Optimal rate control of compressed video

1.  Introduction

This paper concerns the optimal allocation of bits to individual video frames when compressing a video sequence, where the average number of bits per frame is given as a prior constraint. The optimization algorithm developed here is suitable for offline compression; it requires multiple passes through the video sequence.

The raw information content of frames in a video sequence differs. Some frames have high redundancy, showing large expanses of uniform colour. Others show lots of detail, and as such, inherently require more information to encode them. The conditional information of frame n in a sequence given frame n-1 also varies. It is high when a sharp scene change occurs or where there is a lot of rapid motion, lower for fades and wipes, and almost zero for continuous and unchanging scenes. The general problem we are addressing in this paper is how to optimally adjust the bit rate of a compressor to accomodate these changes in conditional information accross a complete video.

A report on this work is found in pdf format here and in html format there..

Tour of Glagow demo

Here we present a couple of video sequences two of which are compressed using the optimal rate control algorithm and two ( on the left ) use constant bit rate.

Short sequences (100 frames)

Constant bit rateMinimized MSE

Long sequences

Constant bit rateMinimized MSE