NormalizeScaleGradient Script

[rockenrock]

Hi John,

I apologize for posting all those 'eye charts' that are really hard to read.  I know you are always very busy so let me simplify my question:

If I have 20 images from light polluted skies and then add to them several more images from dark skies, could (should) the weighted total exposure time decrease? 

 

This did happened with my data:

20 x 10 min from light polluted site:

    -- Total exposure time:                3hr 20min

    -- Weighted exposure time:         2hr 7min

Add to above 4 x 5 min exposure from dark site:

    -- Total exposure time:                3hr 40min

    -- Weighted exposure time:         0hr 42min

 

The integrated images (3 combinations) show that combining all the data is (looks) better, than integrating either group.

 

Thank you for your comments.

Roger

[John Murphy]

On 1/31/2024 at 1:04 AM, rockenrock said:

Hi John,

I apologize for posting all those 'eye charts' that are really hard to read.  I know you are always very busy so let me simplify my question:

If I have 20 images from light polluted skies and then add to them several more images from dark skies, could (should) the weighted total exposure time decrease? 

 

This did happened with my data:

20 x 10 min from light polluted site:

    -- Total exposure time:                3hr 20min

    -- Weighted exposure time:         2hr 7min

Add to above 4 x 5 min exposure from dark site:

    -- Total exposure time:                3hr 40min

    -- Weighted exposure time:         0hr 42min

 

The integrated images (3 combinations) show that combining all the data is (looks) better, than integrating either group.

 

Thank you for your comments.

Roger

Hi Roger,

The weighted exposure time is not an absolute measure. It is relative to the best exposure within the set.

 

For example, suppose you have a set of images that are all very consistent - taken under the same stable conditions - but all these images suffered from significant light pollution and/or low transparency. Since the weighted exposure time is relative to the best exposure, and given that all the images were very similar, the weighted exposure time would equal the total exposure time.

 

If we take the same example, and then add a single excellent frame from a dark site, then the weighted exposure is relative to this new best frame. Since the poor frames are now scored relative to the good frame, their weighted exposure time plummets.

 

Once you understand this, by looking at both the graphs and the weighted exposure time, the results should make sense to you. In your example, you had 20 minutes of really good data. Adding the poor data to this increased the weighted exposure time to 42 minutes. We would therefore expect the results to be noticeably better, which is what you found.

 

To summarize, if you want  to compare how adding different groups of images (multiple nights) affects the weighted exposure time, add the best data first. Then the weighted exposure time stays relative to the same best frame, and the results will be easy to interpret.

 

Regards, John Murphy (NSG, PMM author)

https://www.astroprocessing.com/

Edited February 1 by John Murphy

[zerolatitude]

Hi,

 

This is regarding doing a common measurement of images with different pixel scales (eg Bin 2 and Bin 1). IIRC, some time back you had mentioned in a post that this doesn't work as PI doesn't update noise estimates with rescaling, hence the noise estimates will be wildly off.

 

Is this still the case? If so, it at all  possible to handle Bin2 and Bin1 together in NSG?

 

Thanks

Edited February 12 by zerolatitude

[John Murphy]

On 2/12/2024 at 5:34 AM, zerolatitude said:

Hi,

 

This is regarding doing a common measurement of images with different pixel scales (eg Bin 2 and Bin 1). IIRC, some time back you had mentioned in a post that this doesn't work as PI doesn't update noise estimates with rescaling, hence the noise estimates will be wildly off.

 

Is this still the case? If so, it at all  possible to handle Bin2 and Bin1 together in NSG?

 

Thanks

Yes, I believe that this is still the case.

Hence if you mix Bin2 and Bin1 in NSG, the scaling and gradient normalization will still work, but the weights calculated by NSG will be wrong because it currently relies on the PixInsight noise estimate.

There is currently no easy way around this, with or without NSG (the PixInsight weights are also based on these noise estimates).

 

I would recommend taking all images with the same binning. It is no longer worthwhile taking R, G, B images with more binning than the L image; this was only necessary in the very early days of digital photography because back then read out noise was very high. CCD and CMOS cameras have all improved a great deal since then.

 

Regards, John Murphy (NSG, PMM author)

https://www.astroprocessing.com/

 

Edited February 14 by John Murphy

[zerolatitude]

Thanks for the reply.

 

My situation is not RGB in Bin2 and L in Bin 1 - as you says its not really worth it. But I had some old Bin2 data I wanted to combine with new BIN 1.

 

Right now, my workaround is: Register substacks of BIN 1 and BIN 2 to (say) a BIN1 reference frame. Use weighting formula as number of stars, then integrate manually.

 

Not ideal, but gets the job done I suppose.