Image Retrieval Research And Implementation Combining Color And Texture Feature

With the rapid development of the multimedia technology and Internet, the scale of image resource is growing large and large. Meanwhile, the image retrieval technology based on the content is also developing.This paper firstly introduces the background and the meaning of this topic, making a simple analysis about the current development of the image retrieval technology based on content, and discusses a number of key technologies which are involved in this area.Color is the most direct visual characteristics of the physical characteristics of image. This system extracts the color feature of the image by using color histogram. First of all, this system quantifies the color characteristics of images to eight values. Then calculate the color histogram of image and the similar distance measured by Continental distance.Texture is another important feature among the physical characteristics of images. This system uses Sobel operator and gray-scale co-occurrence matrix to calculate texture feature of the image. It is necessary to gray the image before extracting texture feature. Then, the system can extract the binary image which is used for the calculation of gray image co-occurrence matrix gray image next on the basis of gray image, and the texture features. Finally calculate similar distance measured by Continental distance.The texture, the color and so on in the image retrieval are the extremely frequent used of the first floor vision characteristic . But as a result of person eye's special physiological structure , when observes the image often is not only based on a characteristic, but synthesizes many characteristics. The experiment indicated: Regardless of being inquired the speed or the accuracy, based on many kinds of characteristic's image retrieval systems is always better than the sole characteristic image retrieval system. The third function of this system is the image retrieval integrated color and texture features. First withdraws the color characteristic and the textural property vector separately.Then carries on these two vectors the exterior normalization, causes the different type the eigen feature to be the same in the similar distance computation status. Finally the choice suitable weight causes the image the retrieval effect to be best.This system is truly very good after the test retrieval effect.