Deconvolution

Class: NodeImageProjectedLandweberDeconvolution

Node Icon

Deconvolve an image using the Landweber deconvolution algorithm as defined in Bertero M and Boccacci P, “Introduction to Inverse Problems in Imaging”, 1998. The algorithm assumes that the input image has been formed by a linear shift-invariant system with a known kernel and is best suited for images that have zero-mean Gaussian white noise.

At each iteration, negative pixels in the intermediate result are projected (set) to zero. This is useful if the solution is assumed to always be non-negative, which is the case when dealing with images formed by counting photons, for example.

Inputs

Image

Input image.

Type: Image4DFloat, Required, Single

Kernel

Image to use as kernel.

Type: Image4DFloat, Required, Single

Outputs

Output

Resulting image.

Type: Image4DFloat

Settings

Boundary Condition Selection

Sets the method to use around the boundaries.

Values: ZeroPad, ZeroFluxNeumannPad, PeriodicPad

Output Region Mode Selection

Sets the output region mode.

Values: Same, Valid

Alpha Number

Relaxation factor.

Normalize Boolean

Normalize the output image by the sum of the kernel components.

Iterations Integer

Set the number of iterations.

References

  1. “The Insight Segmentation and Registration Toolkit” www.itk.org