Level 2 Help for MARSDISPCOMPARE

input images.
first: left to right disparity image
second:  right  to left disparity image

Used to test for outliers.
All four neighbors used in the bilinear interpolation must
be within this distance (in pixels) of each  other.

Threshold on the acceptable distance (in pixels) after projecting 
back into the L->R disparity image.

output disparity mask wich is a single banded byte file.  The pixels
dn values can be either 0 or 255, no other values are allowed.
Pixels with dn value of 0 are the ones that passed consistency check.
Pixels with dn value of 255 either failed consistency check or had
no correlation value to begin with.

(Optional)
input R-> L disparity quality image

(Optional)
input L-> R disparity quality image

(Optional)
Two thresholds that the values in the quality images must satisfy.
first  - left disparity image
second - right disparity image
(If only one quality image is provided, only 1 threshold is required)

The default threshold parameters (O_THRESH and MINDIST) assumes a scaling ratio
of about 1 between the images. That is L and R have about the same scale. In 
situation where the scale is different, say L is high resolution compared to
R, then these thresholds can be adjusted to account for it. For instance if 
the two images have a scale ratio of 10, the disparity map accuracy will be 
driven by the low resolution image. When comparing the disparity for the high
resolution image, we cannot expect an accuracy of mindist of 2 pixels (default)
which would require a correlation subpixel precision of 2/10 of a pixel.
If SCALING is activated, the average scale factor contained in the label of the
disparities is accounted for and adjust O_THRESH and MINDIST accordingly (
simple multiplication by the average scaling factor)