Start » Filter Reference » Image » Image Thresholding » ThresholdImage_Dynamic

ThresholdImage_Dynamic


Module: FoundationLite

Thresholds an image relatively to some value calculated in a local rectangular neighbourhood.

Applications

Image binarization when the illumination is uneven.
Name Type Range Description
Input value
inImage Image Input image
Input value
inRoi Region* Region in which pixels are written
Input value
inSourceRoi Region* Region from which pixels are read
Input value
inReferenceMethod ThresholdDynamicReferenceMethod Specifies how the local threshold value will be calculated (see SmoothImage filter family)
Input value
inKernel KernelShape Kernel shape.
Input value
inRadiusX Integer 0 - 65535 Horizontal radius of local neighbourhood.
Input value
inRadiusY Integer* 0 - 65535 Vertical radius of local neighbourhood (Auto = inRadiusX)
Input value
inMinRelativeValue Real* Minimum relative value of a pixel that is considered foreground (Auto = -INF)
Input value
inMaxRelativeValue Real* Maximum relative value of a pixel that is considered foreground (Auto = +INF)
Input value
inFuzziness Real 0.0 - A tolerance for inMin/MaxRelativeValue that results in intermediate output values. Clamped on half of pixel max value (e.g. max fuzziness for uint8 image is 128).
Output value
outMonoImage Image
Diagnostic input
diagBaseImage Image Diagnostic threshold values.

Description

The operation transforms each pixel value to the maximum or minimum level thus creating binary image. The result of the transformation depends on the relative pixel intensity:

  • Pixel values that are brighter than local average of the pixel neighbourhood by at least inMinRelativeValue and at most inMaxRelativeValue are transformed to the maximum level.
  • Other pixel values are transformed to the minimum level.

If any of the parameters inMinRelativeValue, inMaxRelativeValue is not set, it is assumed to be, accordingly, -infinity or infinity.

Pixel neighbourhood used to compute the local average is a rectangle of dimensions (2 \cdot \text{\textbf{inRadiusX} }+1) \times (2 \cdot \text{\textbf{inRadiusY} }+1) centered at the pixel being processed.

Parameter inFuzziness (set to 0 by default) allows to perform fuzzy thresholding which linearly interpolates those pixel values that differ by at most inFuzziness from the border intensities; thus creating smooth transition between minimum and maximum values in the resulting image.

In the multichannel images the operation uses an average of channel values in each pixel, thus the resulting image is always monochromatic.

Hints

  • Define inMinRelativeValue to obtain white pixels for objects which are brighter than the neighborhood.
  • Define inMaxRelativeValue to obtain white pixels for objects which are darker than the neighborhood.
  • Increase inRadiusX (and optionally inRadiusY) to define a bigger neighborhood.
  • Use inFuzziness to add some smooth transitions between black and white pixels in the result.

Examples

ThresholdImage_Dynamic performed on the sample image with inMinRelativeValue = 5.0, inMaxRelativeValue = auto, inFuzziness = 0.0.

ThresholdImage_Dynamic performed on the sample image with inMinRelativeValue = 5.0, inMaxRelativeValue = auto, inFuzziness = 10.0.

Hardware Acceleration

This operation is optimized for SSE2 technology for pixels of type: 1xUINT8.

Errors

This filter can throw an exception to report error. Read how to deal with errors in Error Handling.

List of possible exceptions:

Error type Description
DomainError Incorrect fuzziness value in ThresholdImage_Dynamic.
DomainError Roi exceeds image dimensions in ThresholdImage_Dynamic.
DomainError Source roi exceeds image dimensions in ThresholdImage_Dynamic.

Complexity Level

This filter is available on Basic Complexity Level.

Filter Group

This filter is member of ThresholdImage filter group.

See Also

  • ThresholdToRegion_Dynamic – Thresholds an image relatively to the average pixel value in a local rectangular neighborhood.