woodcock tracking

Today I have tried to compare our rendering of homogeneous media with the same material settings using woodcock tracking and the other calculating the extinction directly. They do (luckily enough) look quite similar, but their was a huge speed difference between the renders. First a render with our homogeneous system of merlot in low concentration:

 This is a render with just one pass, but because we can calculate the amount of light absorbed in the material with e^(-t*extinction) we can get a very clear result with just one pass. The colored dots that can be seen is the result of rays being reflected on the surface of the liquid, which we are doing with russian roulette which requires more samples per pixel to get a smooth image. The woodcock render of merlot in low concentration looks like this:

This is after 35 passes of the render. This render is done using woodcock tracking, and it only approximates the correct light absorbtion in the material. It is an unbiased method, so given enough passes it should convert to a correct solution, but because of the randomness of the algorithm is requires a lot more passes to convert to a clear result. It does however look quite similar to the homogeneous render, so our next project is to speed up the rendering.

homogeneous coke 1 pass:

woodcock tracking, coke 68 passes: