Indirect Illumination

In order to fully understand the illumination settings for Luxology rendering it is critical to first understand the technical process of indirect illumination. While it is intuitive to imagine light coming from a 3D light or a luminous surface and traveling to reach the surfaces and then bouncing around a room, the actual process of rendering with indirect illumination is quite the opposite. As the surface is evaluated, rays are cast outward from the surface randomly and evaluated when they strike other surfaces in the scene. The sum of those evaluations is what contributes to the color and brightness of the original surface.

To get a more precise idea of how indirect illumination is estimated at a point on a surface, imagine the top half of a transparent globe resting on the surface so that the point’s surface normal is poking through the North Pole. Rays are fired from the surface point through random points within each “cell” formed by the latitude and longitude grid lines, with one ray per cell. These rays go out and hit either other surfaces or the distant environment, and the average color that they see is the indirect irradiance estimate. (Irradiance is the technical term used for incoming light.)

Now imagine we need to shade a flat surface, and the environment image is all black except for one concentrated bright region. Each shading point on the surface sends rays as described above. For some points, maybe two of their rays hit the bright region, while for other points only one ray hits it and the rest of the rays see black. With some points getting twice the irradiance of others, you can predict that the surface will look quite splotchy if Irradiance Caching is on (or grainy if it is off). If, however, we subdivide the transparent hemispheres more finely (that is, use more rays), the number of hits and misses will be much more consistent between neighboring surface points, smoothing out the shading. While Luxology’s indirect illumination is based on this hemispheric sampling, there are two very different approaches to the use of these samples, Irradiance Caching and Monte Carlo.

Irradiance Caching Method

The default method of sampling uses a technique called Irradiance Caching. The concept behind this technique is that by leveraging a smaller number of more accurate samples and blending between them, you can achieve an image of perceived quality in a shorter amount of time than sampling every pixel with lesser quality, which often results in a "grainy" image. When Irradiance Caching is disabled, Luxology falls back to generating a hemispherical shading sample for every pixel in the image. Consider the number of rays you use, since this number is multiplied by the millions of pixels in your image. With Irradiance Caching active, Luxology intelligently samples the scene at strategic locations and then interpolates between them for a smoother overall final frame.

When using Irradiance Caching, any variance is spread across from sample to sample, which yields splotches. Irradiance Caching provides several approaches to reducing artifacts, which include increasing the number of rays, adding Super Sampling, and increasing the number of samples required to create a blend (Interpolation Values).

Monte Carlo Method

The Monte Carlo method uses a lower quality (fewer rays) sample at every single pixel whereas the Irradiance Caching method uses fewer much higher quality (more rays) samples and blends them together. When the samples are not accurate enough in Monte Carlo, there is significant variance from one pixel to the next and the image appears as "grainy." One remedy is to increase the number of rays per pixel, but this causes render time to increase dramatically. The Progressive Refinement Render mode uses the Monte Carlo method. It automatically fires additional rays per pixel such that the rendered image is continuously improved until you stop the process or a predefined amount of time has passed.