contrib/brl/bseg/sdet/sdet_denoise_mrf_bp.h File Reference

a processor applying a mrf denoise algorithm More...

#include <vcl_vector.h>
#include <sdet/sdet_denoise_mrf_bp_params.h>
#include <sdet/sdet_mrf_bp.h>
#include <vil/vil_image_resource.h>
#include <vil/vil_pyramid_image_view.h>

Go to the source code of this file.

Classes
class	sdet_denoise_mrf_bp

---

Detailed Description

a processor applying a mrf denoise algorithm

Author:

J.L. Mundy - March 30, 2011

This algorithm selectively smooths the image based on a variance value at each pixel. The smoothing is carried out by a MRF with binary cliques all of equal weight (kappa_) The data cost is related to the variance by

D(fp) = lambda_*(fp-x)^2 ------- var The clique cost is V(fp, fq) = kappa_(fp-fq)^2

The algorithm uses belief propagation based on the paper by

Pedro F. Felzenszwalb, Daniel P. Huttenlocher, Efficient Belief Propagation for Early Vision International Journal of Computer Vision 70(1): 41-54 (2006)

The MRF message storage could be reduced by 1/2 if a checkerboard update scheme is used, but it was decided to update all sites on each iteration.

If a variance image is not set, then the data cost is

D(fp) = lambda_*(fp-x)^2

    Modifications
     <none yet>
   

Definition in file sdet_denoise_mrf_bp.h.