Flat frame normalization for OSC DSLR

[Jason Tackett]

Hi everyone,

I would like to understand a detail of how the ImageCalibration process applies flat frames for OSC DSLR images. Here is how I think flat frames are applied:

For the simple case of a monochromatic image,

1) The master flat frame is normalized by dividing each pixel in the flat frame by the mean value of all pixels in the flat frame

2) The light frame is then divided by the normalized flat frame.

Is this correct?

In my case, I am using a OSC DSLR camera with narrowband and light pollution filters which causes the values of the different color channels to be unequal in the flat frame. It seems like the normalization of the flat frame should be carried out for each color channel separately - particularly when narrowband filters are used and the signal in unused channels may be poor in the flat frame. Is there a way to ensure this happens or does it matter?

Thanks,
Jason

[mschuster]

Jason,

A year ago at least, it worked that way. But I think things like this are subject to change.

IMO if signal in some channel is poor a different normalization won't improve it, because scaling an image by a constant leaves SNR invariant. So maybe in some sense normalization does not matter.

Mike

[Jason Tackett]

Hi Mike,

Thanks for the reply. I have round-about evidence that the master flat is normalized by the mean of the entire master flat, irrespective of color channels. Plus, since there isn't an option to tell the Master Flat section of the ImageCalibration process that it is working with a CFA image, then it is likely normalizing by the whole master flat.

My main concern is if normalizing by the entire flat frame will cause any harm if I execute the ImageCalibration process with narrowband images taken with a OSC camera. I see what you are saying that if the master flat is normalized by a factor containing those unused channels then it shouldn't necessarily add noise into the good channel. I have seen it cause a strong color cast appear when applying the master flat to images taken with a CLS-CCD filter which was my initial concern. The cast is easily repairable with BackgroundNeutralization, so I guess that normalizing the master flat by the entire master flat irrespective of OSC channels should not matter even if narrowband of LP filters are used.

Of course, the alternative (discussed in an active thread right now) is to extract the relevant channels prior to calibration which makes this a moot point. If it is desirable to treat each Bayer pane independently when normalizing the master flat, I wonder if adding a CFA checkbox to the Master Flat section of the ImageCalibration process would make this easier. That way, narrowband images taken with OSC cameras could be run through ImageCalibration without the relevant channels being extracted beforehand. They would just be extracted afterward. But, if normalizing the master flat by the mean of the entire flat frame irrespective of color channels does not matter, then this too is a moot point (I'm full of 'em tonight!).

Thanks,
Jason

[mschuster]

Hi Jason,

A mean normalization across all channels might cause the good channel to highlight clip. This is something to check for.

For the narrowband case I assume you have one good channel and the others have near zero SNR? If so extracting out the good channel pre-calibration makes sense to me. Smaller files. No chance of near zero SNR channels being inadvertently used.

Mike

[Jason Tackett]

Thanks for the word of caution, Mike. I will continue the safe route as you suggested and extract the relevant channel(s) prior to calibration.
 
Best,
Jason