This thesis investigates motion-adaptive temporal transformations and motion parameter coding schemes, for highly scalable video compression. The first aspect of this work proposes a new framework for constructing temporal discrete wavelet transforms, based on motion-compensated lifting steps. The use of lifting preserves invertibility regardless of the selected motion model. By contrast, the invertibility requirement has restricted previous approaches to either block-based or global motion compensation. We show that the proposed framework effectively applies the temporal wavelet transform along the motion trajectories. Video sequences reconstructed at reduced frame-rates, from subsets of the compressed bitstream, demonstrate the visually pleasing properties expected from lowpass filtering along the motion trajectories. Experimental results demonstrate the effectiveness of temporal wavelet kernels other than the simple Haar. We also demonstrate the benefits of complex motion modelling, by using a deformable triangular mesh. These advances are either incompatible or diffcult to achieve with previously proposed strategies for scalable video compression. A second aspect of this work involves new methods for the representation, compression and rate allocation of the motion information. We first describe a compact representation for the various motion mappings associated with the proposed lifting transform. This representation significantly reduces the number of distinct motion fields that must be transmitted to the decoder. We also incorporate a rate scalable scheme for coding the motion parameters. This is achieved by constructing a set of quality layers for the motion information, in a manner similar to that used to construct the scalable sample representation. When the motion layers are truncated, the decoder receives a quantized version of the motion parameters used to code the sample data. A linear model is employed to quantify the effects of motion parameter quantization on the reconstructed video distortion. This allows the optimal trade-off between motion and subband sample bit-rates to be determined after the motion and sample data has been compressed. Two schemes are proposed to determine the optimal trade-off between motion and sample bit-rates. The first scheme employs a simple but effective brute force search approach. A second scheme explicitly utilizes the linear model, and yields comparable performance to the brute force scheme, with significantly less computational cost. The high performance of the second scheme also serves to reinforce the validity of the linear model itself. In comparison to existing scalable coding schemes, the proposed video coder achieves significantly higher compression performance, and motion scalability facilitates effcient compression even at low bit-rates. Experimental results show that the proposed scheme is also competitive with state-of-the-art non-scalable video coders.
| Identifer | oai:union.ndltd.org:ADTP/241433 |
| Date | January 2004 |
| Creators | Secker, Andrew J, Electrical Engineering & Telecommunications, Faculty of Engineering, UNSW |
| Publisher | Awarded by:University of New South Wales. School of Electrical Engineering and Telecommunications |
| Source Sets | Australiasian Digital Theses Program |
| Language | English |
| Detected Language | English |
| Rights | Copyright Andrew J Secker, http://unsworks.unsw.edu.au/copyright |
Page generated in 0.0012 seconds