Analytic ambient occlusion using exact from-polygon visibility

Abstract

This paper presents a new method to compute exact from-polygon visibility, as well as a possible application to the calculation of high quality ambient occlusion. The starting point of this work is a theoretical framework which allows to group lines together according to the geometry they intersect. By applying this study in the context of from-polygon visibility, we derive an analytical definition which permits us to group the view rays according to the first geometry they encounter. Our new algorithm encodes lazily the visibility from a polygon. Contrary to previous works on frompolygon visibility, our approach is very robust. For each point in the scene, the algorithm efficiently calculates the exact fragments of visible geometry. By combining this information with a analytical definition of ambient occlusion, we achieve results that are not sensitive to noise or other visual imperfections, contrary to ambient occlusion methods which are based either on sampling or visibility approximations.

Visibility Silhouettes for Semi‐Analytic Spherical Integration

“…Many additional works leverage similar silhouette and depth complexity sampling in order to efficiently compute soft shadows [FBP08, AAM03, ED07], but none consider semi‐analytic spherical integration. More recent works leverage lazy visibility evaluation and analytic from—polygon form—factor computation to compute soft shadows from geometric area lights and AO effects [MAAG12, AMGA12]. We are motivated by works on semi‐analytic visibility [NN85, SR00, HMN05], which compromise between stochastic sampling approaches that are sensitive to noise, and more complex fully analytic solutions.…”