An Approach For Image Retrieval With Feedback Based On Content

As the rapid development of computer and multimedia technology, the image retrieval technique based on low layer features becomes one of the hottest spot. Speak briefly, it uses the lower characteristics of image such as color, texture and shape features to find a similar image set for the input image in the image database. Notice that the image's basic composition is pixel in a computer, so the subject not only studies the above three characteristics for the image retrieval, it also takes the pixel as a minimum level feature of image and analyses the principle component of the image matrix which is the original data. So it can achieve image retrieval with such "principle component". For getting the better results of the retrieval, it also integrates the low level features and adds the relevant feedback information in the retrieval system.During the process of integrating the lower features, it first combine the color and the texture features, it takes each image for constituting the principal and background part, so separating them out, and then extracting the color feature of the principal portion and the texture feature of the image background for retrieving image.For connecting the users'mind during the retrieval, the paper also designs the image retrieval system with feedback. At first, it does one retrieval bases on the color, texture and shape features. And then, the system can accept the related images that the user clicked on the first retrieval results, it computes the influence factors brought by these images. Finally, the system will output the better retrieval result.