PixInsight Stretching; Masked Stretch or Histogram Transformation ?

[russp]

What's the current thinking on the preferred process for stretching an image - I was under the impression that Masked Stretch was introduced to be a improvement on Histogram stretch but most tutorials I see are still using Histogram stretch. I've tried both on a number of images and can see contrast is reduced using masked stretch but also see references to this being recovered later. 

Thanks for any Guidance

[Filroden]

As you say, I've found Masked Stretch to reduce the contrast which required additional steps later to correct. Also, I found I sometimes would get rings around stars as it would mask the cores and not the halos. I accept that's my fault both in terms of overexposing the cores and also in not spending enough time to refine the process, but I don't see enough benefit for all that extra work compared to the simplicity of Histogram Transformation.

[wimvb]

It depends on the image and your intended result. Masked strerch is great in lifting nebulae, while keeping the stars small. It will always result in a flat image that will need further stretching. I find it excepionally good when doing lrgb, to use in the colour part. For it to work with very bright stars, you will need to repair these first. (HSV repair script + hsv combination).

You should try doing masked stretch in 4 runs, each going higher in mean level up to 0.4. It will really boost colour.

Other stretches you can use are autohistogram, and later on in the process exponential transform to lift dark detail in the background, or gamma stretch to darken background. Both with or without appropriate masks.

As always: experiment. You won't break it.

[Filroden]

7 minutes ago, wimvb said:

You should try doing masked stretch in 4 runs, each going higher in mean level up to 0.4. It will really boost colour.

I'd never thought of doing this. I will have to try it on my next process. And by mean level, do you mean the value for the Target background (defaults to 0.125)?

[wimvb]

Yes. Also, be careful about using clipping in masked stretch. Setting the clipping portion to 0 gives a completely different histogram, as opposed to the default of 0.0005 (0.05%). I usually start with no clipping, but experiment. You will at least need to adjust the black point after masked stretch, using histogram transform, if you don't use clipping. Further, you may have better control when using a preview with only background for background level determination. The same preview as the one used for bg neutralisation and colour calibration will work.

[russp]

Thanks for those pointers Wim, I'll be giving them a try!

[wimvb]

If you move away from the default parameters in Masked stretch, you can get excellent star control in images.

Try setting the clipping factor to a high number like 1 % (0.01). Use a preview that covers background and set this for background reference. Play around with the target median and number of iterations. 1% clipping may seem a lot, but it really isn't in an image with much background. Just note the clipping point in the console window.

Boost colour the same way as described before: redo MS with ever increasing target median up to 0.4, and 100 iterations per run. Then use Gammastretch with a luminance mask (inverted) or Curves transformation without a mask, to bring down the background again.