Fast gradient projection algorithm



x = fast_gp(b, lambda, pars)

Solves min ||x – b||_F^2 + 2\lambda TV(x)

b is the constant term in the Frobenius norm.

lambda is the weight on the total variation penalty.

pars is a structure with additional parameters:

tol is the cutoff for the normed difference between successive iterates. 

If the difference falls below tol, the algorithm terminates.

max_iters is the maximum number of iterations.

bnds is a vector of the lower and upper bounds on the signal.

dom is a binary matrix that equals 1 for all pixels in the non-zero support set.

tv_type is the type of total variation penalty to use. Must be 'isotropic' or 'anisotropic'.



reference :
A. Beck and M. Teboulle, “Fast gradient-based algorithms for constrained total variation deblurring and denoising problems.” IEEE Transactions in Image Processing, vol. 18, no. 11, pp. 2419-2434, Nov. 2009.

A Quadratic Programming based Cluster Correspondence Projection Algorithm for Fast Point Matching

Nonlocal back-projection for adaptive image enlargement



There are no reviews yet.

Be the first to review “Fast gradient projection algorithm”

Your email address will not be published. Required fields are marked *