Optimal quality-aware predictor-based adaptation of multimedia messages

Multimedia Messaging Services (MMS) allow messages composed of different media attachments to be exchanged between heterogeneous devices. In this work, we consider the special case of image-only messages, where the challenge is to adapt messages so that the receiving device constraints are satisfied while maximizing the user´s perceived quality of adapted messages. We propose an adaptation algorithm based on predictors for file size and quality resulting from transcoding parameters that explicitly maximizes an objective function based on the structural similarity image quality index. We show that the proposed method is not only resilient to the imprecision of predictors, but also yields significantly better quality at reduced computational complexity compared to other methods proposed in prior art.