therefore,the texture features cannot be well expressed.Therefore,a novel MLBP operator method is proposed in this paper.Its core is to use(?-?,?+?)a as a local binarization threshold,and in order to express the texture structure of the main part,a certain non-uniform block mode is adopted for the image.(3)In the aspect of extracting the shape features of the image,a novel shape description method based on boundary element features is proposed.The method firstly extracts the color representative value of each boundary mesh on the basis of color quantization;then the boundary mesh of the target object is scanned with the defined four kinds of boundary element descriptors to obtain the boundary element histogram and the boundary element respectively.Related graphs;Finally,the two features extracted comprehensively perform similarity measures.(4)Using VC++ 6.0 design more simple and easy to operate features fusion retrieval prototype system,realized experimental contrast on the same retrieval system based on single feature retrieval and retrieval features fusion experimental contrast.