Temporal-spatial coding and processing techniques for digital video applications.

摘要

This dissertation is about video coding and processing techniques to enhance both compression efficiency and playback quality for various applications such as video streaming, transcoding and frame rate conversion, etc. After a brief review of' background; the motion compensated frame-rate up-conversion technique is first discussed in Chapter 3. This technique is applied to ensure the full frame-rate playback when some frames are missing due to the bandwidth constraint or the poor network transmission. A hybrid global-local motion compensated frame interpolation scheme is introduced to reconstruct those missing frames.; The joint temporal-spatial rate control problem is investigated in Chapter 4, where the temporal frame rate and the spatial quantization step size are jointly optimized during the encoding procedure to achieve good temporal-spatial quality trade-off. Based on a rigorous problem formulation; several solutions are presented to optimize the rate-distortion performance of the compressed video. Furthermore a set of rate-distortion modeling functions are developed in Chapter 5 to reduce the computational complexity of the solutions presented in Chapter 4 so that joint temporal-spatial rate control can be applied to online applications.; Video transcoding from MPEG-2 (high quality, high bit rate, large picture size) to MIPEG-4 (lower quality, lower bit rate, smaller picture size) is examined in Chapter 6. An efficient transcoder with a low complexity is proposed by re-using the motion vectors carried by the MPEG-2 compressed bit stream. It includes both picture and motion vector down-sampling and exploits joint temporal-spatial rate control. The proposed transcoder can provide better transcoding quality than a full decoder cascaded by a full encoder at a lower complexity. Thus, it provides an excellent solution to real-time transcoding applications.