2010 Summer Project Week Best Regularization Term for Demons Registration Algorithm
- MGH: Rui Li, Greg Sharp
- Show comparison of the demons algorithm performance with different regularization terms.
Our plan is to find a way to convert the current research code into an itk module that can be loaded into slicer.
The research code is a mixture of MATLAB and C code. Thanks to the discussion with Luis, we have successfully implemented itkCurvatureDemonsFilter.* and itkCurvatureDemonsFunction.* for a variant of the Demons algorithm with the curvature regularization term.
Our next steps after this project week are: (1). implement locally adaptive curvature based Demons algorithm. (2). implement Demons algorithm that uses L1-regularization term, this step will take more time and effort. itkPDEDeformableRegistrationFilter.* and itkPDEDeformableRegistrationFunction.* need to be rewritten.
All the implementations will be included into plastimatch.
- Demons algorithms for fluid and curvature registration. Cahill, N.D., Noble, J.A., Hawkes, D.J. ISBI 2009.
- A Demons Algorithm for Image Registration with Locally Adaptive Regularization. Cahill, N.D., Noble, J.A., Hawkes, D.J.. MICCAI 2009.
- High Accuracy Optical Flow Estimation Based on a Theory for Warping. Brox, T. et al., ECCV 2004