Denoising

denoise

Purpose

This utility allows you to denoise images, volumes and selfiles with images and volumes. Several kinds of denoising techniques are available, most of them based on wavelets:

• Remove scale all coefficients belonging to a given scale are removed. The finer scale is marked as scale 0 and coarser scales have greater numbers
• Soft thresholding the histogram of the wavelet coefficients is computed, and the value accounting for the[threshold]% of the coefficients is marked (th_DWT). All values, belowth_DWT are removed, and all values aboveth_DWT are substracted this quantity.
• Adaptive soft thresholding The modification proposed byChang, Yu, Betterli. in IEEE Int. Conf. Image Processing is implemented
• Central those coefficients corresponding to a region centered in the middle of the image within a radius r are kept.
• Difussion a difussion process (not wavelet based) is performed. SeeTeboul, et al. IEEE-Trans. on Image Proc. Vol. 7, 387-397

Usage

$ denoise -i ...

Parameters

• __OR__ Denoise a single file

• __OR__ If no output file is given, then the input one is rewritten

• `` Denoise a bunch of files

• __OR__ If no output extension neither root are given, then the input ones are rewritten

• `` Do not produce on screen information

• -denoising [strremove_scale] = Denoising method:

  • ``
  • ``
  • ``
  • ``
  • ``

• -type [strDAUB_12] = If denoising method uses wavelets, wavelet type. Valid types are:

  • ``
  • ``
  • ``

• -scale [s0] = Scale to remove

• -th [th50] of the wavelet coefficients are removed

• -R [r-1] = Radius to keep.r-1= is the default and it stands for half the radius

• ] Diffusion weights:

  • `` data matching (=0)
  • `` 1st derivative smooth (=50)
  • `` edge strength (=50)
  • `` edge smoothness (=0.02)

• `` By default, the difussion process saves the surface image, however, using this option you may see the edge image

• -outer [it10] = Number of outer iterations in the difussion process

• -inner [it1] = Number of inner iterations in the difussion process

• -refinement [it1] = Number of refinement iterations in the difussion process

Examples and notes

Given the sample micrograph shown:

/small_micrograph.gif

we remove scales 0 and 1

$ denoise -i small_micrograph.xmp -o small_micrograph_remove_0.xmp -denoising remove_scale \ 
-scale 0 -type DAUB20
$ denoise -i small_micrograph_remove_0.xmp -o small_micrograph_remove_1.xmp -denoising remove_scale \ 
-scale 1 -type DAUB20

obtaining:

/small_micrograph_remove_0.gif

/small_micrograph_remove_1.gif

If soft thresholding is applied

$ denoise -i small_micrograph.xmp -o small_micrograph_soft.xmp -denoising soft_thresholding -threshold 50

we get:

/small_micrograph_soft.gif

If adaptative soft thresholding

$ denoise -i small_micrograph.xmp -o small_micrograph_adaptive_soft.xmp -denoising adaptive_soft

we get:

/small_micrograph_adaptive_soft.gif

If the central part is kept

$ denoise -i small_micrograph.xmp -o small_micrograph_central.xmp -denoising central

/small_micrograph_central.gif

And applying Shah difussion

$ denoise -i small_micrograph.xmp -o small_micrograph_diff.xmp -denoising difussion
$ denoise -i small_micrograph.xmp -o small_micrograph_edge.xmp -denoising difussion -only_eºdge

/small_micrograph_diff.gif

/small_micrograph_edge.gif

--Main.AlfredoSolano - 17 Jan 2007