112. STEREO PANORAMA STITCH BASED ON POINT CLOUD REPROJECTION

Department: Electrical & Computer Engineering
Research Institute Affiliation: California Institute for Telecommunications and Information Technology (Calit2)
Faculty Advisor(s): Truong Nguyen | Jurgen P. Schulze

Primary Student
Name: Ji Dai
Email: jid046@ucsd.edu
Phone: 857-272-0251
Grad Year: 2020

Abstract
Monocular panorama stitch problem has been existed for a long period. Decent result has been achieve even under challenging circumstances. However, as Virtual Reality revolution is unfolding, stereo panorama stitch has generally replaced monocular one and attract more and more attention. To generate stereo panorama with a pleasant viewing experience while all pixels stay in correct geometric location is not a easy task. Besides common challenges existing in monocular panorama stitching, like parallax handling, in stereo panorama stitch, we also need to take care of depth information preservation: far away objects should still be far away and close by objects should still be close by. In this project, we proposed a novel method for stereo panorama stitching. Our method guarantees that all pixels are in the correct locations and no vertical error is introduced. Besides, viewers can pick any viewing angles view the scene, while most of existing panorama generation method doesn't provide you such option. The algorithm is also potentially capable to do panorama video generation, no temporal alignment is required comparing to using traditional morph based algorithms.

Industry Application Area(s)
Electronics/Photonics

« Back to Posters or Search Results


Contact:   researchexpo@soe.ucsd.edu   (858) 534-6068