10-04-2017, 08:26 PM
An Object Oriented Approach to Image Restoration in Matlab
[attachment=478]
Regularization by Filtering
One method to choose regularization parameters is:
Generalized Cross Validation
For example, in Tikhonov, choose to minimize
Computational Issues:
1. Choose an appropriate basis for U and V
e.g., singular vectors, Fourier vectors, ..
2. Efficient algorithm (speed and storage) to compute
Iterative Regularization
Some examples:
1. Conjugate Gradient (CG)
2. (Constrained) Steepest Descent (EM-LS)
3. Lucy-Richardson (EM-ML)
To-Do List
Stopping criteria for iterative methods.
Better understanding of preconditioning EM-LS.
Include other regularization by filtering methods.
Include other iterative regularization methods.
(e.g., Blind Image Restoration Plemmons, Jefferies)
Test codes using more real applications.