Temporal optimization of quality during video compression

Recent VQM developments have provided a set of methods for evaluating the relative subjective quality of a treated video segment compared to the original video. In this paper we intend to optimize the subjective quality of multi-segment video sequence for a given global bitrate constraint, by adjusting some basic compression parameters such as the Quantization Parameter, the frame rate and the image size along the video sequence. An optimal resource allocation algorithm has been designed to efficiently allocate the global video bit-budget to the different shots (i.e. segments), depending on the content of shots, by using 2 different optimizing criteria: maximize the minimum perceived quality over the sequence of shots, and maximize the average perceived quality over the sequence. We have studied the impact of such an approach on more than 800 shots extracted from 15 videos of different content.