Discussion of results:

This project turned out surprisingly well despite some minor differences with the demo program. Most of the implementation steps didn’t have much interpretation with them. They either worked or didn’t.

As I have noticed with the other projects, there were always some differences between the mathematic results from my version of the program and the demo program. In particular, the lighting direction vectors for my program differed a bit from those generated by the sample program. More significantly, my z-vector values always turned out to be the negative of the ones generated by the demo program. This resulted in the RGB Normals mapping having a greenish-red tint instead of a bluish-purple tint. Not surprisingly the direction of the Needle Map shows opposite directions as well. However, this is compensated for in later steps as the Albedo Map and Reconstructed Surface Maps both look similar between the demonstration program and my version of it.

I noticed some slight discrepancies between my albedo images and the demo program’s. However, I chalked these up to the aforementioned mathematical differences.

The only other difficulty I had with this project involved the least-squares fit when I solved for the depths. This was mostly due to the unexplained relation between the zind data and the way the M-matrix and V-vector are constructed. After puzzling this out and consulting with the TA, I was able to get past this problem.

After that, it was simply a matter of running the program and collecting the screenshots.