Module: Adjusted subvoxel ()

Description:

This module adjusts the given point set or surface to the sub-voxel accuracy by moving points along the gradient direction and snapping them to the local gradient maxima. The points are moved along a path defined by the gradient of the input volume data. The gradient is calculated analytically from the spline equations fitted on the volue data. The calculation of the path is done using numerical methods: Euler's or Runge-Kutta. If there are several gradient maxima within the search distance, gradient magnitude and distance from the starting point are used to make a decision.

Figure 1: Surface adjustment by integrating in gradient vector field

From an initial guess based on threshold or optionally an arbitrary surface definition, the boundary is adjusted with sub-pixel precision to the object edge, i.e. fitting the maximum gradient locally. This method is generally more precise, faster and less memory consuming than ���adaptive subvoxel��� and traditional methods.

Connections:

Input Data [required]
Input point set or surface to be adjusted. This port will accept HxVertexSet or HxSurface.

Volume Data [required]
The 3D volume containing gray values. The gradient will be computed directly from this volume.

Ports:

Precision

Surface determination precision expressed as a fraction of voxel size. Note this is the precision relative to image. The actual physical measure accuracy may depend on many acquisition factors including voxel size calibration.

Adjustment Distance

The distance within which the initial determined surface can be adjusted. This value can be modified depending on image quality, for instance increased in case of strong beam hardening artefact, or reduced for good quality images in particular with thin wall structures.

Adjustment Threshold

The initial surface determination is unchanged where the local contrast (gradient) is below this threshold, for instance in areas where image quality can be deteriorated and strong artefact occur. This can reduce overall computation time and potentially improve surface determination to some extent. However you may want improving the image acquisition instead, if you need accurate measurements in such areas.