Traditional and Scalable Coding Techniques for Video Compression

In recent years, the usage of digital video has steadily been increasing. Since the amount of data for uncompressed digital video representation is very high, lossy source coding techniques are usually employed in digital video systems to compress that information and make it more suitable for storage and transmission. The source coding algorithms for video compression can be grouped into two big classes: the traditional and the scalable techniques. The goal of the traditional video coders is to maximize the compression efficiency corresponding to a given amount of compressed data. The goal of scalable video coding is instead to give a scalable representation of the source, such that subsets of it are able to describe in an optimal way the same video source but with reduced resolution in the temporal, spatial and/or quality domain. This thesis is focused on the presentation, the analysis and the design of some coding techniques already used or still under investigation for both traditional and scalable video coding. Amongst the advanced coding techniques that we analyze, we present the so-called joint predictive transform coding. We will show the superiority of this technique over pure transform based approaches, and point out that the new approach for intra-frame coding of the latest international video coding standard H.264/AVC is a two-dimensional application of this paradigm. Another advanced technique still under investigation for both traditional and scalable compression is represented by trellis coded quantization. We review this technique into different frameworks and give an alternative approach to the design of good trellis coded quantizers that produces slightly increased performance with respect to the previous results found in the related literature. The tools usually considered for scalable approaches to the video coding problem are then reviewed. Some of these tools are used to design a very low-complexity wavelet-based video coder. This coder uses a novel wavelet filter to offer spatial scalability in video coding. A novel approach to the rate-distortion optimization problem for wavelet-based compression of the error frames in predictive coding is presented as well.