An increasing number of 3D VR/AR applications rely on the estimation of a depth map, i.e. each pixel of the RGB image is associated to an accurately sensed depth value. In previous years, we have developed and characterized a high-resolution structured light depth sensing device.
This Master thesis aims at further developing a stand-alone prototype, automating the full process, including the calibration, triggering and image post-processing, ported to an embedded device. Moreover, the algorithm should be parallelized and CUDA accelerated to reach real-time video performance.
Student profile :
The student should combine software programming and hardware DIY skills, be familiar with Linux and show interest in 3D acquisition.
Promotor & contact :
Gauthier Lafruit : gauthier.lafruit@ulb.ac.be