GSoc 2015 brought structured light to OpenCV and I would love to extend the module this year. Among all the available algorithms, phase-shifting profilometry (PSP) is one of the most popular and I believe OpenCV would benefit from its implementation. Unfortunately, PSP does not perform well on dynamic scenes since it belongs to time-multiplexing methods. Indeed, successive patterns are not deformed in the same way which leads to reconstruction errors. To reduce reconstruction errors due to motion, scientists had the idea to combine PSP with Fourier transform profilometry (FTP), a one-shot method that uses the same patterns. My goal for GSoC 2016 is to add PSP, FTP and Fourier assisted phase-shifting profilometry (FAPS) to OpenCV. Spatial phase unwrapping, which is essential to PSP, FTP and FAPS, will be coded as an independent module due to its broad number of applications. Finally, as a proof of concept of scene understanding from structured light, an automatic piecewise-planar surfaces detection will be tested on the results provided by Fourier transform profilometry.



Ambroise Moreau


  • Delia Passalacqua