anton_foy
13th November 2024, 17:56
Using ColorYUV(autogain=true) renders an incorrect representation of the colors as it seems to just adjust the Y-channel and " histogram stretching" (if that even is the correct term) is not applied to the other channels. The saturation will then take a blow if the image is dark. I would think that doing this in rgb would be better since I had some success using python to code and get some good results.This is the first point I want to highlight but the main point is that ColorYUV(autogain=true) has problems with flickering and I would speculate that it could be avoided if the code would analyze a downscaled (bicubic or bilinear maybe) version of the image to lets say 64x48 or similar resolution in order to more smoothly skip the small fluctuations in luminance. Is my logic totally off in this reasoning?
Not to forget, also some kind of limiting or highlight roll-off would be great since autogain cuts off too much information in the highlights. I am aware that "ex_autolevels" can give good results in this area but I believe a more simple solution could be available aswell.
Not to forget, also some kind of limiting or highlight roll-off would be great since autogain cuts off too much information in the highlights. I am aware that "ex_autolevels" can give good results in this area but I believe a more simple solution could be available aswell.