Web Image Adaption for Heterogeneous Network Connection Clients Based on Semantic Value

Abstract

Web content transcoding on proxy is a promising way to make Web documents adaptive to heterogeneous network conditions and browsing terminals. Its effectiveness is not only making the servers maintain just one or a few versions of the content, but also reducing the data size delivered to clients. As images have much larger size compared with texts and the transcoding of them can significantly reduce the totally delivered bytes, in this paper we mainly pay attention to the adaptive transcoding of Web images. Firstly, in order to quantify the information provided by Web images, we proposed the concept of ”semantic value”. Secondly, the idea of image classification combined with user study is taken to assign the semantic value to Web images. Then an adaptive model is designed based on a survey of the transcoding characters of Web images and the semantic value assigning strategy. The model will optimally deliver the Web contents over heterogeneous networks by deciding proper transcoding way for each image so as to maintain as much semantic value of transcoded Web document as possible. Extensive experiments on 100 Web pages containing 4211 images have shown the effectiveness and efficiency of this work.