Efficient Clustering-based Algorithm for Predicting File Size and Structural Similarity of Transcoded JPEG Images

The problem of adapting JPEG images to satisfy constraints such as file size and resolution arises in a number of applications, from universal media access to multimedia messaging services. Visually optimized adaptation, however, commands a non-negligible computational cost which we aim to minimize using predictors. In previous works, we presented predictors and systems to achieve low-cost near-optimal adaptation of JPEG images. In this work, we propose a new approach to file size and quality prediction resulting from the Transcoding of a JPEG image subject to changes in quality factor and resolution. We show that the new predictor significantly outperforms the previously proposed solutions in accuracy.