Dx gauss. Horizontal first differences followed by horizontal gaussian
smoothing with distribution width = sigma1.
Dx. Horizontal first differences.
DoG. Difference of gaussians filtering with DoG = k1 x G_sigma 1
+ k2 G_sigma 2.
grad Summed squares of vertical and horizontal derivatives (needed for complete 3D data extraction).
None. No preprocessing.
The DoG parameters are used by preproc.
The edge detection is performed automatically and the number of edge pixels within the block is used (in the absense of an input disparity image) to determine whether a block contains enough information to attempt correlation.