A Multi-domain Motion Vectors Prediction Method for Spatial Scalability

In scalable video coding, spatial scalability is a layered video representation with different spatial resolutions for the same video signal. Inter-layer motion information prediction is a method of improving coding performance for spatial scalability. In this paper, a multi-domain motion vectors prediction coding method for spatial scalability is proposed based on inter-layer motion information prediction. It consists of two spatial layers: one base layer and one enhancement layer. The proposed coding method is fully compatible with existing coding standard in base layer, while in enhancement layer a motion vector prediction mode is added. If this mode is selected, the motion vectors of temporal and spatial and inter-layer domain are exploited for motion vector prediction of current macroblock in enhancement layer. It will obtain a more accurate motion vector by using multi-domain motion vectors. The effectiveness of the new technique is demonstrated by simulation experimental results. With almost no changes on both computation complexity and encoded bitrate, the average increase of Y-PSNR is about 0.22 dB for sequence Foreman. and 0.17 dB for Subway.