11.96
#... |
|---|
|
So I do some test other night and take images st ISO 400 and 800 and process each group separately. Works fine. Then I think to combine the two sets which I did by stacking the two autosave.fits files which seem work fine. But wondering if this is correct or would it be better to restack all the subs from both ISO into a final autosave.fits then process that? Any ideas appreciated! Sigga |
6.49
#... |
|---|
|
Hi Sigga, Due to some algorithms used by these softwares for stacking the images (like sigma rejection algorithm) I believe it is better to stack images that has the same characteristics (exposure time, ISO settings etc.), and similar signal / noise ratios and distribution of noise. Even if the software is smart and normalize the images (therefore no problem because ISO is just a scaling of the signal), if you are not sure how it works under the hood : then don’t take the risk. If the images are too disimilar in terms of distribution of the noise, then these rejection algorithms may not work properly, because they make assumptions on the distribution of noise the across the bunch of images to say « we take this pixel in the average » or not. Would be good to have feedback from some developers of these softwares if they are in the community of Astrobin :-) Jeremie |
22.40
#...
·
2
likes
|
|---|
|
Sigga, It shouldn't matter how you combine the images as long as you combine the data using noise weighted averaging. Noise weighting will scale the data to maximize SNR in the averaged result. If you are using PI to do integration, simply select the "Noise Evaluation" option for "Weighting." EDIT: Ah, I realized that you are using DSS, which is really limited. Unfortunately, I'm not aware of a good way to handle that issue in DSS without a fair amount of external pre-processing involving noised measurement and data manipulation. Grouping the data before combination won't make any difference. Mathematically that's the same as just dumping it all into one big stack. And as Jeramie has pointed out, you may also run into problems with the stacking filters as well. John |
6.89
#...
·
4
likes
|
|---|
|
Hi, Let's take simple average, no rejection no weights no nothing. NxISO400 subs, MxISO800 subs First way, each sub contributes the same, namely 1/(N+M). Simple, that's what average means  Second way, the N subs that are ISO400 each contribute 1/N to autosave400.fits and the M subs that are ISO800 each contribute 1/M to autosave800.fits. Then you have autosaveFINAL.fits = (autosave400/autosave800)/2. Ergo, each ISO400 sub contributes 1/2N to the final image and each ISO800 sub contributes 1/2M to the final picture (differently). If N=M the two ways are the same. If N=10 and M=40 for example, every ISO400 sub counts like four ISO800 subs. So the result will be shifted towards the set with the fewer subs. Now, is one way more correct than the other? Maybe yes, maybe no. If the set with the few subs is of much higher quality, then second way is better. If you have very few subs in the smaller group, it is almost certainly not better. N=2 and M=100 means just two subs count as much as the other 100! I'd say do not overthink it. Try to integrate groups of same ISO and approximately equal number of subs and you should be good. if N=10 and M=50, then split M into 5 groups. If N is very very small (like N=2 and M=30) then it's better to just throw the two subs into the whole lot and do way 1. Cheers, Dimitris |
|
11.96
#... |
|---|
|
Thank you all for ideas and assistance, I try more experiments. Best wishes and be safe, Sigga |
2.71
#... |
|---|
|
Sigga, Dimitris beat me to it. It shouldn't matter how you stack the subs if the number of 400 and 800 ISO are the same or close to the same. But if there are significantly more of one set than the other, then you should stack the frames together to get one autosave file rather than combine two autosaves after stacking. The easiest way in DSS is to put all the lights and calibration frames for one ISO into one group and all the other lights and calibration frames into another group. Any frames that are common to both sets of subs are put into group 0. HTH, Ron |
6.65
#... |
|---|
|
Hi Sigga I find that stacking autosave files works great, when I have images from different nights, each night is a separate autosave that are stacked together. As Ron mentioned, each different ISO should go in a different tab. First tab is anything that applies to all frames, which could be flats, darks etc, as long as ISO, etc. are the same in all lights. Put lights with one ISO in the 2nd tab, then lights for the other ISO in the third tab. Different ISO need to be in different tab so that you can apply the appropriate bias and darks for each ISO, assuming you are using those. So if ISO 400 lights are in tab 2, put their ISO 400 bias and darks in that tab. Same for Tab 3, all your lights, darks, bias for ISO 800. Flats can go in the 1st tab and be applied to all, they don't need to be taken under the same conditions as lights. For stacking there is a special setting if you have different ISO or sub exposure lengths. Go to stacking parameters, light tab, and click entropy weighted average (high dynamic range), which is meant to actually take advantage of different settings like ISO, as in other HDR methods. One thing to note in stacking subs, sigma-clipping actually won't be used even if you click it, if each sub is too different. You will find DSS just averages. Sub exposures need to be similar (though not identical necessarily) . The reason is kappa-sigma methods need the subs to be similar, because they reject anything that is different to the level you specify. DSS knows this and will not use kappa-sigma if it knows it won't work. The program is quite smart. I use these kappa-sigma methods sparingly. Kappa-sigma is good to reject satellites etc in one image, but can hurt your S/N. There is a great user group for DSS for questions, I have always gotten good answers to mine there: https://groups.io/g/DeepSkyStacker Rick |
