I think the trouble makers here are highlights when at least one channel has clipped. Different demosaic algorithms or highlight reconstruction methods could help. Careful EV compensation or exp. bracketing when shooting to prevent near clipping.
Years ago, when I was using K10D and some old version of RawTherapee, sometimes weird maze-like patterns would form on near-clipped highlights, that weren't there on jpegs. K10D has only 10M pixels and strong AA filter, so sensor resolution or lack of a filter probably are not affecting this. With Bayer sensors, when one or two channels clip, demosaicing has big gaps of information to fill, producing artifacts. Just my theory.
|