Pattern in lens flare

In this (bad) picture I noticed a strange effect in the light lens flare (see area in DT screenshot).


From a magnification of 100% you can see a slight pattern in the image, that can’t be removed by denoising etc. either. Is there a name for something like this and a way to remove it?

The effect only occurs in the lens flare, the rest of the image looks normal.

Raw file: 20190913-1841-57.raf (47.7 MB)

It looks like a well known issue more obvious in some of xtrans sensors. It’s often called “purple flare with grid artifacts” and you can find many discussions about it on the web.
Unfortunately there’s no easy way to remove it, but if you like shooting in conditions which tend to produce lens flare, try to vary the composition of your image slightly with slightly different angles with regards to light Ray’s, and hopefully some images will be free of artifacts.

1 Like

Moire-removal routines might help with this sort of grid artifact. In dt I use Denoise Bilateral Filter: maximize all the values (radius and RGB channels) and apply the Color blending mode. If it affects other colours in the frame you could apply this with a drawn mask. Then I’d change the demosaicing to FDC and apply a Denoise Profiled luma preset to get something like this:

Not perfect, but the grid is minimized.

I’m using Darktable 2.6.2. Zoomed in as close as I can to your screenshot, I got it to go away using the profiled denoise on default settings. Hope this helps, but I’m sure someone here will know what’s going on.

Thanks for your replies!

@sguyader Thanks for the hint, now I have a name for the problem :slight_smile: I understand now that it is a sensor problem, which also seems to occur with some bayer sensors. I will experiment a bit here to find out what I have to watch out for when shooting with my camera to prevent that.

@sankos / @sls141: Thanks for your tipps to minimize this issue. I also tried arround with Denoising, but weren’t happy with the results. I will try your suggestions this evening.

@sls141: DT currently doesn’t have a noise profile for the X-H1, thus as far as I understand the profiled denoise shouldn’t do anything, so I wonder what it did in your case. I used the X-T2 noise profile for the X-H1 and did get small improvements, but not neraly as good as the ones on your screenshot.

@noboevbo the noise profiles are the easiest ones to generate, you should do it!

It’s true that the X-H1 and the X-T2 share the same sensor, but the X-H1 has IBIS mechanism and electronics, a different body, and maybe all of this has some effect on the noise profile. So, indeed you should make your own profile.

Sorry for the late response. I just found time to create an X-H1 noise profile with the (great :slight_smile:) tutorial on this site. With it, I get pretty similar results to the X-T2 profile, but if I apply a profile for a higher ISO than 200 (which was used in the image), like 1250, it removes the pattern in the lens flare. So thanks again for your support!

I would guess it’s caused by the filter stack in front of the sensor. There are many layers housing the AF sensor, anti aliasing filter, etc.