Core Optical, Inc.
AdaptiveDWMTMfilter ClassIndustrial Strength, Scientific Grade
Encapsulates the functionality associated with an Adaptive Double Window Modified Trimmed Mean (Adaptive-DWMTM) filtering operation.
Inheritance Hierarchy

OnlineSystem Object
  PrecisionImage.SpatialFiltering SpatialFilteringOperator
    PrecisionImage.SpatialFiltering AdaptiveDWMTMfilter

Namespace: PrecisionImage.SpatialFiltering
Assembly: PrecisionImage (in PrecisionImage.dll) Version: (

public class AdaptiveDWMTMfilter : SpatialFilteringOperator

The Adaptive-DWMTM is a hybrid median-mean filter that first computes the median value within the median kernel neighborhood and uses this as a seed value - in conjunction with an estimation of the standard deviation of the noise - to determine a range for the mean computation. The mean for all kernel values within this range is then computed and returned as the filtered result at the given location in the source data.

The kernel radius for the mean kernel must be greater than or equal to the radius of the median kernel. The default value for the median kernel radius is 1 (3x3 kernel) and the default value for the mean kernel radius is 2 (5x5).

Two additional properties of the filter affect the result: the K property and the NormalizedStdDev. The NormalizedStdDev is an estimation of the noise magnitude in the image and is a value ranging from 0.0 - 1.0. This corresponds to a percentage of the range of values within the source data object. It can be thought of as the percent noise standard deviation. Larger values are used for greater noise in the source data, whereas smaller values are used for smaller noise levels and cause less smoothing of the data. The default value for NormalizedStdDev is 0.15f. The K parameter controls the scaling of the standard deviation. A value of 0 for K forces the filter to function as a classic median filter, whereas larger values of K introduce greater blur characteristics to the result. The default value for the K property is 1.25.

See Also