PDF version of this entire document

Contents

• Introduction
  • Image Registration
  • Statistical Appearance Models
  • Exploiting the Connection
  • Contributions
  • Thesis Organisation
  
  
• Non-rigid Registration
  • Motivation
  • Overview
  • Representing Image Deformation
    • Objective Function
    • Optimisation
    • Pairwise versus Groupwise Registration
  • Assessment of NRR
    • Recovery of Deformation Fields
    • Overlap-Based Assessment
    • Motivation for Ground-Truth-Free Assessment
  • Summary
  
  
• Active Appearance Models
  • Motivation
  • Shape Models
    • Representing Shapes
    • Aligning Shapes
    • Obtaining Modes of Variance
    • The Correspondence Problem
  • Appearance Models
    • Modelling Shape
    • Modelling Intensity
    • Combined Models
  • Active Appearance Models
    • Learning to Fit the Model
    • Model Fitting
  • Summary
  
  
• MDL Shape Models
  • Overview
  • Minimum Description Length
  • Measuring Model Quality
  • Methods of Manipulating Correspondence
  • Optimisation
  • Summary
  
  
• Model-based Registration
  • Overview
  • Applying the MDL Approach to Images
    • Image Data
    • Manipulating the Correspondence
    • Objective Function
    • The Registration Algorithm
    • Experimental Framework
  • Experiments - Registering 1-D Images
    • Basic NRR Results
    • Studying NRR Methods
    • Weighting
    • Optimiser Tolerance
    • Subsets
  • From Registration to Models
    • Building Appearance Models in 1-D
  • 2-D Model Construction
    • Registration Algorithm
    • Exploratory Experiments in 2-D NRR
    • Using the Results of NRR
    • The Model Construction Algorithm
  • Summary and Discussion
  
  
• Assessment of Non-Rigid Registration
  • Overview
  • Measuring Model Quality
    • Specificity
    • Generalisation
    • Caveat
  • Image distance measures
  • Summary and Discussion
  
  
• Validation Methodology and Experiments
  • Method
    • Image Data
    • Perturbing the Initial Registration
    • Validation using Warped Images
    • Sensitivity
  • Results
    • Sensitivity
    • Effect of Noise
  • Discussion
  
  
• Application to Evaluation of NRR
  • Comparing Registration Algorithms
    • Pairwise Registration to a Reference
    • Groupwise Congealing Algorithm
    • Groupwise MDL Algorithm
  • Results of Comparison
  • Exploration of 3-D NRR
  
  
• Future Work
  • Automatic Model Building
  • Assessment of NRR
    • Specificity and Generalisation
    • Investigating Robustness
    • Extension to 3-D
  
  
• Summary and Conclusions
  • Discussion
  • Conclusions
  
  
• Bibliography