Research On Image Retrieval Technology Based On Color And Texture Features

With the rapid development of network and computer technology, image data showing rapid growth in geometric progression, so large capacity image database emergede. This also brought the image management and retrieval unprecedented difficulties. Content based image retrieval is the most common and reliable way to image database management and retrieval. Content based image retrieval is based on the use of color, texture, shape and other features,then complete matching, quickly find the target image. But digital devices can not always be intelligent and flexible like the human brain, it can not exactly match low level features and high level semantic features of the image, thus prevent people integrately use visual perception and mental perception to find images. Therefore, the human brain naturally strong cognitive abilities introduced into image retrieval system is widely recognized as the best way to effectively improve the accuracy of image retrieval.This paper introduced the human computer interaction relevance feedback into content based image retrieval system to achieve a more effective image retrieval. The main work and achievements are as follows:Firstly, comprehensively and integrately presented the content based image retrieval technology. Introduced the color, texture, shape feature representation methods and similarity matching,and focuses on the use of color and texture features of images for image retrieval. This paper used the partition and block histogram to express color feature, not only retains the convenience and robustness advantages in global histogram, but also applicated the spatial information in block histogram, improved the retrieval accuracy.Secondly, use GLCM to extract image’s texture features for image retrieval based on texture features. And integrately use color and texture features of images to achieve a comprehensive feature based image retrieval, and improved the retrieval accuracy.Thirdly, research for relevance feedback technology in image retrieval. The support vector machine theory is applied to the image relevance feedback, the realization of human computer interaction improved retrieval satisfaction, and the selection of SVM parameters has been discussed.Experimental results show that image retrieval system based on integrated features and relevance feedback shows a better retrieval precision than image retrieval system based on a single feature,it experienced better at retrieval effect.