Affiliation
American Association of Variable Star Observers (AAVSO)
Sat, 01/28/2017 - 20:39
In the past I have always used AIP4WIN to do basic calibration before uploading to VPHot. I am exploring other options, in the case CCDStack.
In creating master frames, CCDStack does not have a median combine, but does have "sigma reject mean" and "clip min/max mean." From what I have read, various types of sigma clipping are used to substitute for median combine. May I assume this is true and preserves the linear nature of the processing?
Thanks, Ed
Hi Ed
A quick look reveals that these are fancy methods that are used to not loose S/N when you use median combining to eliminate cosmic rays.
They come with caveats:
http://mirametrics.com/help/mira_mx_7_ue/source/image_combining_methods.htm
I just take 15-25 flats and let MaxIM combine them after I buzz through them to make sure they all have donuts. It is easy to take enough bias and darks to overcome the S/N loss from median combining.
Ray
Thanks for the comment, Ray, that link is most useful. However, I am still not sure that it addresses my original question. Do these clip methods represent linear or non-linear processing? If non-linear, then they have no place in photometric reduction, If they do not introduce some non-linear processing of inages, then they are OK and a alternate to median combine.
Ed
Hello Ed
I would defer to Professor Henden on this one, but both clip methods are non linear. In fact, Professor Google says that Median combine is not a linear operator. Average is definitely a linear operator.
Even though a non linear operator is not desired, it "Depends" on what you are doing. Clipping a big cosmic ray event may give you better PT that just linearly averaging it.
I typically use 30 flat frames and average. When pushed for multi band, I drop back to 15 frames in each filter. I never use the clip operators.
Gary
Thanks, Gary. I agree that Dr. Henden needs to be consulted. I don't want to recommend to students any program that introduces non-linearity into the process.
Ed
Hi Ed
My understanding of sigma mean reject and clip min/max mean is that they do an average stack and then remove large excursions from the mean. I think that the details of how they remove the excursions would make all the difference in the world. I have always stayed away from methods like this just because I don't know what's is going on under the hood.
A cursory look at the instruction manual for CCDStack seems to indicate that Sum, Median, and Mean combines are all available. It isn't a great manual and I didn't spend a lot of time looking.
You might give the CCDWare User Groups a try. http://www.ccdware.com/support/
Jim
Thanks, Jim:
I have a long message from the developer on the CCDStack Community forum. I am going through that now. There is hope.
Ed
The comments by Mike Newberry (Mirametrics), given earlier, are really good in describing the various techniques than can be used to combine images. Any rejection or ranking method changes the image statistics, though this is not as important for flats as it is for the science frames. For all rejection methods, the best approach is to take lots of frames (10+) so that the population of possible pixel values is well sampled. Once you have lots of frames, the impact of the rejection on image statistics is minimized and so all of the schemes are acceptible. My preference has always been for mean/average first, followed by median, followed by sigma clip.
As a really poor, but real, example, I remember taking long exposures of a CV system, in which I wanted to stack, say, 10 exposures to get a good signal/noise detection. In the midst of these exposures was an eclipse that affected a single sub-frame. Stacking them with mean gave me a resultant value that was lower than the other methods, but which was correct.
With flats, expecially twilight flats, you often want to reject poor data. For example, in twilight flats, you may have a star streak across a frame if you are unguided, or star images to reject if you are dithering while guiding. So a straight mean is not appropriate for twilight flats. Likewise, you may have a cosmic ray hit that you want to remove (if you don't, then every pixel underneath that cosmic ray will have a wrong value, and will get applied to all subsequent science frames). So I highly recommend filtering flatfield images with some technique. I usually choose median because it is simple and works well with these high signal/noise frames. I think sigma clip will do fine for flats, especially if you follow the above guideline of taking 10+ flats.
Arne
Thanks, Arne, that was the answer I needed.
Ed