LinearFit, Color Calib, ATWT, and HDR Multiscale Transformation

[vkohli]

Hi All,

I am quite new to pixinsight and I have a processing question that I needed clarification on.

At present my R,G,and B images have been calibrated, subframe selected and drizzled. They were registered with staralignment, and dynamically cropped. Next I performed ABE/DBE, and then peformed LinearFit to match the histogram. The reference image for LinearFit was the image that was shifted more toward 'white' as viewed via histrogram transformation. After LinearFit I did a channel combination. Lastly,  I performed a background neutralization (aggregated background), and a color calibration.

My questions are:

1. I wanted to use ATWT for noise reduction. Am I correct that it is best to use ATWT while the images are in linear form? If ATWT is performed on the linear images, should I be doing ATWT on the individual channels before LinearFit, channel combination, bg neutralization, and color calib? Or, is it best to use ATWT after all of this on the combined (RGB) linear image? My method for generating non-linear is through maskedstretch.

2. Following ATWT, I wanted to do HDR multiscale transformation. Is HDR multiscale transformation meant to be done on non-linear images.  Again, should this be done on the channel combined images, OR, on the individual images before combination, OR after combination and non-linear stretching?

I hope the above is not too confusing to follow.

Thanks for the help.
Vik 

[pfile]

on #1, i don't know - ATWT is pretty well depreciated at this point and the successor tools are MMT or MLT (multi scale median/linear transform).

there are some posts in the forum detailing both of those tools. i do believe that they can be used for NR on either linear or nonlinear images, but don't quote me on that. the new hotness in noise reduction is TGVDenoise or MureDenoise (script). MureDenoise only works on mono images so in that case you have to separate. MMT/MLT/TGVDenoise can work on RGB images.

on #2, HDRMT should be done on non-linear images. it can be done on mono or RGB images. one thing to look out for is that if you have overexposed data in one or more RGB channels then the result of HDRMT can look very weird, color wise. you'll need to mask off the blown-out parts to avoid this.

rob

[vkohli]

Rob,

I'm familiar with MMT. So, is it best to perform MMT before LinearFit, and channel combination, OR, after LinearFit and channel combination. I would think the former; I would want fairly noiseless images before the combine, and then after combine have good room for non-linear stretching. But, not sure if I am right.

VIk

[pfile]

well i am not sure if there is a consensus... but i really don't know. i use MureDenoise most of the time and TGVDenoise usually later in the process after stretching if there's too much chrominance noise...

hopefully someone else can chime in.

rob

[jkmorse]

Vik,

One thing to be careful with is being too aggressive in your noise reduction in the linear stage.  The problem comes if you introduce color patterns in linear clean-up.  When you stretch them you make that problem much worse.  In the linear phase I would be hesitant to hit your images too hard with anything other than MLT and Mure Denoise (and remember, Mure only works if you are not using drizzle).  Others may take a different view but I save TGVD and MMT until the non-linear phase.

Hope that helps,

Jim

[vkohli]

Thanks Jim.