Taking DSS Autosave.tif into PixInsight

[M Covington]

Esteemed colleagues,

I am using PixInsight to postprocess the output of DeepSkyStacker (originally coming from a Canon 60Da).  It works well, but the DeepSkyStacker Autosave.tif file is very dark when opened with PixInsight, as if it weren't using the whole bit depth of its file format.  I can manually adjust it with Intensity Transformations -> Histogram Transformation but was wondering if there was a shortcut.

Also, I am under the impression that if I adjust brightness, saturation, etc., in DeepSkyStacker and save the file, it doesn't have as much bit depth as it claims to.  I see banding in shaded areas, for instance.  This was described as a problem with earlier versions of DeepSkyStacker some time back; does anyone know if it's still present?

I am using the latest DeepSkyStacker, version 3.3.4.

Thanks!

[M Covington]

Partly answering my own question, by using Image Inspection -> Statistics, I see that the autosave.tif from DSS *does* use the full bit value range; what I'm up against is simply gamma correction (nonlinearization).

What is the best way to do this?  I can do a correction empirically in Histogram Transformation, but it would be handy to have a ready-made correction available that would do most of it for me.  Can I save a Histogram Transformation and re-apply it?  Is gamma correction lurking somewhere else in PixInsight?

By the way, when I open a Canon raw file directly with PixInsight, it uses 1/4 of the dynamic range.  That is easily understandable because it's a 14-bit image being opened as 16-bit.  Presumably Deep Sky Stacker adjusts for that before generating the Autosave.tif.

[M Covington]

And further answering my own question, a good "first cut" gamma correction, to be followed by minor adjustments as required by the individual picture, is:

DSS Autosave.tif opened in PixInsight:  Histogram Transformation, shadows 0, highlights 1, midtones 0.1

Canon 60Da raw file opened in PixInsight:  Histogram Transformation, shadows 0.03, highlights 0.25, midtones 0.1

Here the 0.03 reflects partial correction for bias (leaving a little to be taken out in the hand adjustments later) and the 0.25 reflects using only 14 out of 16 bits.

Query: is there an easy way for me to save these transformations to use them easily for future pictures?  I understand that there is but am relatively new to PixInsight (or rather have stayed at the beginner level for a long time).

Thanks!

[M Covington]

Also found that Auto Histogram does it well, with settings:
   Clipping off if it's from DSS, on if it's Canon Raw with shadows=0.03, highlights=1.00
   Target median MTF, 0.2

Since Auto Histogram remembers its settings, it's convenient.

[pfile]

you can set up the histogram tool, then drag it's triangle to the workspace desktop... that will create a process icon which can either be double-clicked to re-load the Histogram process, or simply dragged on to an image to apply the HT contained in the icon to the image.

i seem to remember that Autosave.tif was somehow different than just asking DSS to save the file. however, it could be that Autosave.tif is by default a 32-bit tiff which a lot of programs can't handle.

generally speaking though, any stacked image from CR2 files is going to look completely dark compared to what Photoshop/Lightroom/Camera Raw would give you... as you say it's because you need to apply a histogram stretch to the image. this is a normal part of astronomical image processing, since the calibration of the CR2 files must be done on the raw sensor data, meanning that the data is in linear form when the stack is finished.

rob

[M Covington]

Yes... Not realizing that PixInsight's MTF curve is rather different from a gamma curve, I was balking at the amount of correction needed.  But I've worked it out.  Thanks.

[M Covington]

...And wrote a tutorial that others may find useful.

http://www.covingtoninnovations.com/dslr/NewDSLR/index.html#Autosave

[pfile]

great page! i guess i'd only add that there are a number of processes that need to be run on images while they are still linear, like some forms of noise reduction, deconvolution and dynamic background extraction, to name a few. so don't fear the STF; it's necessary to do those jobs and still be able to see your data

forgetting about the STF is definitely a thing, and for that reason Juan put in a visible marker when the STF is applied - a green line appears between the tab on the left side of the image (containing the image/preview name) and the image, as a visual reminder that STF is applied. same goes with masking; the whole tab turns kind of a dark copper orange when a mask is applied and is active.

by the way that whole in-camera darks vs. calibration masters was quite a lively topic on Cloudy Nights a year or two ago. there's one guy there who absolutely insists that the in-camera system is 100% equivalent to making and subtracting a master dark from regular lights, and claims there's a statistical basis for it. however, his proof seems to be "i know a lot about statistics" rather than the actual math.

my argument against this was the same as yours/sander's; the fact that the pre-dark subtracted subs have been aligned means that the dark signal does not overlie itself in the aligned lights and so you're not averaging together the same pixels as you would when you average together the dark subs. however i suspect that it works 'well enough' because generally dithers are not huge, and there's probably some correlation in the dark signal between nearby pixels. someone on CN actually went thru the exercise of doing it both ways and the traditional way gave slightly less noise.

rob

[M Covington]

Good points, Rob.  I am adding an explanatory note to the page pointing that out.