Sampling Algorithms

The rendering of effects like motion blur, jittering, area light sources, or depth-of-field of a camera lens, is based on multiple sampling with varying sample locations in time, 2D or 3D space. mental ray offers a proprietary implementation of the Quasi-Monte Carlo method to achieve these variations in optimal ways. The sample locations chosen in time, 2D, and 3D space are deterministically chosen on fixed points that ensure optimal coverage of the sample space. The algorithm is similar to fixed-raster algorithms, but avoids the regular lattice appearance of such algorithms. The resulting images are exactly identical if the scene is re-rendered with the same options due to the deterministic nature of the algorithm.

Quasi-Monte Carlo methods can be succinctly described as strictly deterministic sampling methods. Determinism enters in two ways, namely, by working with deterministic points rather than random samples and by the availability of deterministic error bounds (Niederreiter92).

Unified Sampling 3.9

The unified sampling mechanism provides a simplified and consistent user interface to the different primary sampling settings of the various rendering algorithms, like adaptive sampling in ray tracing, fixed sampling for the rasterizer, as well as sampling in progressive rendering and iray. At the same time, it enables the use of enhanced sampling strategies in mental ray to avoid typical artifacts like visible regular patterns. In this mode, sampling is performed both in space and time at once, making it much easier to tune a rendering with motion blur for desired quality. Once this interface is enabled, a set of few controls becomes available to adjust the primary sampling strategy. In particular, a single convenient direct dial for sampling quality is provided that controls adaptivity to perceptual detail in the image when supported by the rendering algorithm currently used. The original controls and settings remain active until unified sampling has been enabled in the rendering options of the scene or on the command line of a standalone mental ray.

Copyright © 1986, 2013 NVIDIA Corporation