Overall project went Ok, Here are the sample pictures

Buddha:

Normals and Albedo

Surface Views

Surface Views with Albedo

Rock:

Normals and Albedo

Surface Views

Surface Views with Albedo

Owl:

Normals and Albedo

Surface Views

Surface Views with Albedo

What worked and What did not:

This project was quite simple, though I had some problems in matching my light directions with the solution light directions. Also in calculating the initally I was populating all possible rows and getting Ok surfaces but not same as solution. Then I figured out that, if one of the elements of contraints is out of bounds not to populate the entire row.

Expriments and Changes:

1. Code: In the Depths.cpp, the functions "makeMmatrix" and "makeVvector" can be combined into one function "makeMmatrixAndVvector" to avoid running the loop twice. I have put that code in "Depths.cpp" under "Experiment" directory.
2. You can visualise the surface by loading just the normals and then computing the depth. Since there is no mask, the output looks something like this. Interesting.

   

3. Tried comuting the Light Direction and Normals based on the overall (all the components of RGB image) intesity rather than just based on single channel. Here is the result for comparison.

numrecompute = 174

delta_new = 1218.406594

z range (-40.9851, 40.3287)060 < /P >

numrecompute = 174

delta_new = 1195.056132

 z range (-39.4988, 37.9564)206

• I also treid some scehemes for weigting the normals with pixel intesity and smoothing out the normals themselves, I did get much further into it to produce significantr results.