[RCC] NGC6888 - The Crescent Nebula Requests for constructive critique · Amos B. · ... · 4 · 433 · 2

I don't know what others think, but for me, you stretch the data as much as it can "handle" and not more. A good, deep, clean signal can handle a LOT of stretching, and often this can reveal fainter structures that most images do not expose. Not every integration can handle such a stretch though. If you stretch more than the data can handle, it'll be noisy, and depending on the quality of the data, that could look pretty bad (interpolated data will often show more mottling and worse mottling in the background than bayer drizzled or mono data.) 

What do I mean by "handle" here? There is intrinsic noise, and the intrinsic SNR of your integration, before stretching. Then there is the APPARENT NOISE. This is the noise as it presents once stretched. Consider this:




Same data, different stretches. The second only appears to have more noise, because the faint background details were stretched more than the data could really handle. 

The amount of noise an image has, and its final SNR, often has little to do with the SNR of the original signal, and a lot to do with just how you stretch. There are definite techniques to stretching, which is the conversion of a linear image into a non-linear state. You can stretch in many ways. Purely algorithmically with a variety of algorithms, or purely manually with curves, with a histogram tool, etc. Or some combination of both. You can finesse your skill with stretching, and MANAGE the various ranges of the signal, so that you can stretch those that can handle it, perhaps more than areas that cannot. Further, there are ways of stretching that can brighten data, without increasing contrast beyond the point the data can handle. Sometimes its enough to brighten, without increasing contrast (FWIW, most of the time, the basic tenant of a stretch is that you are VASTLY increasing the contrast of the signal, which initially is very low contrast, and very dark. By applying a curve, you pull up parts of that signal, increasing contrast first and foremost (making some parts brighter than others, i.e. ratio if bright vs. dark increases.)) Sometimes you may only need to increase the contrast of some parts of your image a little, while also increasing brightness (which, in comparison to increasing contrast, is more of just a shift in all the tones, vs. a change in the ratio of bright to dark.) 

There is a lot of technique in stretching, and it can leverage many different stretching tools to get the job done. I usually use a histogram stretch for my basic initial stretching, then a combination of histogram, curves, and maybe some other tools such as local contrast enhancement, to finesse my stretched image. 

If your data can handle it, then there is nothing wrong with a stronger stretch. Conversely, if stretching just makes parts of the image really noisy, then those parts probably shouldn't be stretched as much, and that may mean stretching less overall, or just getting more creative and "localized" in your stretching process. There is also nothing really wrong with making an image brighter, without necessarily increasing contrast much. In the case of the image you linked, it appears "brighter" than a lot of images of the same region. It also has improved contrast, but the region overall is brighter in general. I don't think there is anything wrong with that. The image does seem to lose a bit of the color contrasts in crescent itself, and the oxygen signal dominates the hydrogen, which is a little unusual. CERTAIN PARTS of the image might benefit from some increases in contrast, but the image overall looks good to my eye. 

FWIW, I am viewing on a monitor that has a very wide gamut, and in its native color space mode, it has a tendency to greatly enhance the colors and sometimes contrasts of images of deep space. So, I may well be seeing a more contrasty image than you, and that is something to take into account. The screen I use, an ASUS ProArt very wide gamut screen, allows me to switch to AdobeRGB and sRGB screens, which helps me see what most people are likely seeing. The image definitely has less contrast in sRGB, the brightest areas of red are not as bright or vibrant, and the blue (which overpowers the red/Ha in the crescent nebula itself) is not quite as overpowering. It helps when processing to have a screen that allows you to switch between gamuts, and has a wide native gamut itself. You can see how a wider range of potential viewers will see your images.

Thank you both for offering extremely valuable & constructive feedback. It is one thing I did struggle with on the processing side of this image - the 'over' saturation of the Oiii, and to an extent the Ha as well.

When I was working with this imaging set in PI, I was surprised at how well things were coming along, but perhaps a less extreme stretch of the data will make the ha/oiii balance better. I will go through and give it another go, and take both posts of feedback and see if there's much improvement. 

Thank you kindly.

However in regards to framing, it's something I will admit I struggle with. When it's an object such as this one, part of me of course, wants to mostly frame the crescent as the center. Though, to that end, if I can get the soap bubble to shine through a little better, I could see changing the frame of view with that in mind. Either way, it's good feedback and much appreciated.