Scalable Lossless Video Coding Based on Adaptive Motion Compensated Temporal Filtering

In this paper, we propose a lossless video coding system with temporal scalability support for preview purpose. The backbone of the compression consists of an adaptive MCTF scheme and a spatial lossless predictive coding scheme. The former aims at both inter-frame coding and temporal scalability while the latter focuses on lossless coding of temporal prediction residuals in the spatial domain. Unlike the conventional MCTF scheme, the proposed one can dynamically select either inter-frame or intra-frame predictions subject to the property of the frame image. The inter-prediction is further divided into three modes, i.e. forward, backward and bidirectional to reduce side information of motion vectors. The prediction mode is determined on a block basis and can thus effectively explore data redundancy temporally or spatially subject to the variations in blocks. The adopted lossless spatial prediction scheme is basically based on JPEG-LS followed by arithmetic coding. Experimental results indicate that our system can outperform rival systems based on JPEG-2000, JPEG-LS and H.264.