Hello,
thanks for the data. I tried it and got a result I'd expect. It's green but the color is same for Bilinear an VNG. My guess is that you applied STF with linked channels to VNG result and STF with unlinked channels to Bilinear result. I tried it and got something very similar to images attached to your post. Can you please try it again and stretch both Bilinear and VNG images with same STF process instance?
On the attached image (which has moderate manual STF), you can see that same dark pixels are in the Bilinear image as well, they are just less pronounced. So they are part of the original data noise. Bilinear debayering tends to smooth things, acting as low-pass filter. For smooth surfaces, it works great. But on sharp edges, it produces color artifacts - you can see these strange colors on star edges. VNG on the other hand tries to detect and keep edges. Result is sharper and there are no color artifacts. But the noise is also sharper as it is not smoothed out by low-pass filtering of Bilinear interpolation. So individual pixels, which are darker because of noise tend to stay individual dark pixels after VNG debayering.
I personally don't consider this as an disadvantage. Small-scale noise is easier to deal with when compared to large-scale one.
best regards, Zbynek
