Axial-Cones: Modeling Spherical Catadioptric Cameras for Wide-Angle Light Field Rendering

摘要

Catadioptric imaging systems are commonly used for wide-anglernimaging, but lead to multi-perspective images which do not allowrnalgorithms designed for perspective cameras to be used. Efficientrnuse of such systems requires accurate geometric ray modeling asrnwell as fast algorithms. We present accurate geometric modelingrnof the multi-perspective photo captured with a spherical catadioptricrnimaging system using axial-cone cameras: multiple perspectiverncameras lying on an axis each with a different viewpoint andrna different cone of rays. This modeling avoids geometric approximationsrnand allows several algorithms developed for perspectiverncameras to be applied to multi-perspective catadioptric cameras.rnWe demonstrate axial-cone modeling in the context of renderingrnwide-angle light fields, captured using a spherical mirror array. Wernpresent several applications such as spherical distortion correction,rndigital refocusing for artistic depth of field effects in wide-anglernscenes, and wide-angle dense depth estimation. Our GPU implementationrnusing axial-cone modeling achieves up to three orders ofrnmagnitude speed up over ray tracing for these applications.