That's right and that's basically where I took this idea from. But nevertheless I experimented to use it in downsampling as well (I could imagine preventing color loss that way; think at barely visible detail in the highlights, clipping while reducing image size ... then think at changing contrast to make this little things move away from clipping border before reducing ...). It's different here, but funny since it works as well. The test image shows gradients made of color stripes and by common filters/algorithms they are reduced to grey stripeless gradients. But what if I want to retain little more color ... my aim is a gradient of slight color.snibgo wrote:From this thread and viewtopic.php?f=22&t=21415, it seems sigmoidal processing is suggested for reduce halos when enlarging, and not for reducing.
//edit: shame on me! I just enlarged the test image and saw the gaps between the colored stripes are not grey (as I assumed) but filled up with the complementary color! I guess in usual downsampling these complementary colors eliminate each other to grey. By contrasting I destroyed this balance and the colors got visible, on the one side to RGB, on the other to CMY. Now I know, there the colors come from. =)
So sorry, everything is allright! I was so confused where these colors could come from and didn't watch closely at the test image!