I'm processing some astronomical images. These are Canon CR2 raw files.
When I open them in Camera Raw I find that the bright stars have a dark offset halo around them in the green channel. Red and blue channels look normal.
The halos have pixel values of zero (general sky background around them is ~3000) and then jump to values of ~20000 on the star itself. It looks as if a severe sharpening process has been carried out but only on the green channel.
If I open the image with a different raw converter it all looks normal. Here is a 200% screen grab.
Is anyone aware of a problem with Camera Raw??
what camera are you shooting with and what does the other RAW converter image look like?
I'm using a Canon EOS 5D Mk3. This is a 90 second exposures at ISO1600.
The image below shows the R,G and B image planes from DCRAW (top line), and the similar ones from ACR (bottom line).
There is some nebulosity in the image - so the noisy background is real. My big concern is how the dark rings can have occurred in the green channel.
Thanks for any help or suggestions you can make!
Sorry, I should have added that I'm working with the latest versions of Photoshop CC and ACR!
Hi Graham! Thanks for reporting this issue. We'd like to investigate it further. Can you share a CR2 file that shows the issue via a file sharing service (Dropbox, etc.)?
I've copied a raw file to Dropbox: Dropbox - ACR - Simplify your life
I've also copied the related xmp file in case it's useful.
The group of stars I was using for the screen-grabs are towards the top right corner above the nebulosity. It's worth noting that as you look around the green channel the darker part of the ring around the stars generally points towards the centre of the image.
Hope this helps, and thanks for your support.
PS, I forgot to mention that this is on 64 bit Windows 10, version 1803
I've downloaded the files and opened up a bug for investigation. Thanks again, Graham!
It’s been a couple of months now, have you had any feedback about the bug investigation?
Having just now looked at this thread, my opinion is that the stars with green-channel dark halos are the same ones that have a magenta-colored halo to being with, and since magenta is the opposite color to green, the uneven sharpening is because there is harder edge in the green channel so the sharpening bites into it a little harder. There might be just as much of a halo in the other channels but those channels still have a red and blue fade to them that is not an edge to sharpen. That said I'm not really seeing what you're seeing as far as halos, either, using the supplied XMP. Maybe ACR 11.2 that I'm using does things better, now.
Hi, thanks for your comments.
I hadn't tried reprocessing the image in ACR11.2.1, but having just done it I agree that the result does look slightly better. However, my main problem still remains.
When I carry out the processing I set all the ACR sharpening sliders to zero as I don't want to do any sharpening at all. I initially just want to be able to register and combine about 40 individual images to improve the signal to noise - the sharpening etc. all comes later in the workflow.
In the image below I've done a cross-section through one of the bright stars and plotted the R,G and B channels. The red and blue channels look much as I would expect them to (bearing in mind that these bright stars are heavily saturated). However, the green channel looks completely wrong. I'm only guessing that it's had some sort of sharpening carried out, but whatever it is, it's taken some of the pixel values right down to zero - way below the background level and certainly not something that could occur in-camera.
Do you know if ACR still applies some form of base-line sharpening even with the controls at zero (e.g. to offset the effects of the anti-aliasing filter?), or am I missing something in my settings? I've experimented with turning off the chromatic aberration and profile corrections but the effect is still there (actually slightly worse as the black area is more offset towards the centre of the image - consistent with some chromatic aberration). Whatever it is, it just seems wrong to me that one channel is being so heavily modified compared to the others. I can imagine that it might not be detectable in a normal image but an unresolved point source like a star really does show up any optical or processing faults !!
Thanks once again for looking at this.
I would suspect it's related to the RGGB layout of a standard Bayer array - that is, twice as many green-sensitive pixels as blue/red-sensitive. That naturally yields better definition in the green channel - so if there is some baseline sharpening applied equally to all channels, you might get something like this.
But let's see what Jeffrey comes up with.