Proposals:Registration Framework in MATITK

Registration Framework to add to MATITK

This document is intended to collect the commonly used combinations of registration components so that they can be selected to be wrapped for MATLAB using MATITK.

Image Registration

The most common registration problems are:

Linear Transformation

3D to 3D images

• Rigid Transform (usually done with Quaternion Optimizer or VersorRigid3DOptimizer)
• Similarity Transform (any optimizer except Levenberg-Marquardt)
• Affine Transform (any optimizer except Levenberg-Marquardt)

2D to 2D images

• Rigid Transform (any optimizer except Levenberg-Marquardt)
• Similarity Transform (any optimizer except Levenberg-Marquardt)
• Affine Transform (any optimizer except Levenberg-Marquardt)

Non-Linear Transformation (Deformable Registration)

2D to 2D Images

• BSpline deformable Transform
• Demons Registration
  • Non-Symmetric Forces (default one)
  • Summetric Forces
• FEM Deformable registration filter
• LevelSet deformable registration filter

3D to 3D Images

• BSpline deformable Transform
• Demons Registration
  • Non-Symmetric Forces (default one)
  • Summetric Forces
• FEM Deformable registration filter
• LevelSet deformable registration filter

Point Set Registration

• ICP examples (the patent has expired now)
  • This is a combination of Levenberg-Marquard optimizer, any transform, two point sets, and a PointSetToPointSet metric.

---

ITK: [Welcome | Site Map]