Undergrad Research Project - Using polarization cues to enhance depth information

Fall 2016

Student
Biqi Zhao
Advisor
Aswin Sankaranarayanan
Project description

Consumer 3D cameras such as Kinect have become rather popular in scientific research and product development. However, the coarse depth information they provide has become inevitably problematic. Previously proposed solutions to this problem have been combining depth map with surface normals obtained from Photometric Stereo and Shape-from-Shading. This project aims at combining surface normal information obtained from polarization with coarse depth.

Since this area is rather new, there are very few existing papers to refer to. The project will be started with studying existing papers and implementing their algorithms. Not only are these algorithms under many constraints, there also remain plenty of unresolved problems. Thus there are several things to be done after reproducing the algorithms from existing papers: (1) Identify Illumination and material constraints of these algorithms. (2) Improve algorithms by solving the existing problems. (3) Study the feasibility of this method under extreme conditions. (4) Identify real world usage of this method.

Once an ideal algorithm is developed, it can be applied to many real life usages such as scene reconstruction.

Return to project list