Optimal Quality-Aware Predictor-Based Adaptation of Multimedia Messages

The Multimedia Messaging Service (MMS) platform allows messages composed of various multimedia attachments to be exchanged between users of mobile devices. However, these heterogeneous devices exhibit different capabilities regarding what media types, resolution, and maximal message size they support making the adaptation of messages mandatory to ensure compatibility between sending and receiving devices. The challenge is therefore to adapt messages so that they satisfy the receiving device’s constraints in a way that both maximizes the user experience and minimizes the computational cost of adaptation. Minimizing computational cost will help cope with high-volume traffic while maximizing user experience, as estimated by the perceived quality of adapted messages, will secure the service provider’s user base. In this work, we propose a generic adaptation scheme based on predictors for file size and image quality resulting from transcoding parameters applied to a given image that will explicitly maximize perceived quality as estimated by the structural similarity image quality index. We will further show that our proposed method is resilient to the imprecision of the predictors and that it yields significantly better quality at a greatly reduced computational cost compared to other methods proposed in prior art.