Date of Award

Summer 1999

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Computer Science

Committee Director

Hussein Abdel-Wahab

Committee Member

Kurt Maly

Committee Member

Shunichi Toida

Committee Member

Christian Wild

Committee Member

Martin Meyer

Abstract

As the demand for image and video transmission and interactive multimedia applications continues to grow, scalable image and video compression that has robust behavior over unreliable channels are of increasing interest. These desktop applications require scalability as a main feature due to its heterogeneous nature, since participants in an interactive multimedia application have different needs and processing power. Also, the encoding and decoding algorithm complexity must be low due to the practical considerations of low-cost low-power receiver terminals. This requires image and video encoding techniques that jointly considers compression, scalability, robustness, and simplicity.

In this dissertation, we present a family of image and video-encoding techniques, which are developed to support conferencing applications. We achieve scalability, robustness and low computational complexity by building our encoding techniques based on the quadtree and octree representation methods.

First we developed an image encoding technique using the quadtree representation of images and vector quantization. We use a mean-removal technique to separate the means image and the difference image. The difference image is then encoded as a breadth first traversal of the quadtree corresponding to the image. Vector quantization is then used to compress the quadtree nodes based on the spatial locality of the quadtree data. Our next step was to use the quadtree-based image encoding technique as a base for developing a differential video encoding technique. We extended it to encode video by applying the well-known IPB technique to the image encoding system.

Then, we explore another method of extending our image encoding technique to encode video streams. The basic idea was to use exactly the same three steps used in our image encoding technique, mean removal, conversion to tree structure, and vector quantization, and replace the quadtree structure with an octree structure. The octree is the three-dimensional equivalent of the quadtree. We divide the sequence of frames into groups and view each group as a three-dimensional object. By encoding frames together, we can obtain substantial savings in encoding time and better compression results.

Finally, we combined both the differential quadtree and octree approaches to generate a new hybrid encoding technique. We encode one frame using the quadtree-based image encoding technique, and then encode the following group of frames as a differential octree based upon the first frame.

Using a set of experiments, the quadtree-based image encoding and differential video encoding techniques were shown to provide reasonable compression in comparison with similar techniques, while the octree and hybrid video encoding techniques gave impressive compression results. Furthermore, we demonstrated that our encoding techniques are time efficient compared to the more common frequency based techniques. We also compare their scalability feature favorably with other well-known scalable techniques. Moreover, we demonstrated their ability to tolerate and conceal error. The new encoding techniques proved to be efficient methods of encoding for interactive multimedia applications.

DOI

10.25777/1tn3-sg67

ISBN

9780599525115

Share

COinS