Non-rectangular region-of-interest for analysis

Do you take flats? These will normally correct for specifically this. It's vignetting.

Yes of course. But the correction is not perfect, and in particular with the polluted city conditions I usually photograph in it still becomes visible after correction.

Also, vignetting causes diffraction which affects the star shape and contrast. And the SNR is lower. Lots of things that show up, even after flatfielding.

I understand that indeed, however I would say that proper flats would then still produce a better result overall and effects that still show up can then maybe be corrected with the light pollution tool. The other effects are there ofcourse, but less noticable than the vignetting I think. Also, cropping in a circular manner would prevent you from producing mosaics, even if they just consist of 2 panels. I think though, if it's something people would want, that it's not that hard to implement. I'll tag Mabula for him to chime in on it. @mabula-admin

Of course I'd use all the available calibration tools and not try to replace them with this option 😉.

To illustrate the issue, I've attached three pictures of the same mosaic, the only difference being 1st or 2nd degree LNC, or none. All calibrations have been applied, darks, flats, bias. MBB is set to 5%. The observing conditions weren't optimal for all of the panels, in particular the ones on the left hand bottom-ish side show the reason why I'd like this region-of-interest option: the bright corners sticking out. Depending on the way the sky brightness varied, they sometimes are darker instead of brighter, and frequently darker on side and brighter on the other.

LNC can to some degree deal with these corners, and MBB will fudge them out to something palatable, but skipping the corners all together will give LNC much more freedom to properly remove the gradients in the images (instead of having to find a balance between the gradients in the central part of the image and the deviating corners).

It does indeed mean that for mosaics the panels should overlap more, in order not to create gaps between the panels. But this is basically the same difference as using a smaller versus larger sensor on the same telescope.

I've gone through the ordeal of removing the corners with the light pollution tool in a large mosaic, but that's something I'd prefer to keep as a once-only learning experience... 😉

Posted by: @ralph

I'm using a full frame camera (Canon 6D) and frequently run into degraded image quality at the corners of the frames, while large parts of the frames are fine. In the mosaics I make, these corners affect the final image in a negative way.

It would be nice to have the option of selecting a (circular) region of interest to crop the image to include only those regions for further processing.

 

Hi @ralph & @vincent-mod,

The option of only using a circulair part of the calibrated light frames for subsequent processing should definitely be possible and I can understand the idea. I know that sometimes flat-field calibration with full-frame DSLRs like the Canon 6D can be rather troublesome to get correct all the way into the corners, but this almost always means that a suboptimal way is used to create proper flats in my experience... so my question to start with is: how do you create your flats and what optics are used? A telescope or an regular objective?

But the correction is not perfect, and in particular with the polluted city conditions I usually photograph in it still becomes visible after correction.

Also, vignetting causes diffraction which affects the star shape and contrast. And the SNR is lower. Lots of things that show up, even after flatfielding.

If the correction is not perfect, then that is indicative of a suboptimal flat creation method. You should be able to get perfect flat-field calibration, it is not easy, but definitely not impossible. So I am curious about how you make your flats? (so skyflats, tshirt, flatpanel etc.... but include also iso, exposure time of flats and of the bias and/or darkflats )

Now, light pollution is in all data, even in 3nm H-alpha exposures from the darkest observatories.. but it's a separate additive signal. It will not affect flat-field calibration which works multiplicatively. The light pollution in the corners of the sensor are low due to vignetting. With correct flats, the light-pollution will be multiplied correctly in those corners. If you create proper flats, the light pollution off course will still be there after calibration, but the vignetting and dust bunnies etc... should all be corrected perfectly I would think. I shoot all the time from very polluted skies (city center Arnhem, The Netherlands) and have always been able to get good flats even with  widefield (12-50mm) opjectives and full frame camera's. The camera and optics combination do warrant different methods to create proper flats. Some setups are very forgiving, others not. I also know that the camera sensor can be a problem to get good flats, this is a known problem with some Nikon's and Sony DSLRs. Those sensors can show ringing after flat-field calibration, that is a problem outside of this discussion I think...

Regarding "vignetting causes diffraction which affects the star shape and contrast", I think the optics will create different diffraction patterns towards the corners, vignetting or not. The vignetting is not the cause. If that is the problem, the optics are not able to properly illuminate the whole sensor, or put in a different way, the optics don't meet your expectations? It is well known, that practically all optics will be of less quality the further away you are from the optical axis and bad diffraction will happen and contrast will suffer, and colors as well.

Noise, and SNR are clearly affected by vignetting, yes, with strong vignetting, the corners will have clearly higher noise values, relative to the sensor center, from the flat-field division/multiplication. But again, rarely I have seen this to produce a visual problem. Maybe you can illustrate this with some images?

Personally, for your problem, I would simply calibrate all my frames, save in 2) calibrate.

