Linear and Non Linear Processing

[ashworthacca]

Happy New Year to you all!

I have been into astroimaging for a couple of years now and have a grasp of the basic principals. I keep coming across the terms Linear and Non-Linear Processing and realise that I don't actually know what they mean, especially in relation to processing steps I would carry out in Pixinsight or Photoshop.

Can anyone explain what the terms mean and give examples of a linear and non-linear process in Pixinsight or Photoshop.

Thanks

[harry page]

Hi

very simple linear refers to images that have not been stretched with things like curves and hisogram transformation

This is why in pixinsight you use STF to help see whats going on

Regards

Harry

[ollypenrice]

Yes, some modifications to the image straight from camera are best performed before it's stretched. Dynamic Background Extraction, Background Neutralization, Colour Calibration, SCNR etc. Why correct these things after you've made them worse by stretching them when PI can correct them on the unstretched (linear) image? The great thing about Screen Transfer Function (known to the rest of the world as a screen stretch) is that it previews the result that your modifications will give when stretched later. While I can't live with PI for a full processing job I do find these pre-stretching functions utterly essential nowadays.

Harry despairs of me, I know...

Olly

[themos]

If the intensity of an image pixel (its numerical value) is still proportional to the number of photons received, it's still a linear image. Otherwise, it's a non-linear image. The "line" in linear is the straight line you get if you plot pixel values against photons received. The reason we resort to non-linear processing is that the dynamic range of astro photos is very large, too large to be rendered accurately on displays. In terms of numbers, for example, a nebula may have pixel values between 20 and 40, while the brightest stars have values of 60,000. So we boost the low-light regions while keeping the bright regions at about the same value, in order to bring out the detail. The nebula ends up being between 20,000 and 30,000 while the bright star goes to 60,500. We then rescale everything to a max value of 256 and end up with a nebula of 100-120 and a star with 256. The linearity has been destroyed but it looks better.

[ashworthacca]

Ah, thanks guys......great explanations.