This project uses Reinhard Tone Mapping on the Cornell Box, which uses a log scale for the Y value of the tristimulus curve, then uses L = Y/(Y+1) for a normalization. I compare with a simple clamping strategy that divides intensities by a max intensity. The Reinhard image was not given sufficient time to converge, thus the noise. In hindsight, I should have saved out the raw data for each image, which could then be tone-mapped with different strategies as they were developed. This would have given a better comparison of the techniques and ensured that the values used where identical.
Tone mapping by clamping (left) and with Reinhard tone mapping (right).
Projects | Home