On May 15, 2012, at 12:09 PM, Gerhard Fuernkranz wrote: > Am 15.05.2012 16:27, schrieb Stephen: >> >> I have an idea to improve highlights for a camera input profile: can I scale >> XYZ data linearly to match an increase in exposure? >> >> For example, +1 EV of exposure compensation = 2 x linear device RGB. Will >> XYZ also increase by the same factor of two? > > XYZ is a linear light space, so the numbers are expected to scale linearly > with intensity. > > I would wonder though, how accurate various cameras can realize the > prescribed f-stops and shutter speeds, how well they are repeatable, and how > uniform the steps between different adjacent f-stops and shutter speed are > distributed in reality. The biggest problem is flare in fact. The good profiling strategy is to shoot in a way that prevents flare almost completely (but it is hard to achieve for a casual user). Shoot a target, open RawDigger and check that the white patch is close to saturation, about 90% of max value in raw. Once that is done, shoot a series of images adjusting the exposure down 2 stops at a time, 3 times. That allows to make a synthetic target covering 11 stops. More is rather useless as flare is too strong. -- Iliah Borg ib@xxxxxxxxxxx