Too bin or not to bin...that is the question

Astrophotography: share your photos & discuss techniques
User avatar
menardre
Vice President
Posts: 886
Joined: Tue Jan 17, 2012 6:09 pm

Too bin or not to bin...that is the question

Unread post by menardre »

Having a little photography background, I have always tried to image with the highest resolution possible. Using Photoshop it is much easier to 'soften' images than sharpen them.

When I started astrophotography I kept the same philosophy i.e. always image at the highest resolution. I have read many articles related to the pro's and con's of 'binning'. To me 'binning' is just throwing away resolution. There is a difference with astrophotography, and that is there is a point where higher resolution does not provide more detailed image. This is related to sky quality and 'seeing'. Based on atmospheric conditions the best resolution that you can attain is between 1 arc-sec to 2 arc-sec even with good clear skies. So imaging at higher resolution than that does not increase the details in the image.

One of my imaging setups utilizes an 11 inch SCT and a ZWO ASI2600 camera. Based on the focal length of the scope and the size of each pixel, that setup has a 'resolution' of about 0.396 arc-sec/pixel. Astronomy call this 'over-sampling', since the best seeing limits 'resolution' to 1-2 arc-seconds. To compensate for this, you can select a different 'binning' mode for the camera. Binning is essentially adding the output of adjacent pixels to make one larger pixel. Binning 1x1 takes every pixel as a single entity. Binning 2x2 takes four pixels and adds them to make one pixel four times larger. Binning 3x3 takes nine pixels and adds them together, etc. Binning reduces read noise and also allows for shorter exposures.

But does binning sacrifice image quality??

All of the experts say you should do binning if your setup is over-sampling. So I decided to run my own test. I took one hours worth of images (thirty 2 minute exposures) with the camera set for 1x1, then 2x2, then 3x3, then 4x4. All in the same night under same conditions. I also collected darks, flats, and biases for each configuration.

I then processed them the same using Pixinsight. In order to keep everything fair, I only processed the images up to the point of stacking them. This included flat calibration, flat master, image calibration, star alignment, local normalization, stacking, background extraction, denoise, autocolor, and ArcSinH stretch. I stopped at this point because going any further is simply 'photoshopping' and I wanted to evaluate and compare the 'raw' stacked version.

So attached is the result of imaging at 1x1, 2x2, 3x3, and 4x4 binning. It should be noted that files sizes are dramatically different. The 'raw' FITS 1x1 file is 50,969KB, the 2x2 file is 12,747 KB, the 3x3 file is 5,668 KB, and the 4x4 is 3,183 KB.

Tell me if you can see any difference.

This pretty much convinces me that the experts are correct. Binning up to the point of being within the 1-2 arc-sec/pixel does not reduce image quality. Do you agree????

I will next complete the image processing of each version and see if there is any difference. I may then even have each printed 8x10 to see if there is any difference when printed. I know this sounds like an over-reach, but I think it is worth the effort.

Roger
M64 Master 1x1_DBE_EzDN_Autocolor_ArcSinH.jpg
M64 Master 1x1_DBE_EzDN_Autocolor_ArcSinH.jpg (16.99 MiB) Viewed 652 times
M64 Master_2x2_DBE_EzDN_Autocolor_ArcSinH.jpg
M64 Master_2x2_DBE_EzDN_Autocolor_ArcSinH.jpg (3.28 MiB) Viewed 652 times
M64 Master 3x3_DBE_EzDN+Autocolor_ArcSinH.jpg
M64 Master 3x3_DBE_EzDN+Autocolor_ArcSinH.jpg (1.12 MiB) Viewed 652 times
M64 Master 4x4_DBE_EzDN_Autocolor_ArcSinH.jpg
M64 Master 4x4_DBE_EzDN_Autocolor_ArcSinH.jpg (713.96 KiB) Viewed 652 times
Roger M.
Celestron CPC1100 EDGE, Stellarvue 130T refractor dual mounted on iOptron CEM120 on permanent pier mounted in Observatory. Imaging camera ZWO ASI2600 OSC, guide camera Lodestar or ZWO ASI290MM.
User avatar
Pete
Astro Day Coordinator
Posts: 3999
Joined: Fri Jun 13, 2003 9:03 pm

Re: Too bin or not to bin...that is the question

Unread post by Pete »

Thank you for this Roger. I've gone from 1X to 2X binning on the 14" as it's still grossly oversampled and theoretically there's no loss in image quality. Last year I was running 4X for a while and while images still came out good a forum posting about binning in post processing pushed me into reverting to 1X for a while.

I'd remembered that at 2X the pixels were added and not averaged. But failed to recall that adding only happens with RAW8. Haven't even tried astro imaging at 8 bits as I typically stretch the Hell out of my images and stretching 256 bin depth ain't gonna cut it. I sort of figured the averaging with RAW16 when doing the Hockey Stick the other day and not seeing a huge increase in brightness.

Post processing with 2X binned images is ever so much faster and is still conservative when it comes to sampling, so I guess that's where I'm at with the 14. Got an adapter coming that'll allow mounting my old focal reducer and that'll be a big help if the vignetting isn't serious.
Pete P.
User avatar
mark.m
President
Posts: 298
Joined: Fri Feb 03, 2017 1:13 pm

Re: Too bin or not to bin...that is the question

Unread post by mark.m »

Roger:
Thanks for a great writeup. I usually bin 3x3 with my QHY268M, and the primary motivation is reducing the amount of disk space that I need for archiving. (I permanently archive every image that I use during my variable star work.) With 3x3 binning and compression, I can squeeze an entire night onto a single backup DVD disk. Binning does nothing to either improve or degrade my brightness measurements; it has no effect on SNR if you're measuring brightness.

Your post is coincidentally timely; I just finished rewriting the part of the AAVSO observer's guide that deals with binning. (Talk about heated debates with the review team!) The one twist that's awkward with binned images is trying to determine if any pixel has saturated within the core of a star's image. (When you're measuring brightness or performing astrometry, that's a fatal mistake that should cause the entire image to be deleted.) Once multiple pixels have been binned (either by adding or by averaging), you can no longer easily tell if the pixel has saturated by looking at its binned value. For astrophotos, this probably isn't a big deal. And as Pete has pointed out, binning and file format interact -- it's important to use a file representation that's consistent with how the binning is being done (adding vs averaging) or you will degrade image quality by binning.

- Mark M
Mark M, AJ1B
Portsmouth, RI
Celestron 14" and Meade 10" SCTs
QHY268M + SBIG ST-9
GM2000 (10Micron)
American Association of Variable Star Observers (AAVSO) observer code: MMU
User avatar
menardre
Vice President
Posts: 886
Joined: Tue Jan 17, 2012 6:09 pm

Re: Too bin or not to bin...that is the question

Unread post by menardre »

Pete/Mark

Great comments.

This is one area where on-line posts are almost useless.... there is so much miss-information and assumptions. Some posts say to always bin up to the sky resolution limit and others say to never bin (or only bin during processing). That is why I decided to run my own test.

I think the idea of binning was easier to understand with CCD camera where the binning was done in the analog domain. Binning with CMOS cameras is more a function of how the manufacturer has implemented binning, especially with OSC CMOS cameras.

I plan on using 2x2 binning with my 11 inch setup and 1x1 binning with my 5 inch refractor.

Thanks for your responses.

Roger
Roger M.
Celestron CPC1100 EDGE, Stellarvue 130T refractor dual mounted on iOptron CEM120 on permanent pier mounted in Observatory. Imaging camera ZWO ASI2600 OSC, guide camera Lodestar or ZWO ASI290MM.
Post Reply