Research On Method Of Images Classification In World Wide Web

With the rapid development of information technology,the construction of information thruway and the broad application of Internet, people has been in society of information. In information society, Internet has supplied people lots of usable resources,such as text, picture, video and audio ect.And Internet has already been the important resources of eduction. But kinds of Internet information arrange complex, this will bring diffictult to the retrieval of information.Nowadays, Content-based image retrieval has been improved rapidly, mainly based on the low-level features from images, but product based on the high-level features has't successd. The thesis focus on the classification of images collected on the World Wide Web.The classification consists in seprarating the images into two classesricon and picture. The size of icons ususally is small, the effect of them is to provide clue and identification to users. And pictures usually contain contents.Pictures can be classed into two classes:photograph and graphic. And photograph can be seprarated into two classes:artifical scenery and natural scenery.Chart and brushwork form graphic. For different images, analyse them and draw the important parameters, use the parameters to classification. This thesis also has improved Hough transformation, and has applied it to the classification of graphic.The thesis has also made study on the classification of texture. Using Gabor filters to extract texture features from the texture image, with Tamura features, apply the BP Neural Network to classification.The whole test plat is based on Win2000 OS, and explored by Visual C++6.0. The result of the test indicate that the arithmetic of image classification is effective.