I have had occasional color-banding that didn’t have too much effect on the final processed image, but I’m trying to track it down. The screenshots below are from 310 30s subs at gain 5, offset 8 -10°C, all with no processing other than the screen Histogram stretch. No filter was used, other than the UV/IR window in the QHY camera (which has a similar shape and bandpass to the Astronomik L2 filter, I believe). I wouldn’t normally shoot OSC so close to the moon, but I was testing my new CEM40 and its raspi-usb connection issues (but that’s a whole different discussion for other places).
(red is the about like green, no obvious pattern in the stacked image) edit: I can see cross-hatch in the posted green stack jpeg that wasn’t so obvious looking at in Siril itself.
Where is the cross-hatching most likely coming from? If it is from the camera, why is it not visible in a single sub, and why does it survive reasonably aggressive dithering?
A bit tricky to know from those examples alone, since those are the exact kind of things introduced by JPG compression itself. I assume you aren’t processing with compressed JPG though…
I see the pattern within Siril, when looking at the histogram-strech image. What sorts of things Siril does with the image to allow it to be displayed on the screen, I do not know.
Hello.
Generally the blue channel is the channel with less signal. In this condition, because of the lack of signal you can have an increase of some artifact during registration. Especially moiré pattern as you can see.
With a closer look you can ever see it on the green channel.
I think that your subs have too short exposure with gain to low.
I’m wondering if the demosaicing algorithm contributes to this effect. It sure looks like a lot of compression or interpolation artefacts in the images you showed us. Which algorithm are you using and can you try another to see if this changes?
Thanks.
I used Rice fits compression, and the DCB de-bayer algorithm (based on trying to interpret some of Ivo Jager’s comments, but he would not have been talking about Siril and might not have been thinking about the full range of algorithms you have implemented. I had forgotten about the “may cause background artifacts” comment in the algorithm pop-up.
I will re-process tomorrow with RCD, and also with fits compression off.
I was only able to sort-of test the effect of RCD vs DCB, in that I wasn’t expecting to keep the subs. I had not yet deleted the subs, but I had deleted the flats. I re-processed using RCD and a mismatched flat and the cross-hatching was much, much less evident. This was with Rice fits compression still on.
I’m not too surprised because the pattern looks like the tiling pattern the demosaicing typically uses. Keep in mind that you’re looking with a histogram equalization, so it’s probable that details like these will not be visible in final images.
