Random selection of evaluation points in the reference image space.
Probabilities of selecting a given point can be weighted by the spatial gradient norm of the input image. This process insure that exactly n points are selected, at the cost of computational complexity.
Description
Figure 1. Points randomly selected with balance in {0, 0.5, 1}, with and without the inclusion of points whose probability was originally (before balancing) null.
Image in which points are selected. In order to minimize deformations of the input image, in Primatologist's framework, the atlas is registered towards the target MRI space. This should thus be the target MRI volume.
Modulatory parameter of the probabilities. If 0, probabilities are directly proportional to the norm of the gradient in the image. If 1, all voxels have the same probability of being selected. Any value between 0 and 1 pushes probabilities towards one of these extremities.
include_zero: Booléen ( optional, input )
If false, no point with a null gradient can be selected, even if the balance parameter is not 1.
Atlas Registration: Select Evaluation Points