You are viewing the course site for a past offering of this course. The current offering may be found here.
Lecture 12: Monte Carlo Integration (38)

Solid angle sampling involves dividing the hemisphere around a point into discrete sections and calculating the contribution of each section to the overall lighting of the scene. This technique is useful for simulating point light sources, such as light bulbs or candles. On the other hand, light area sampling involves dividing a light source into small patches and calculating the contribution of each patch to the overall lighting of the scene. This technique is useful for simulating area light sources, such as fluorescent lights or LED panels.


While this may work quite well for samplig over uniform areas of regular shapes, taking 100 random points of light from the light source area for every pixel being evaluated seems incredibly inefficient unless supersampling or other forms of optimization are implemented. I have another concern, though - for any nonuniform light sources, such as circles of light that are especially bright in the center (e.g. the sun), this random sampling still has the chance of completely ignoring the most significant regions of the light source, and given the number of pixels being evaluated, that's almost inevitable. E.g. if a model involved a pinpoint hole to let a single sliver of light from a sun light source through, given even 100 points of random sampling, odds are good that the area beneath the pinhole would be shadowed, or that a small, dull circle of light would be cast underneath it rather than a single, bright point as it should be. Is there a better way to calculate lighting for finer details than random sampling?

You must be enrolled in the course to comment