|M.Sc Student||Dar Yehuda|
|Subject||Spatio-Temporal Bit-Allocation for Low Bit-Rate Video Coding|
|Department||Department of Electrical Engineering||Supervisor||Professor Alfred Bruckstein|
|Full Thesis text|
The digital video is a 3D spatio-temporal signal formed as a sequence of 2D images (frames) captured over time. Hence, a sampled video represents a large amount of information. As a result, video transmission and storage systems require efficient coding and should be analyzed from a rate-distortion perspective. Specifically, good quality video coding for low bit-rate applications has great importance for transmission over narrow-bandwidth channels and for storage with limited memory capacity.
In this work, we propose a spatio-temporal scaling-compression system for improving video coding at low bit-rates. A spatio-temporal analysis of the compression is proposed, and the optimal spatio-temporal down-scaling factors are examined. We show, both analytically and experimentally, that at low bit-rates, we benefit from applying a spatio-temporal down-scaling, i.e., reduction of frame-rate and frame-size, before the compression and a corresponding up-scaling afterwards. This suggested procedure improves the compression efficiency at low bit-rates only by means of estimating the optimal down-scaling using analytic model of the compression, and performing scaling operations outside the codec.
Our analysis of profitability of spatio-temporal down-scaling for video compression is based on a theoretical model of the compression at low bit-rates in a wider scope than usual. A video signal model that is suitable for multi-resolution analysis. Then, we analyze low bit-rate compression as carried out in H.264's baseline profile that serves as an exemplary hybrid video codec. Specifically, the used coding modes, motion-compensated coding, transform coding and bit-allocation are studied. We examine the overall compression-scaling system and formulate a bit-allocation optimization problem for finding the optimal down-scaling factor of a given video by its second-order statistics and a bit-budget.
Another contribution of this work is a thorough study of motion-compensated prediction. We examine the effect of frame-rate and compression bit-rate on block-based motion estimation and compensation as commonly utilized in inter-frame coding and frame-rate up conversion (FRUC). This joint examination yields a comprehensive foundation for comparing motion-compensation (MC) procedures in coding and FRUC. First, the video signal is modeled as a noisy translational motion of an image. Then, we theoretically model the motion-compensated prediction of an available and absent frames as in coding and FRUC applications, respectively. The theoretic MC-prediction error is further analyzed and its autocorrelation function is calculated for coding and FRUC applications. Furthermore, we propose two constructions of a separable autocorrelation function for prediction error in MC-coding.