Quality-aware predictor-based adaptation of still images for the multimedia messaging service

Abstract

The Multimedia Messaging Service (MMS) allows users with heterogeneous terminals to exchange structured messages composed of text, images, sound, and video. The MMS market is growing rapidly, posing the problem of MMS adaptation, which is necessary to ensure terminal interoperability. Message adaptation involves technological challenges, especially considering the high volume of messages that this service can handle. In this work, we propose novel predictor-based dynamic programming approaches to MMS adaptation, which provide a framework for explicit maximization of the user experience, rather than relying on heuristics to deliver adapted messages satisfactorily. We show that the proposed solutions lead to noticeably superior image quality and faster transcoding times than comparable algorithms offered in products currently on the market and those described in the literature.

Predicting and Optimizing Image Compression

“…Coulombe and Pigeon present their own predictor of file size and image complexity for JPEG over a number of papers describing similar techniques [14,5,15,16,17,11,18]. The key idea underpinning these works is to use the compression ratio of an already encoded image as input to nearest neighbor and table-based predictors, under the assumption that file size and quality grow monotonically with increasing quality factor, and neighbor values of quality factor correspond to similar compression ratio and quality level.…”