3D Reconstruction

Discussion

Reconstruction generally worked well for the given source images. Normal and depth calculations were able to capture the overall geometry of the objects as well as finer details. Albedos were quite accurate at distinguishing colors from the surface. The results for the rock image set seemed to turn out particularly well, given the rough and irregular surface.

Some problems with reconstruction are most apparent in the owl image set. The limited viewing angle results in distorting the owl's face at the edges of the reconstructed figure. There seemed to be a small amount of specular highlights that created artifacts in the normals as well as the albedos, seen in the ridges of the eye and pupil and the torso. Perhaps smoothing images before calculating normals would help reduce this effect for slightly specular materials.