STF readout for BackgroundNeutralization/ColorCalibration

[ruediger]

Hi all,

I noticed that the auto STF with unlocked channels often (nearly always) produces a very pleasing background and color balance, often much better than what I can achieve with background/white previews and manual BackgroundNeutralization/ColorCalibration, especially in wide angle DSLR/photolens shots.

So is there a possibility to transfer the STF values to BN/CC?

What "harm" is done to the picture when BN/CC is skipped, instead STF values are dropped to HistogramTransformation for a very small first stretch (i.e. keeping median roughly the same), continued with MaskedStretch etc.

Rüdiger

[niteman1946]

Good question.

Mark

[Alejandro Tombolini]

Hi Rüdiger, from BN Documentation:

The background neutralization task equalizes the average red, green and blue components of a color deep sky image to yield a neutral gray rendition of the sky background. Neutralization is carried out by applying per-channel linear transformations computed from a set of pixels sampled on a background reference image.

A neutral background is a necessary precondition to perform a correct color calibration. The BackgroundNeutralization tool can be applied prior to ColorCalibration or other color calibration tools and procedures.

I noticed that the auto STF with unlocked channels often (nearly always) produces a very pleasing background and color balance, often much better than what I can achieve with background/white previews and manual BackgroundNeutralization/ColorCalibration, especially in wide angle DSLR/photolens shots.

In general in DSLR images I can achieve similar result with linked channels after BN and CC than before them with unlinked channels.

These videos can help. 1 - 2

Saludos, Alejandro

[ruediger]

Hello Alejandro,
thanks for your comment.

Yes maybe I'm applying BN/CC not correctly. See the attached example. The upper right picture was the result after traditional BN/CC followed by MaskedStretch, the lower right with using my described method of transfering STF to HT (keep median), then MaskedStretch.

The BN/CC result looks "wrong" for me, with a green/brownish coloring and is overall very sensitive to where I place the black and white previews.

The STF/HT method instead doesn't need any user guidance --why should I "hint" PixInsight of what I assume should be black when the STF in full auto mode does better than me?

My wish for linear data processing is that ideally there should be no user interaction at all... let's see what I normally do in linear stage:
- background extraction: none necessary if taken under good skies and calibrated with proper flats, otherwise ABE with reduced polynomial degree (2). I nearly never use DBE and set samples manually -- normally this is an indicator that flats are not correct or stray light entered optical system
- background neutralization and color calibration -- discussed this thread
- deconvolution: could be highly automated. Why do I have to pick 30 or more stars manually, generate masks with standard procedures etc. etc. I rather would like to enter high order commands like "reduce FWHM from 2,5 to 2 pixel" or so...
- denoising: for me this is in PixInsight often an endless parameter tweaking and trial and error procedure. I would prefer e.g. an anscombe transformation and let the computer do a wavelet based denoising based on only one parameter (sigma).
etc. etc.
Creativity can start once the data is stretched!

Rüdiger

[jkmorse]

Rudiger,

I have to say, I share some of your thoughts.  I often have trouble getting things looking right with BN/CC, especially when nebulosity or other color sensitive images are in the mix.  For example, when I use it on something like M51, it tends to wash out the blues in the galaxy arms. 

I always compare the results I get from BN/CC with a separate run using a straight up Linear Fit (got that idea originally from Gerald; thanks again!).  In almost all cases the LFit image has superior color balance in both the target and the background than I get with BN/CC. 

Would love to hear what I am risking by sticking with LFit over BN/CC?

Best,

Jim