Average, Median, Sigma Clip or other?

Hi Lads,

When combining images, which of the above do you select? I usually pick Median, but just wondering as to the advantages and disadvantages of each.

Cheers,

Michael,

This is a LOADED question. Some quick comments:

(1) Median combine should be avoided in general and used only when there is a problematic exposure (such as a satellite/plane track crossing the field of view). Of course, you can toss it out such an exposure but this may not be easy if you only have a handful of exposures (ex. 2x15 min R, 2x15 min G and 2x15 min . Now, why do we avoid a median combine? See (2) below ...

(2) A mean (or average) combine is much better due to the fact there is more data in the master file. I read a lengthy (mathematical) paper which showed that there is 1.6x more data in a master file from an average combine than a median combine.

(3) A sum combine (you did not ask about it but lets be complete) ... this is the best of (1)-(3) but needs some attention when saving your master file. Since an average combine will produce figures after the decimal point and which are eventually dropped (ex. 42,049.342 ADU -> 42,049 ADU), there is a small amount of information (0.342 ADU in the example) which is dropped. In contrast, a sum combine does not drop any information (ex. 87,783 ADU) but one should be careful to save the master file as 32-bit or IEEE floating point since the ADU's are not within the interval 0 to 65,535 as is the case with (2) but the interval 0 to x where x equals a multiple of 65,535.

(4) Sigma-clip ... this is another good technique and basically drops the values that fall outside k standard deviations. In other words, if you choose k=3 which translates to 99% of the gaussian distribution (ie. x-3s to x+3s), you will drop 1% of the extreme values (really the 0.5% lowest and the 0.5% highest). This is especially useful if you have problematic pixels (ex. cold and/or hot pixels) which cannot be corrected for when using darks and flats.

(5) Sigma-clip median combine ... a combination of (1) and (4) ... this is also a good method since you drop problematic pixels and then work with the good pixels to derive the median ADU

(6) Sigma-clip sum combine ... a combination of (3) and (4) .... in my opinion this is the best way to go since it represents the best of all worlds, for you drop the problematic pixels (via the sigma-clip) and then sum the good pixels without dropping information due to rounding (vis a vis average combine).

Hope the above helps and I did not add any confusion.

Anthony,
Many thanks for such a detailed answer.
I'm not sure though how to do a Sigma Clip Sum Combine. My guess of it is as follows:
1) Combine the images by Sigma Clip and call resultant image, say, "Combined Sigma Clip".
2) Combine the images by Sum and save resultant image as, say, "Combined Sum".
3) Then combine "Combined Sigma Clip" and "Combined Sum" by Average and save as "Combined Final". It is this image that is then processed.
Is this correct method of doing it?

Cheers,.

Michael,

I hate doing this to you but far from it ....

Please remember that the sigma-clip is done at the pixel level. Thus, you cannot save a master frame which is sigma-clipped and then somehow average thereafter Put another way, for pixel(1,1) frames 1, 2, 3, and 5 may be left after the sigma-clip whereas for pixel(1,2) the corresponding frames may be 2, 3, 4 and 5 (trivialized example).

This is totally a software operation and you must have the software that will itself perform such a combine from start to end. I am afraid there is no way around it!

PS. The master from (2) above in your flow WILL include pixels dropped from (1).

Anthony,
Mind me asking, what software do you use to produce a Sigma Clip Sum Combine? I haven't seen that option in Maxim or Registar, only a Sigma Clip OR a Sum but not both simultaneously - or have I missed something?
Cheers,