A Study of Functionally Layered Video Coding and Its Applications (映像の機能的階層符号化とその応用に関する研究)

学位の種類 博士(工学)
学位記番号 博甲第527号
学位授与の日付 平成22年3月25日
学位論文題目 A Study of Functionally Layered Video Coding and Its Applications (映像の機能的階層符号化とその応用に関する研究)
 主査 准教授 岩橋 政宏
 副査 教授 吉川 敏則
 副査 教授 山田 耕一
 副査 准教授 湯川 高志
 副査 准教授 木村 哲也

平成21(2009)年度博士論文題名一覧] [博士論文題名一覧]に戻る.

Chapter 1 p.1
Introduction p.1
 1.1 Backgound of the research p.1
 1.2 Objective of the research p.17
 1.3 Organization of the dissertation p.18
Chapter 2 p.19
Video coding system and its components p.19
 2.1 Introduction p.19
 2.2 Digital video signal and its data volume p.19
 2.3 Signal Transformation(DCT,DWT) p.22
 2.4 Entropy coding (Variable length coding) p.33
 2.5 Layered video coding (scalable video coding) p.34
Chapter 3 p.37
Water Level Detection p.37
 3.1 Introduction p.37
 3.2 Water level detection by using video signal p.39
 3.3 Problem of the existing method p.41
 3.4 Water level detection utilizing spatial FIR filtering p.43
 3.5 Water level detection utilizing JP2K wavelet transform p.56
 3.6 Conclusions p.62
Chapter 4 p.64
Functionally layered coding for water level detection p.64
 4.1 Introduction p.64
 4.2 Overview of river monitoring systems p.64
 4.3 Functionally layered video coding for water level detection p.68
 4.4 Experimental results p.73
 4.5 Conclusions p.91
Chapter 5 p.92
Functionally layered coding for robot vision network p.92
 5.1 Introduction p.92
 5.2 Overview of robot vision network p.93
 5.3 Functionally layered coding for robot vision network p.96
 5.4 Experimental results p.101
 5.5 Conclusions p.109
Chapter 6 p.110
 6.1 Summary p.110
 6.2 Future works p.111
Biblography p.114
List of Publications p.129
 ‐Journal Articles p.129
 ‐Conference Proceedings p.129

Digital "data compression" (source coding) has been playing an important role as a key technology for storage and communication of image and video signals. It has wide variety of applications such as digital TV, movie, DVD, mobile phone, remote surveillance, intelligent car navigation and so on. This dissertation proposes a functionally layered video coding which reduces data amount to be transmitted via a digital network.
A conventional layered video coding based on MPEG international standard compresses data amount of a digital video signal. It outputs two kinds of bit-streams - "base layer" and "enhancement layer" to be stored and transmitted. A rough and low quality signal is reconstructed from the "base layer" bit stream. Additionally decoding the "enhancement layer", high quality signal is reconstructed. It is utilized for a broad casting to two types of receivers - low quality and high quality.
The "base layer" of the conventional approach is designed for a user (people) so that we can browse a rough image or video from it. On the contrary, in the proposed approach, it is designed for a robot to perform a "function". In this dissertation, "water level detection" and "auto location of a robot" are introduced as examples of the function. It aims at constructing a new concept of layered coding specially designed for communications among "people and robots".
Chapter 1 describes how the following chapters are related each other. It also includes brief explanation of basic technical terms necessary for understanding the following chapters.
Chapter 2 describes background of the main theme of this dissertation. Its necessity and essential difference between conventional approaches are explained.
Chapter 3 is a preliminary section to the chapter 4. It proposes an optimization procedure of an image signal processing filter to detect "water level". It concludes that a combination of temporal low pass filtering and spatial band pass filtering achieves robust and stable water level detection.
Chapter 4 proposes a functionally layered coding for "water level detection". The optimum basis functions for this functionality are found utilizing new findings in the previous chapter. In this approach, a video signal is decomposed into some band signals and bit planes based on discrete wavelet transform of JPEG 2000. Only a part of the components necessary for this function is embedded into the base layer so that data amount is much more compressed.
Chapter 5 proposes a functionally layered coding for "auto location of a robot". The optimum basis functions for this functionality are determined. The chapter 4 and 5 conclude that the proposed approach successfully achieves both of data amount reduction and high precision recognition.
Chapter 6 summarizes the dissertation. It also describes discussion and remaining issues as future works. As described in the dissertation, utilizing the proposed approach, "compression" and "recognition" are synthesized into a unified simple signal processing. It contributes to construct a small size signal processor chip with long time battery operation in an intelligent sensor node

 本論文は、「A Study of Functionally Layered Video Coding and Its Applications (映像の機能的階層符号化とその応用に関する研究)」と題し、6章より構成されている。


 第2章「Video Coding and its Components」では、映像の信号処理を構築するための要素技術について、基底分解、符号化、認識処理を中心に、次章以降を理解する上で必要な事項を簡潔にまとめている。

 第3章「Water Level Detection」では、水位検出という機能を実現する上で必要となる映像信号処理について、信号処理の各要素に対する最適化問題を設定している。また、工学的に有用な解を導き、それが十分実用に耐えることを、各種映像実験を介して検証している。

 第4章「Functionally Layered Coding for Water Level Detection」では、ディジタル映像の階層符号化において、信号波形を複数の基底成分に分解し、その中からある特定の機能を実現する上で必要最小限となる成分を抽出し、基本階層として伝送することで、通信に要するデータ量を大幅に削減できる方式を提案している。機能としては、河川水位の自動検出が取り上げられ、提案法の有効性を、各種の実験を介して検証している。

 第5章「Functionally Layered Coding for Robot Vision Network」では、前章に述べられた機能的階層符号化法を、探査ロボットによる環境マップの生成問題に応用している。