I want to determine an appropriate ambient value (-av) to approximate higher-order reflections in a geometrically complex scene.
The scene consists of a tree with multiple overlapping leaves, and I sample irradiance on a per-leaf basis. As expected, many leaves receive very low irradiance. However, some diffuse light should always be present, which is what -av is intended to approximate.
I am aware of the approach described in Rendering with Radiance (Chapter 5), but I am wondering whether the following could be an alternative.
I would run two reference simulations with rtrace, both using -av 0, but with different -ab: one with -ab 2 and one with a higher value (e.g. -ab 4). The difference in sampled irradiance would be used to estimate a suitable -av value for future runs with -ab 2.
I am certain that this difference will vary across leaves and under different sky conditions. Nevertheless, is this approach reasonable to calibrate -av or am I missing something?
There is in fact an option I added years ago when I realized that the -av value could be approximated by a broad average of interreflection calculations going on, anyway. It’s not super-reliable, but if you set -aw 40 or so with a -ab 1 or -ab 2 calculation, you should see a general raising of indirect values.
This option is turned off by default due to the hazards it presents when computing both indoor and outdoor interreflections at the same time. This causes the interior lighting values to be raised more than they should be. If you are doing everything outdoors or indoors, then there is usually no problem with this option.
Thank you for the suggestion; however, I am trying to avoid using -aw, since I’ve turned off interpolation (-aa 0).
I compared simulations with -ab 2 and -ab 4, but I did not observe a significant difference that could calibrate -av.
I suspect this is due to the low ground albedo and leaf reflectance. In a scene with higher-reflectance surfaces, the outcome would likely be different.