Design And Implementation Of Color-based Image Retrieval System

With the rapid development of computer technology and the advances in technology and Internet multimedia, image resources have become more abundant, and the image usually contains a lot of rich information, how to search the user needs information from large image databases accurately and efficiently, the CBIR (Content-Based Image Retrieval) solved this problem better than other methods, the color characteristics of human vision has the most obvious and most intuitive features in content-based image retrieval, and its calculation of color feature simple, and has robust characteristics, now it has became an important visual image information, the current Color Image Retrieval in forest fire prevention, weather forecast, trademark search, and other fields has been widely used.This thesis discusses and compares several common color space, and on this basis, focusing on the color histogram similarity measurement and color feature extraction method, for the color histogram method only considered the color on the image collection distribution, without taking into account the local color on the spatial distribution in the image, an improved method by irregular blocks of the image has been proposed, calculate the color histogram of each block, and given different weights, after testing it has improve the image retrieval accuracy.Finally, based on the above method, adopting the JSP technology and MVC model, established color feature based image retrieval system as a test platform for image retrieval. Compared the color histogram algorithm and the improved weighting block algorithm; the result of the experiment has proved the last is better than the first method.