Image Calibration - best practice? And a question

[joelshort]

Returning to the original post, I'm still unclear as to whether I should be using darks or just using CosmeticCorrection with a master dark.  (Dimitris touched on this but I'd like more info).  Since I'm using a low noise Sony camera it has been suggested to me that using CosmeticCorrection is better because darks can add unwanted noise back into the image.  So my practice so far has been to adjust the cosmetic correction to remove hot and cold pixels and use that process icon in the BPP script when calibrating lights, and not using darks at all.  Someone tell me what I'm doing wrong. 

[RickS]

Now another question: within the BPP script, is it better to use master bias, master dark and master flat for calibration, or use all the raw bias, dark and flats?  Why or why not, or doesn't it matter?  Obviously I can create the master bias and master dark and keep those on hand for future BPP runs and I assume that BPP will calibrate faster this way, but is it best practice to use masters instead of raw subs?

It doesn't matter.  If you supply raw calibration files it will build the masters then perform the rest of the actions.  If you supply masters it just skips this initial step.  As you suggest, it is quicker if you already have masters.

Cheers,
Rick.

[sctall]

So adding on to this
I usually have a Master Dark and a Master Bias. I have a Starlight Xpress 814C, so I have pretty good control over some noise.
But it's still there.
Anyway, I always have fresh Flats, ( of course ).  Is it OK to add the 2 masters in BPP and then all of the Flats subs, and Lights?
Or should I make a Master Flat first, and then either add the Lights, or use Image Integration for the lights, now that I have all the masters?

I usually put the 2 masters in BPP, add in the Flats and Lights, and GO!!!.
Then I take the registered subs created by BPP. and run them in II, and see if I can get a better integrated image using the sigma sliders.
I have not seen a difference by doing this, but I will admit I really do not know what I am doing.
Any advice on what to look at when doing this to compare?

Scott

[Warhen]

Joel, A few things to back up some of what the guys have said. Also, I think your questions will be answered in our PIP-1 'Stack' tutorials you just ordered.

BPP is very good, and 'calls out' to the other independent modules to do its work. While it is powerful, one limitation is that you can only use one master bias or dark, so if you have varied exposure times or temperatures, you'll have to choose only one.

Additionally, it allows for only one pixel rejection algorithm for light frames. So if you have say 9 subs each for color channels, but 20 for luminance, you would again have to choose one algorithm, to the detriment of either L or RGB. These are the reasons why integrating with ImgIn may be better. Also, though BPP offers the same control as ImgIn with regard to Sigma rejection sliders, it can't use a premium, user-selected reference image as can ImgIn.   

BPP calls out to Star Alignment and ImgCal so for those functions, the results may be equal to what you can accomplish with them independently. So, while you may be well pleased with the BPP produced masters, many of us use BPP to generate the calibrated, cosmetically corrected, and registered subs, then integrate with ImgIn.

As to your question on CosmeticCorrection- this is very different than calibration. After light frames have been bias corrected, and/or dark subtracted, and flat divided (or not!), CosCor takes care of nonlinear issues such as hot pixels, and also user-defined column defects. Take the time to create a template for each camera. Post calibration, the artifacts mentioned are 'filled in' with the average values of the surrounding pixels. Whether called out by BPP, or run separately after ImgCal, watch CosCor make residual artifacts disappear!

Using darks or not is your call. It has been demonstrated that using darks with some of the low-noise sensors (Sony Exview/Super HAD) can indeed be more trouble than they are worth. If using flats however, you'll need to take at least a master bias.

Nothing wrong with using pre-made masters when using BPP, whether generated by BPP or independent modules.

Lastly, as you 'd see in Jordi's PowerPoint, or our tutorials, noise quotes can be read by the Process Console or the Noise Evaluation or other Scripts.

[dnault42]

I haven't been monitoring the PixInsight forums until recently and just came across this thread and felt I should respond.

My tutorials shouldn't be considered the definitive way of doing things.  They are how I approach image processing as best as I can describe.  There are certainly people that know more about PI than I do and I will defer to their knowledge.  That being said, I had some specific reasons for what I presented in that tutorial and why.

Rick, I agree it is a bit masochistic!   However, my point was to show all the steps in as much detail as possible so that people could understand what is going on under the covers of BPP and could more effectively trouble shoot when problems do occur.  For 90% of the data we collect BPP does a great job, especially with the calibration portion of the process.  I definitely failed to make that clear and will update the tutorial accordingly.

I did specifically leave off details about image integration parameters and tuning them to maximize SNR.  I did this because it's a complex subject and I felt it needed a separate article especially since the tutorial was so long already.

I'm less concerned about the sub used as the alignment reference than the one used as the ImageIntegration reference.  I do cover picking the II reference frame in detail.  I do specify how I pick the alignment reference as well, but it's just a simple statement about how I usually use the first frame in the sequence since that is the one I used to frame the target.

Regards,
David

[RickS]

Rick, I agree it is a bit masochistic!   However, my point was to show all the steps in as much detail as possible so that people could understand what is going on under the covers of BPP and could more effectively trouble shoot when problems do occur.  For 90% of the data we collect BPP does a great job, especially with the calibration portion of the process.  I definitely failed to make that clear and will update the tutorial accordingly.

I did specifically leave off details about image integration parameters and tuning them to maximize SNR.  I did this because it's a complex subject and I felt it needed a separate article especially since the tutorial was so long already.

Fair enough, David.  This thread started when you'd only done the first article and it seemed like you'd spent a lot of time on what I consider the least important part of the calibration/integration process.  Glad to hear you have now addressed the important integration phase.

Cheers,
Rick.