PDF version of this document

Possible Outlines: BMVC 2005
and MICCAI 2005

R. S. Schestowitz

Research Student

Imaging Science and Biomedical Engineering

Marked in RED are bits that need to be re-generated or produced from scratch.

BMVC

Introduction

• What appearance model are
• The need to evaluate models
• The benefits of evaluation

Background

• Correspondence problem
• Appearance models: example
• Different methods of different research groups for creating appearance models and the like (e.g. SDM)

Methods

• Show model/s (the famous '3 or 5 images' view)
• Synthesis from the model (show face, brain examples)
• Introduction of the shuffle transform
• Using existing graphics files to explain about the framework

Experiments

• Show some hand annotated brains/faces. Probably include just brains in MICCAI.
• Show perturbation applied to correspondences
• Show CPS splines applied to images (points are immutable)
• Show how models change after noise has been applied to points/images
• Explain about the increasing-sigma/elevating-warp-magnitude (or number) experiments
• Questionable: Possibly include a comparison between automatically-generated models, e.g. group-wise and pair-wise
• Addendum:
  • Consider inclusion of face images 'treatment', e.g. masking and rigid alignment
  • Alternatively, explain about 104 brain volumes being aligned and sliced
  • Acknowledge the Wellcome trust for faces, German source of faces with description of the database
  • Properties such as a high number of glasses and beards within

Results

• Show 'number-of-modes' graph for increasing sigma
• Show different shuffle distances and Euclidean; discuss this
• Questionable: Show 'number-of-modes' graph for different models
• Present the way that CPS-based noise can be applied to the images
• Show the impact of the above on the model/s

Conclusions/Summary

• We have introduced a model evaluation scheme.
• The demonstrated that it can distinguish between (what is claimed to be) good models and worse models - referring to group-wise, points moving to regions of high variance, etc.

MICCAI

Introduction

• As in BMVC, emphasis on brain models; citing brain-related publications

Background

• Ditto

Methods

Differences w.r.t. BMVC:

• Different figures need to be produced
• Show brain models in more depth to make use of space
• Undecided: Distinguish between hand-annotated model and automatically-built models
• Possibly refer to TFC's work on automatic brain model construction
• Normalisation

Experiments

• Experiments with different brains models from TFC
• Experiments with sigma changes, number of modes
• CPS warps and their effects on brain models
• Possible: Comparison with Bill Crum's methods of validation

Results

• Similar to BMVC, but referring to relevant experiments and discussing each in more depth

Conclusions/Summary

• We can evaluate the quality of a brain model
• Questionable (needs rephrasing): Can use this to validate registration?
• Good models can be identified and evaluated quantitatively

この文書について...

Possible Outlines: BMVC 2005
and MICCAI 2005

この文書はLaTeX2HTML 翻訳プログラム Version 2002-2-1 (1.70)

Copyright © 1993, 1994, 1995, 1996, Nikos Drakos, Computer Based Learning Unit, University of Leeds,
Copyright © 1997, 1998, 1999, Ross Moore, Mathematics Department, Macquarie University, Sydney.

を日本語化したもの( 2002-2-1 (1.70) JA patch-2.0 版)

Copyright © 1998, 1999, Kenshi Muto, Debian Project.
Copyright © 2001, 2002, Shige TAKENO, Niigata Inst.Tech.

を用いて生成されました。

コマンド行は以下の通りでした。:
latex2html -no_subdir -split 0 -show_section_numbers /tmp/lyx_tmpdir11920dOZaoq/lyx_tmpbuf0/Outlines.tex.

翻訳は Roy Schestowitz によって 平成17年3月29日 に実行されました。