What is the ultimate purpose of the ScreenTransfer Function?

[dsnay]

Hi,

As I go through Harry's tutorials (very nicely done, by the way!) I've noticed that the ScreenTransferFunction is active in some of them and not in others. Harry goes out of his way to indicate that it's active when such is the case. Is there some underlying reason why it should be active at some times and is not necessary at others? These are the kind of fundamental questions that really need to be addressed with some sort of operations philosophy documentation. You know, a basic "this is how we, the authors, envision PI being used" kind of document.

Thanks,
Dave

[Philip de Louraille]

The ScreenTransfer Function goal is to allow you to see how an effect applied to a linear image will look like without changing the linearity of the image itself.
So, for example, say that you want to denoise your linear image. Yep, it is quite dark. You have no clue as to what your levels of denoising are doing.

So you apply the ScreenTransfer Function, now you can CLEARLY see what the denoising is doing to your linear image. Now you can apply another effect to your still linear image (perhaps remove a gradient.) You end up clearly seeing what the effect of the gradient removing filter is. Yet your image is still linear.

So the Screen Transfer Function is a temporary in-memory on-screen only effect which allows you to see the faint subtle effect your linear image contains without having to expand your image histogram.

Some effects you want to apply to a linear image, others to a non-linear one. The hardest to see are the ones you apply to a linear one.

[dsnay]

Thanks. That clears things up a bit.

[dsnay]

The ScreenTransfer Function goal is to allow you to see how an effect applied to a linear image will look like without changing the linearity of the image itself.
So, for example, say that you want to denoise your linear image. Yep, it is quite dark. You have no clue as to what your levels of denoising are doing.

So you apply the ScreenTransfer Function, now you can CLEARLY see what the denoising is doing to your linear image. Now you can apply another effect to your still linear image (perhaps remove a gradient.) You end up clearly seeing what the effect of the gradient removing filter is. Yet your image is still linear.

So the Screen Transfer Function is a temporary in-memory on-screen only effect which allows you to see the faint subtle effect your linear image contains without having to expand your image histogram.

Some effects you want to apply to a linear image, others to a non-linear one. The hardest to see are the ones you apply to a linear one.

What is the benefit of not making the screen transfer changes permanent, once you've made them the first time? I would assume you'd want them to be there in the end anyway, so why not stretch the data early, make the change permanent and then save the file as an updated version? You'd still have the original data unmolested if you change your mind. Is there an advantage to performing things like gradient removal and denoising before you actually stretch the data, or am I missing something here?

As I type that it occurs to me that the screen transfer might be viewed as a preview of what you might do when you use the histogram tool. Am I anywhere close to understanding this?

Thanks again,
Dave

[georg.viehoever]

Dave,

some processes work best when you are working on linear data, such as deconvolution. A histogram stretch is usually a non-linear operation. So to see something while doing the operations on linear data, you use STF that just changes the view, not the data. When you are done with the linear data, you can do your histogram stretch, resulting in non-linear data, switch off STF, and continue from there.

Georg
 

[dsnay]

Dave,

some processes work best when you are working on linear data, such as deconvolution. A histogram stretch is usually a non-linear operation. So to see something while doing the operations on linear data, you use STF that just changes the view, not the data. When you are done with the linear data, you can do your histogram stretch, resulting in non-linear data, switch off STF, and continue from there.

Georg
 

Thanks Georg,

Okay, then that leads me to the next question. Is there some general set of guidelines regarding which functions are best done on linear data and which can wait until after the data is non-linear?

Dave

[georg.viehoever]

Hi Dave,

just search the forum for "linear data", you will find plenty of material, for example http://pixinsight.com/forum/index.php?topic=1331.0 .

Georg

[Harry page]

Hi

I know its not finished , but its got some of the info http://www.harrysastroshed.com/pixuser/glossery.html

Harry

[dsnay]

Hi

I know its not finished , but its got some of the info http://www.harrysastroshed.com/pixuser/glossery.html

Harry

Thanks Harry!
Would it be fair to say that if I'm working with a monochrome camera (I almost always am) then I would likely benefit from inserting the "Channel Combination" right after the "Image Integration" step and then proceed through the steps as you've outlined?

I realize this is not a complete guide on how to process data - no two images need the same processing - but this does sound like a pretty good plan. It does sort of line up with what I've always done using a variety of programs. However, it would be really nice to have them all in one place and I have seen for myself what the DBE tool can do for even simple open cluster images.

Dave

[Harry page]

Hi

Yes I would combine after integration , ( obviously the images have to aligned )

Then I would do Background neut and DBE I see no need to do separate DBE on the RGB


Harry

Hi

I know its not finished , but its got some of the info http://www.harrysastroshed.com/pixuser/glossery.html

Harry

Thanks Harry!
Would it be fair to say that if I'm working with a monochrome camera (I almost always am) then I would likely benefit from inserting the "Channel Combination" right after the "Image Integration" step and then proceed through the steps as you've outlined?

I realize this is not a complete guide on how to process data - no two images need the same processing - but this does sound like a pretty good plan. It does sort of line up with what I've always done using a variety of programs. However, it would be really nice to have them all in one place and I have seen for myself what the DBE tool can do for even simple open cluster images.

Dave