Radon

MLModule

genre	Projection
status	stable
author	Bernd Kuemmerlen
package	MeVisLab/Standard
dll	MLTransformation
definition	MLTransformation.def
keywords	Radon, transform, Sinogram, Projection, CT, xray, x-ray

Purpose

The module Radon performs a 2D Radon Transform for each slice of the input dataset.

The code is based on Peter Toft's implementation of the Radon Transform.

Details

The number of Directions and Shifts greatly influences the computation time of this module.

Often, rather small values are sufficient, especially for Directions.

Windows

Default Panel

Input Fields

input0

name: input0, type: Image

The input image can be any square 3D image (note that there is no check if the image really is square).

Output Fields

output0

name: output0, type: Image

The output is the Radon transform (Sinogram) g(theta, rho) for each slice, where theta is the projection angle and rho is the offset.

For a more detailed explanation see Peter Toft's Page.

The x-axis of the output image is the theta-axis from 0 to PI (0..180 deg), the y-axis is the rho-axis from -(rhoSteps-1)/2 to +(rhoSteps-1)/2.

Parameter Fields

Field Index

Apply: Trigger

Auto: Bool

Directions: Integer

Interpolation: Enum

Scale voxel values: Bool

Shifts: Integer

Visible Fields

Directions

name: directions, type: Integer, default: 256, deprecated name: thetaSteps

Sets the number of projections.

The sinogram will be calculated for theta=0..180 deg, so if Directions = 180, there will be 180 projections with 1 degree increment.

Shifts

name: shifts, type: Integer, default: 256, deprecated name: rhoSteps

Sets the number of rays used for each projection.

The central ray will always go through the center of the image. If you set Shifts to the X-size of the input image, all rays will cover a circle with radius X-size/2 centered on the image.

Interpolation

name: interpolation, type: Enum, default: Linear, deprecated name: interpolationMode

Defines the interpolation mode for gathering values along rays through the image.

Values:

Title	Name
Nearest Neighbour	NearestNeighbour
Linear	Linear
Linear Fill	LinearFill

Scale voxel values

name: voxelScaling, type: Bool, default: TRUE

If checked, the result of the Radon transform will be scaled with the voxel size of the input image.

This "normalizes" the result so that scans of the same object with different resolutions lead to (basically) the same Radon transform.

Should be set checked.

Apply

name: apply, type: Trigger, deprecated name: Apply

If pressed, the module computes anew.

Auto

name: autoApply, type: Bool, default: FALSE

If checked, the module computes anew on each change of a field or the input.