Please use this identifier to cite or link to this item:
Title: Model-based video coding techniques
Authors: Siu, Manson
Degree: M.Phil.
Issue Date: 2001
Abstract: This thesis is primarily concerned with how to integrate a model-based coding system. A large portion of this thesis is devoted to our proposed algorithms, which address the solutions for the difficulties encountered in a model-based coding system for real world applications. Apart from that, this thesis also covers our proposed model-based coding system. In this thesis, we first discuss some existing model-based coding systems, and pinpoint the areas that limit the practicality of the application of model-based coding techniques. Typical approaches used in existing systems employ dedicated graphic models for different objects of interest. The objective of our work is to investigate and tackle the limitations that commonly encountered in practical applications of model-based coding. Accordingly, a robust model generation technique and a universal texture extraction technique are developed. Integrated with the aforementioned techniques, our system is capable of generating object models by deforming a graphic model in a model library according to the given image source. The deformation results are object models that well describe the objects of interest. This eliminates the necessity to install a specific model for each object of interest. Whereas, the proposed universal texture extraction technique provides a solution to obtain the texture information of an object without limits of the object's orientation and size. We also investigated some facial animation techniques during the integration of our system. There are two advantages by implementing these facial animation techniques in our system. First, it results in a further reduction in the channel bandwidth required. This is achieved by the concise expression of the animation parameters for describing the change of a graphic model. Second, the possibility that extends the system to a synthetic coding system is preserved. Finally, we provide some solutions to realize the critical parts of a system. These critical part significantly affect the practicality of a model-based coding technique. In a practical system, minimal knowledge requirement is necessary. Unlike most of the existing model-based coding systems, tailor-shaped head models are not required in our system. It is not necessary to provide a specific head model for a particular user. Furthermore, our system is designed to be scalable in alleviating the heavy computation overhead, and therefore it is favourable to real world applications.
Subjects: Digital video
Coding theory
Hong Kong Polytechnic University -- Dissertations
Pages: xvii, 94 leaves : ill. (some col.) ; 30 cm
Appears in Collections:Thesis

Show full item record

Page views

Last Week
Last month
Citations as of Jun 4, 2023

Google ScholarTM


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.