I think it is a matter of deleting the depth factor. The default of one hundred is too much. If I reduce it to thirty (about what it was for stereograms that were once printed every week in comics, here), then simplify the code according to a constant depth, then I will probably find the bug in it, too. Then I can either put the depth factor back in or let someone else do a good job of it. What is daunting is the sophistication of the code, now, since I see that there are graduations between most heights. Maybe I should see if the code can make steps between all heights. So, in sum, it is like: 1. Delete depth factor code. 2. Delete edge-blending code. 3. Choose a minimum and maximum depth that make for 256 heights that blend when only one difference in colour is. 4. Consider making it scale and interpolate, again. 5. Start with what I actually promised to do, elsewhere. :) _______ http://ecn.ab.ca/~brewhaha/img/STEREOD.GIF Stereogram diagnostic--256 numbered (and coloured) shades of gray that should look like a mountain road in RDS. It would be trivial to arrange it in a square spiral, which might be more informative, yet. Fun: Try reducing the colours in a 24bit grayscale to the map in that diagnostic. The hue will always be wrong, so you will get a lot of noise to diffuse.