Title page for 83345008


[Back to Results | New Search]

Student Number 83345008
Author Chwen-Jye Sze(I³)
Author's Email Address No Public.
Statistics This thesis had been viewed 424 times. Download 5 times.
Department Computer Science and Information Engineering
Year 1997
Semester 2
Degree Ph.D.
Type of Document Doctoral Dissertation
Language zh-TW.Big5 Chinese
Title Wavelet-based Early Vision Processing
Date of Defense
Page Count 148
Keyword
  • Differential Geometry
  • Heat Conduction
  • Range Edge Detection
  • Selective Image Smoothing
  • Wavelet
  • Abstract In this dissertation, a systematic study of wavelet-based early
    visionprocessing using different theoretical foundations has
    been presented.The work applies the heat conduction theory and
    the differentialgeometry theory, respectively, to the selective
    image smoothing problemand the range edge detection problem. In
    the first topic, a new wavelet-based selective image smoothing
    scheme using an image flux conduction model is proposed.
    Basically, an image smoothing process can be simulated by a heat
    conduction process through mathematical formulation. But a heat
    conduction process borrowed from thermal physics may not be
    applied to image smoothing directly. Therefore, a discrete image
    flux conduction model has been developed by modifyingthe
    original heat conduction model. For computing the image flux
    valuesmore accurately, a differential operator derived from
    wavelet-based approximation is introduced into the developed
    model and a new wavelet-based discrete conduction equation is
    proposed. This proposed new equation is proved to satisfy the
    so-called maximum-minimum principle. The principle provides a
    more restrictive stability criterion than the von Neumann
    stability. When an iterative function system satisfies this
    principle, the system will eventually converge.The second issue
    which have been discussed in this dissertation wasthe range edge
    detection problem. In this work, differential geometry was
    introduced as a theoretical foundation. A new 3-D invariant
    feature,the magnitude of normal changes alone two orthogonal
    directions, was introduced to detect the boundaries of different
    surface patches. Since the image data were discrete by nature,
    several surface approximation schemes including quadratic
    surface fitting, orthogonal wavelet-based approach, and non-
    orthogonal wavelet-based approach were applied to approximate
    the continuous data surface. Thus, the multiscale wavelet
    transform was introduced to detect the set of edge points. These
    points were selected once their magnitudes of a normal change
    exceed a preset threshold. Experimental results have
    demonstrated that the proposed wavelet-based approach is indeed
    superb in range edge detection.
    Table of Content Cover
    Contents
    List of Figures
    1. Introduction
    1.1 Early Vision Processing
    1.2 Related Works
    1.2.1 Survey on Nonlinear Diffusion/Conduction Filtering
    1.2.2 Survey on Range Edge Detection
    1.2.3 Survey on Wavelets
    1.3 Overview of This Dissertation
    1.4 Contributions of THE dISSERTATION
    1.5 Organization of this dissertation
    2. Background Knowledge on Heat Conduction ,Differential Geometry and Wavelets
    2.1 Heat Conduction Transfer
    2.1.1 Fourier's Law of Heat Conduction
    2.1.2 Heat Conduction Equation
    2.1.3 Thermal Conductivity
    2.2 Differential Geometry
    2.3 Wavelet Theory
    2.3.1 Preliminaries and Notation
    2.3.2 Wavelet Transform and its Inverse
    2.3.3 Multiresolution Analysis
    2.3.4 Orthogonal Wavelets
    2.3.5 Orthogonal Wavelet-based Differential Operators
    2.3.6 Non-orthogonal Wavelets
    3. Selective Image Smoothing Based on An Image Flux Conduction Model
    3.1 Intreduction
    3.2 COnductivity Design for Selective Image Smoothing
    3.3 Image Flux Conduction Analysis
    3.3.1 Some Potential Poblems Which May Be Encountered When Solving A Continuous
    Heat Conduction Equation Based on A Discrete Data Set
    3.3.2 An Image Flux Analysis Model
    3.3.3 Isotrpic Wavelet-based Image Flux Conduction
    3.3.4 Anisotropic Wavelet-based Image Flux Conduction
    3.4 Experimental Results
    3.4.1 Experiments on One-dimensional Ideal Signal
    3.4.2 Experiments of Isotropic Conduction on Real Images
    3.4.3 Experiments of Anisotropic Conduction on REAL iMAGES
    3.5 Discussions
    3.6 Concluding Remark
    4. Multiscale Range Edge Detection Based on Normal Changes
    4.1 Range edge detection via normal change
    4.2 Calculation normals from discrete surface points
    4.3 Detceting edgees based on abrupt normal changes
    4.4 Experimental results
    4.5 Concluding Remarks
    5. Conclusions and Future Direcions
    5.1 Conclusions
    5.2 Future Works
    Bibliography
    Reference
    Advisor
  • Kuo-Chin Fan, Hong-Yuan Mark Liao(SM, )
  • Files No Any Full Text File.
    Date of Submission

    [Back to Results | New Search]


    Browse | Search All Available ETDs

    If you have dissertation-related questions, please contact with the NCU library extension service section.
    Our service phone is (03)422-7151 Ext. 57407,E-mail is also welcomed.