Hero Wavelength Sampling Bias

[PearCoding]

Our hero wavelength approach has a bias which lets the whitefurnance test or an empty scene with D65 environment illuminant return a value close to 0.9 instead of 1. This may resolve at very high spp, but should be not visible at all at least in such a simple scene.

[PearCoding]

Carefully choosing the spectral domain is reducing the bias. [360, 760] returns around 1.05 instead of 0.9 compared to the CIE spectral domain. More investigations are needed, but it is clear that the used CIE + Illuminant Tables are the culprit.