PROJECT TITLE :

Binocular Light-Field Imaging Theory and Occlusion-Robust Depth Perception Application

ABSTRACT:

The depth information can be reconstructed using binocular stereo vision (SV), although it is susceptible to scenes with heavy occlusions. Light-field (LF) imaging, a new computational photography technique, offers a fresh solution to the problem of passive depth perception by simultaneously capturing several angle perspectives in one exposure. The binocular-LF imaging system is the focus of this research, which examines SV and LF imaging. Based on the geometrical optics theory, an imaging theory can be derived by modelling the imaging process and examining disparity properties. By utilising multi-baseline stereo matching cues and defocus cues, an accurate occlusion-resistant depth estimate algorithm is proposed. Binocular SV and LF imaging-induced occlusions are discovered and dealt with to remove matching ambiguities and outliers.. A database of real-world scenes is then built to evaluate the system's accuracy and resilience, and the results are published online. High-quality depth maps with smooth surfaces and exact geometric shapes may be recovered using the proposed approach, which concurrently addresses both SV/LF imaging disadvantages.