Photometric Stereo Artifacts

CSE455 Project 3

By: Huiqi Wang

  • Cat Images
  • Horse Images
  • Buddha Image

  • Discussion

    In general, all the main parts for my project worked well.
    At the beginnning,  it was hard to figure out what M and b as well as M and v stood for from the project specification. It was solved after I found out they were for two different functions--compute normal and compute depth.
    The reconstructed surface has a little problem with the left edges (as shown in the third views of Buddha reconstructed surface and also the other two image sets). It could be my problem of the algorithm to compute M matrix and V vector, or it was the defect of this application which I'm not sure of. But I think if we can use images shoot from more than one angle (more light angles), we can see better reconstructed 3D images.

    Cat Images

    original
    RGB-encoded normal needle map
    albedo map
    reconstructed surface 1 without albedos reconstructed surface 2 without albedos
    reconstructed surface 1 with albedos reconstructed surface 2 without albedos

    Horse Images

    original
    RGB-encoded normal needle map
    albedo map
    reconstructed surface 1 without albedos reconstructed surface 2 without albedos
    reconstructed surface 1 with albedos reconstructed surface 2 without albedos

    Buddha Images

    original
    RGB-encoded normal needle map
    albedo map
    reconstructed surface 1 without albedo
    reconstructed surface 2 without albedo
    reconstructed surface 3 without albedo
    reconstructed surface 1 with albedo
    reconstructed surface 2 with albedo
    reconstructed surface 3 with albedo