I am working on my own DXT compressor, and at the heart of it I need to be able to perform a linear fit on a set of 4 dimensional points, preferably with perpendicular offsets. Right now I am taking the channel with the biggest delta as the independant variable and using each of the other channels as a dependant variable in a series of 3 linear fits using the formula indicated on this page: Least Squares Fitting Perpendicular Offsets What I am not sure about and what I would like to find out is exactly what it means if the denominator in step 15 equals 0. And if anyone knows of a description of a 4 dimension linear fit anywhere, I would like a link. edit: Can't seem to get the URL clickable. [edited by grhodes_at_work to make link clickable] [Edited by - grhodes_at_work on August 30, 2004 1:02:34 PM]

I misread the DXT5 format documentation. Looks like alpha is treated seperately (which makes a lot more sense) so I only need 3 dimensional linear fits.

I also decided to use vertical offsets instead of perpendicular offsets as the perpendicular offsets seemed to involve a lot more work than they were worth.

You have it working now?

Looks like it.