Adaptive Smoothing - noise removal with edges preserving.

Namespace:  AForge.Imaging.Filters
Assembly:  AForge.Imaging (in AForge.Imaging.dll) Version: (


public class AdaptiveSmoothing : BaseUsingCopyPartialFilter


The filter is aimed to perform image smoothing, but keeping sharp edges. This makes it applicable to additive noise removal and smoothing objects' interiors, but not applicable for spikes (salt and pepper noise) removal.

The next calculations are done for each pixel:

  • weights are calculate for 9 pixels - pixel itself and 8 neighbors:
    w(x, y) = exp( -1 * (Gx^2 + Gy^2) / (2 * factor^2) )
    Gx(x, y) = (I(x + 1, y) - I(x - 1, y)) / 2
    Gy(x, y) = (I(x, y + 1) - I(x, y - 1)) / 2
    , where factor is a configurable value determining smoothing's quality.
  • sum of 9 weights is calclated (weightTotal);
  • sum of 9 weighted pixel values is calculatd (total);
  • destination pixel is calculated as total / weightTotal.

Description of the filter was found in "An Edge Detection Technique Using the Facet Model and Parameterized Relaxation Labeling" by Ioannis Matalas, Student Member, IEEE, Ralph Benjamin, and Richard Kitney.

The filter accepts 8 bpp grayscale images and 24 bpp color images for processing.

Sample usage:

// create filter
AdaptiveSmoothing filter = new AdaptiveSmoothing( );
// apply the filter
filter.ApplyInPlace( image );

Initial image:

Result image:

Inheritance Hierarchy


See Also