Efficient Near-Optimal Dynamic Content Adaptation Applied to JPEG Slides Presentations in Mobile Web Conferencing

In the context of mobile Web conferencing, slide documents are generally transcoded into JPEG format and wrapped into a Web page prior to delivery. Given the diversity of these devices and their networks, dynamically identifying the optimal transcoding parameters is very challenging, as the number of transcoding parameters combinations could be very high. Current solutions use the resolution of the target mobile device and a fixed quality factor as transcoding parameters. However, this technique allows no control over the resulting file size, which, if too large, might increase the delivery time and negatively affect users´ experience. Another solution (content selection) which leads to better quality consists in creating several versions and, at delivery time, selecting the best one. However, such a solution is computationally expensive. In this paper, we propose a prediction-based framework which computes near-optimal transcoding parameters dynamically with far less computations. We propose five methods based on this framework. The first predicts near-optimal transcoding parameters, while the others improve their accuracy. From the set of documents tested, two of the proposed methods reach optimality 14% and 30% of the time, respectively. Moreover, the average deviation from optimality for the proposed methods varies from 6% to 3%, with a complexity varying from 1 to 5 transcoding operations.