This is documentation for Mathematica 8, which was
based on an earlier version of the Wolfram Language.
View current documentation (Version 11.2)


iteratively reduces noise while preserving edges in image.
assumes a regularization parameter value param.
  • TotalVariationFilter works with arbitrary grayscale and multichannel images, as well as real arrays of any rank.
  • In TotalVariationFilter, the value of regularization parameter param is typically in the range 0 to 1.
  • The type of noise to be removed can be specified by setting a Method option. Possible settings include , , and .
  • By default, Gaussian noise is assumed. For removing salt-and-pepper noise or impulse noise, the Laplacian noise model gives best results. Poisson noise typically describes low-light images as well as various types of multiplicative noise.
Remove Poisson noise from a low-light photo:
Remove Poisson noise from a low-light photo:
Click for copyable input
Denoise a grayscale image:
Use total variation filtering to denoise one-dimensional data:
TotalVariationFilter works with numerical sparse arrays:
Remove salt-and-pepper noise:
New in 8