Adaptively layered statistical volumetric obscurance

Abstract

We accelerate volumetric obscurance, a variant of ambient occlusion, and solve undersampling artifacts, such as banding, noise or blurring, that screen-space techniques traditionally suffer from. We make use of an efficient statistical model to evaluate the occlusion factor in screen-space using a single sample. Overestimations and halos are reduced by an adaptive layering of the visible geometry. Bias at tilted surfaces is avoided by projecting and evaluating the volumetric obscurance in tangent space of each surface point. We compare our approach to several traditional screen-space volumetric obscurance techniques and show its competitive qualitative and quantitative performance. Our algorithm maps well to graphics hardware, does not require the traditional bilateral blur step of previous approaches, and avoids typical screen-space related artifacts such as temporal instability due to undersampling.iii PREFACE

Optimization of screen-space directional occlusion algorithms

“…Instead performing calculations on the geometry, it uses rendered frame's depth buffer, with an optional normal vector buffer. As a replacement for the aforementioned ray tracing, many techniques were invented by the developers, including point sampling, line sampling or horizon-based sampling [3]. Similarly, amount of occlusion in the given pixel equals a ratio of the number of "occluded" samples (i.e.…”

“…A sample box is build around processed point, instead of a sample hemisphere. The authors of [3] assume that the bigger is the difference between µ (the mean value of the screen space depth within a sample box) and d, the more occlusion occurs in each pixel. This is computed in relation to the Z T and Z B parameters and can be seen in equation (2).…”

“…It generates unwanted dark halos it such places and special steps must be taken in order to avoid this artefact. One such approach, described in [3] is to split depth buffer into separate layers, adaptively to the local geometry. Borders of the layers are created around depth discontinuities.…”

“…A key element in the StatVO technique is fast and errorfree generation of the averaged depth buffer. A data structure named Summed Area Table (SAT) was chosen for this purpose by the authors of [3]. SAT is a two-dimensional array, where each element corresponds to the sum of the elements above and to the left of him.…”

“…Then it is assigned to one of the layers, depending on the factor, whether its depth is higher or lower than the average. The idea of this algorithm is that around depth discontinuities, every pixel closer to the camera will have depth value lower than the average and vice versa, therefore they will be assigned to different layers [3]. Process is repeated recursively until four input layers are created.…”