Image transmission presently is the main bottleneck for real-time web browsing in low-bandwidth settings. Our goal is to substantially reduce the amount of data transferred when a web page with embedded images is transmitted by one to two orders-of-magnitude. Potential applications include wireless handhelds and remote networks of workstations with low bandwidth connections to web servers on another network. One scenario is that of an off-shore network in a command-and-control center, with many independent users accessing remote data.
Our target systems are standard browsers, servers, and publishing tools. We propose to use foveation-based multi-resolution images to dramatically reduce bandwidth requirements for image transmission. Foveal image rendering systems initially transmit low resolution images. Specific portions of thes multi-resolution foveal image are successively refined in response to the user's indication of interest by pausing or clicking of their mouse.
A key feature of our design is that neither the browser, the web pages, nor the server where they reside need be specialized for our mode of web browsing. This allows our solution to be compatible with existing intra- and inter-net infrastructures. We propose to implement a web proxy that replaces in-line images that would normally require full transmission via the low-bandwidth link with foveated renderings served by a foveated image server on the remote site. Our proposed implementation is scalable: cacheing foveal image servers can be cascaded so that multiple users viewing the same image do not transfer the same data multiple times over the same low-bandwidth data links and can benefit from data already transmitted.
For further information, please contact either Eric Freudenthal or Chee Yap.