Then with the Batch Modify tool in 9) perform a batch crop of the calibrated light frames and remove the problems that you don't want to have. Then load the calibrated and cropped light frames as lights and process as usual from 3) Analyse stars.

Finally, I would invest more time to perfect the creation of your flats. To me, it seems to be a big waste of data to shoot with a full frame Canon 6D and then remove a big valuable part  because your flats are not working as they should be... Use a smaller sensor (or crop in the camera) with the same optics, or get better optics if you want to use a full frame sensor.

The RFC to select a circulair part of the image for processing is possible, but it will be a very low priority when I compare it to my extensive list of ToDo's, so I think that it will take a long time if any before it can and will be implemented I am afraid. I hope that makes sense?

To conclude, an example of very strong vignetting in the corners of the sensor using a Nikon D800 with a Takahahsi Epsilon at F/2.8 (lower f-ratio is more difficult for flats), corrected perfectly with bias & darks and proper flat frames using a flatpanel.

Raw frame and calibrated frame with extreme stretch and saturation to show the possible problems

The same, but with Background Neutralization enabled in the preview filter (setting below the histogram)

The Nikon D800 sensor chamber is making proper flats rather difficult for the full-frame sensor, but it still can be corrected perfectly I think even at F/2.8. I don't see any visual issue with star diffraction or noise in the extreme corners after calibration as well and in this case, because of full-frame with focal ration of 2.8 I would expect it to be, if it was a real serious issue. What do you think Ralph?

Cheers,

Mabula

I very much understand the low priority you are giving it. And I by no means want to push it higher up on the list, there are plenty of other things much more important than this, also to me.

Depending on the situation, I take flats in various ways. I have an integrating sphere, I occasionally take sky flats, but the flats used here were taken with a light panel. Normally I take many tens of flats at low ISO (50 or 100), but sometimes at the same ISO setting as the lights were taken since some sensor effects are probably ISO dependent. Flats are always taken with matching darks.

The optics I use vary from 180 degree circular fish eye (basically impossible to take flats), Samyang 14mm F/2.8 and 135mm F/2, and very often a Canon EF400 F/2.8L, occasionally a William Optics FLT132, and lately a GSO RC8 without flattener (used in the examples above).

In particular the RC8 is giving me some issues, but then I'm pushing it to the limit regarding flatfielding. The problem with the RC8 is that it's not properly baffled for the 6D that I'm using it with. Near the corners significant stray light arises, see the picture of an RC6 with same problem:

In a properly baffled system only the entrance pupil of the optics should be visible, wherever you look into the optics from the focal plane. Here you can see a crescent appearing in the dark part of the secondary, as the camera was moved more towards the edge of the field, and eventually another crescent appears next to the entrance pupil of the primary.

What this means is that light from outside the direct field of view ends up on the sensor in the wrong spot. Put a bright star in that out-of-field spot and you end up with arcs in the image. I've seen those in my exposures.

Now, if there is a gradient in the light pollution, the directly imaged light will of course show that gradient. But the stray light as it is in this RC8 causes a reversal of the gradient in the stray light affected part of the image.

Below two exposures of the same field, both calibrated with the same flatfield, same optics, two different nights, one in twilight with lots of sky background (yes, I know I'm pushing it to the limit here).

This image shows no problems with flatfielding, just like the examples you showed:

However, when continuing into twilight, or very light polluted conditions, this is what I see happening:

There is a slight gradient in the image, not that annoying, but as soon as we end up in the heavily vignetted part the stray light starts to dominate the image and the gradient is reversed. Now this is very difficult if not impossible to correct with LNC of whatever degree. And correcting this by hand with light pollution correction is possible but a lot of work. And I think you'd share the idea that it's better to attack the problem at the source than try to patch it up afterwards.

So my suggestion is that the user is given the option to cut out the really bad parts of the image anyway, and leave the LNC to deal with the remaining benign gradients.

Finally, an example of diffraction due to vignetting:

This is a crop of an M31 image, taken with the EF400 F2.8 and 6D. The mirror in the 6D casts a slight shadow at the bottom of the image for these fast lenses. The picture is taken at F2.8, so no diffraction from the aperture blades is expected. And most stars are nicely round, with interesting "cat's eye" diffraction effects due to gentle vignetting due to the lens design (e.g. the top left bright star). But look at the bright star on the bottom towards te right: this one has a very pronounced diffraction spike in the vertical direction. And this is due to the vignetting caused by the mirror, also visible in the flats if you look for it. Also in your Nikon D800 image by the way.

I don't have any examples at hand of similar vignetting effects in other optics, but in general: as soon as the sensor extends beyond the designed corrected field, the optical quality drops rapidly. And instead of cropping this circular good part of the field by a square or rectangular crop, why not mask it in (pre-)processing with a matching circular mask? Or another shape of e.g. an off-axis guider protruding just a bit into the image?