The main approach and algorithm are summerized and descibed in Note.pdf inside Note folder.
The sample output results are in wiki page.
Main algorithm is from Tai-Pang Wu and Chi-Keung Tang, Dense Photometric Stereo Using a Mirror Sphere and Graph Cut, IEEE Computer Society Conference on Computer Vision and Pattern Recognition (2005) pp 140-147.
Graph Cut Fold-Fulkerson maxflow algorithm toolbox downloaded from http://vision.ucla.edu/~brian/gcmex.html
Shape from shapelet reconstruction is from Peter Kovesi, Shapelets Correlated with Surface Normals Produce Surfaces. IEEE International Conference on Computer Vision. (2005) pp 994-1001. Toolbox downloaded from http://www.csse.uwa.edu.au/~pk/Research/MatlabFns/
================
Get started: Inside preProcessData.m, change the directory of input data, set the image size appropriately, and copy the light source direction file into code folder.
Compile the mex file.
Load lightvec.txt
surfaceNormal = initialNormal(VarName1, VarName2, VarName3); This will take around 1-2 minutes.
refineNormal = graphCutSurfaceNormal(surfaceNormal); This will take quite a long time.
Use lightImg = diffuseLight(surfaceNormal); or lightImg = diffuseLight(refineNormal); to check N-L image result.
Use recsurf = shapeFromShapelets(surfaceNormal); or recsurf = shapeFromShapelets(refineNormal); to check 3D reconstruction result.