This topic comes up often in various ways and I have my own take on it. I don't think it is necessary to place imaging approaches in various labeled bins and it's more productive just to describe what is going on.
The imaging I do, and that interests me, is a general form of imaging the natural world. It could be imaging a bug or a flower or a canyon - or a nebula. When I look at it I want to have a sense I am seeing the object itself and I can interpret what I'm looking at. There is no need to resort to words like "true" in this case - it's a simple matter of knowing what I am looking at and how to interpret the scene presented.
One way to do this is to allow the imager to do anything at all to the image to make it more compelling. There are no constraints at all - and there is particular skill, both technical and artistic, in pulling this off so people like the result. I think this is by far the main way images are presented in CN - and that is perfectly fine. But I personally don't like it - not because it is 'fake' - but because it is more about what can be done with given imaging data, vs. revealing the scene itself with minimal modifications.
So what I strive to do is the very minimum manipulation of the image needed to make the object visible. This greatly simplifies the post-processing - but it means you need to have good data.
As for "doing science" - rather than label things science vs. art, you can ask if the processing done is consistent with allowed processing for a top scientific journal. That puts constraints on the image that serve a valuable purpose both for science and for general nature photography: it limits the ability to introduce artifact and maintains a sense of directly presenting raw data.
And what is allowed by these top journals? In general just global operations on the pixels - as individual measurements - including a global nonlinear stretch. And rejection of satellites in a stack or cosmic rays isn't just allowed - it's required if you are claiming to capture data from a distant object. Those other things aren't manually removed - they are rejected as outlier bad data that would introduce artifact. The result is "cleaner" - but done in a data-driven way based on a noise model.
If you have a jungle scene and a tiger is in a shadow area - it is fine to do a global stretch so the tiger is visible. But you can't highlight just the tiger in a selective manner. In this way, the dodge/burn techniques of Ansel Adams would not be allowed, but a global stretch in Photoshop would be ok.
So I can codify the various minimal stages of processing for deep sky: First you need to calibrate, align and stack many frames - and you end up with a linear image true to the data. It is probably not very visible so you need to set a white and black point - and you now have a linear image. Does it look good? Here is an example:
That is directly captured and the only processing done was to set white and black point. No cropping, no nonlinear stretch and no sharpening, etc.
Here is a similar example, but with a global nonlinear stretch to make more of the scene visible:
Here is a similar narrowband semi-false-color view of the famous "pillars" in M16:
When I do narrowband I map each color channel to the way the eye would respond to that color - and I sum them up for all channels involved. If one is very faint then I boost it with a stated factor - all in a linear manner. The pillars image is completely linear in each channel and uncropped - only black and white point were set for the final color image, and the Oiii channel was boosted by 1.5 so it is more visible.
I want to emphasize that I'm not saying this is better or worse - but it does depart from the somewhat extreme labels and assumptions made in this thread. Most astro imaging is done with no constraints at all - but that isn't what I'm doing or want to see. I want to see objects in a compelling way with the very minimum, and stated, manipulation done to the data so I know what I'm looking at. And it is largely for the same reasons such restraints are imposed by top scientific journals.
Edited by freestar8n, 14 October 2021 - 06:14 PM.