Wolfram Language & System 10.4 (2016)|Legacy Documentation
applies a nonlocal means filter to image by comparing a range r neighborhood to its nearby neighborhoods.
assumes an additive noise power value ns for comparing neighborhoods.
compares neighborhoods in a range w window.
- NonlocalMeansFilter is a nonlinear filter used for edge-preserving smoothing.
- In NonlocalMeansFilter[image,r,ns,w], normalized similarities between the range r neighborhoods are used as weights to replace each pixel by a weighted average of all the pixels in the range w window. The similarity between two neighborhoods is computed by Exp[-Max[0,SquaredEuclideanDistance[n1,n2]/len-2ns]/(k^2 ns)], where len is the total number of channel values in the neighborhood and k is .
- NonlocalMeansFilter[image,r] is equivalent to NonlocalMeansFilter[image,r,Automatic,5r].
- NonlocalMeansFilter returns an image of the same type as image.
- NonlocalMeansFilter takes a Padding option. The default setting is Padding->"Fixed".