Research Of Image Retrieval Based On Color And Texture

With the multimedia and communication technology rapid developed, digital image has been involvedin all areas of people’s lives, and becomes an important component in multimedia information. Along withthe development of network technology and the increasing popularity of image collection equipment, thetype of digital images is expanding continually, and how to find the user needs images quickly andaccurately to become an important aspect of information retrieval. Content-based image retrieval has beenwidespread concern, its basically is to use the color, texture, shape, spatial relationships and semanticcharacteristics to image retrieval. The quality of these features extraction has a direct impact on the resultsof image retrieval, so this article based on the characteristics of image retrieval do the following twoaspects.1. This paper proposed a image retrieval algorithms based on the refined edge histogram. Color is animportant visual feature of images. However, when extracting color features it will loss spatial informationand lead to false retrieval. In this paper, we present a "Back"-shape regional division approach and combinewith pyramid histogram of orientated gradients (PHOG) to extract image edge features, termed refinededge histogram. Moreover, the refined edge histogram is applied to color image retrieval. Experimentalresults show that the proposed EDH are suitable for color image retrieval and has higher precision andrecall compared to other existing methods.2. This paper proposed a image retrieval algorithms based on the texton fusion histogram. Naturalimages include rich color and texture informations, from the micro-structure that there is a closerelationship between color and texture. In this paper, we use four texton templates to extract the texton image of the color image. It is not only reflects the color and texture information fully, but also avoid theloss of color space information, and combined with the gradient angle of the image, construct a fusionhistogram for image retrieval. Experiments shows that this method retrieved similar images more priorityand has higher precision and recall than single features and traditional composite character.