Time-resolved inverse radiosity: simultaneous estimation of shape and reflectance

Area of Science:

• Computer Vision
• Computer Graphics
• Computational Imaging

Background:

• Inverse rendering aims to recover scene properties like shape and reflectance from images.
• Inter-reflections pose a significant challenge, complicating accurate material and geometry estimation.
• Conventional methods struggle with complex lighting and inter-surface light bounces.

Purpose of the Study:

• To introduce a novel time-resolved inverse radiosity method for joint shape and reflectance estimation.
• To address the limitations of existing techniques in handling inter-reflections.
• To enable robust photometric analysis in complex lighting environments.

Main Methods:

• Incorporation of transient light propagation into inverse radiosity.
• Utilizing a single-photon avalanche diode (SPAD) for time-resolved imaging.
• Development of a temporal consistency measure for joint optimization without prior shape knowledge.

Main Results:

• Successful simultaneous estimation of shape and reflectance.
• Effective mitigation of inter-reflection artifacts in real-world scenes.
• Demonstrated robustness in challenging scenarios with complex lighting.

Conclusions:

• Time-resolved inverse radiosity offers a powerful approach for high-fidelity inverse rendering.
• The method enables accurate reconstruction even with significant inter-reflections.
• Opens new avenues for analyzing and recreating complex visual environments.