Layers

[Juan Conejero]

Hi all,

Just a few notes, in the hope that they will clarify some misunderstandings that (IMO) I've read on this thread, with respect to PixInsight, to the future development of PixInsight, and to my personal opinions. The following are just facts about PixInsight and descriptive comments about my own way of understanding astrophotography.

- I am not against painting tools. There will be painting tools in PixInsight, including bitmap-based, vector-based and text drawing tools. Painting and drawing tools will be integrated with a layer composition system (namely LayeredComposition) based on PixelMath. In fact, several important steps toward these tools have already been made: implementation of SVG graphics, extensive XML support, and vector-based graphics (the VectorGraphics class in PCL and its counterpart object in PJSR) in recent versions of PixInsight.

- I am against hand-painted masks in astrophotography. This is just my opinion. Everyone is free to use the available tools and resources as he/she wants. But for me, a hand-painted (= hand-painted-masked) image is incompatible with astrophotography, as I understand it.

- An algorithmic mask can be used poorly. In general, any algorithm can be used well or can be used poorly. Algorithmic masks and well implemented tools are necessary conditions, but not sufficient conditions.

- Applying the result of a process --e.g. a sharpening procedure-- through a well designed, algorithmic mask that has been constructed with homogeneous criteria for the whole image is not the same thing as applying it by painting holes by hand with a brush tool. In the latter case, the user is generating false structures --or the false impression that some structures exist-- because the process is not being applied in the same way to all objects sharing a set of properties in the image. Yes, an experienced user can paint masks with extreme care, but that doesn't change the fact that manual intervention is subjective, unrepeatable, and hence unreliable. For the same reason, photography replaced visual observation during the end of the XIXth Century. In the digital era, it is interesting to note how purely subjective data handling practices are replacing automatic data processing techniques in astrophotography. Regression?

- It is extremely discouraging having to compete with hand-painted images because a hand-painted one can always be much nicer, much more APOD-prone than an image of the same object acquired with the same equipment, but processed in a documentary way. No matter how smart we can be, no matter our processing skills, the efficiency of our algorithms and techniques: nothing can compete with a brush.

- Talk about shades of gray. (i) Using the clone stamp to fix small artifacts and defects is perfectly OK, as long as no significant image structures are invented during the process. Personally, I prefer leaving a blooming in all of its glory, if while retouching it I find myself guessing nonexistent or dubious structures. (ii) Removing traces of nonstellar objects in a star mask is OK and often necessary. (iii) Increasing color saturation for a particular color, for example red to improve HII regions, can be admissible under certain conditions, but for me this is a borderline practice. (iv) Applying morphological transformations to improve star profiles is OK, as long as some common sense is used (e.g., removing existing stars is inadmissible). I could put more examples, but I think you get the idea with these. Of course there are grays.

- I am not here to make money. Damn, now that I've just ordered my new Ferrari...  8) When I design new tools or new features or development ways for PixInsight I am not thinking on how to gather more users, but on how to improve PixInsight as I see and like it. I probably should be different, but as I once said before, I am older than 19 so there's little hope that I can change.

My 0.02 €

[Nocturnal]

Hi,

good discussion on a crucial and at times dividing topic in AP. I have at times published images where I commented on the ability of the image to support the somewhat absurd level of color saturation because the data was there to do it. Increase the saturation and as long as you're not blowing out the colors it's free game as far as I'm concerned. With my note I meant to indicate that I did not 'invent' the colors. They were there and I amplified them to exaggerate small saturation differences.

In a second image (M101) I specifically said that the arms aren't as blue as I've seen them elsewhere. The color simply wasn't there in my data. Using a brush it would have been easy to colorize my image to match what we're taught to expect but I couldn't do that.

I think that's one of the crucial things I've learned here and that I see as a significant difference between the 'documentary school' style and 'old school' style. I will never use a brush to arbitrarily sharpen, color or otherwise adjust my image. At most I will allow a few clone stamps to wipe out remaining hot pixels but that action causes me to review my aquisition and calibration steps as I don't want those either. I feel that if I can describe my processing in data specific terms like "I used a luminance masked noise reduction pass to remove small scale noise in the dim areas of the image while leaving the bright ones alone" I am being true to the data and not making it better than it really is. When I increase saturation I typically do this with a luminance based mask so I can again describe the process in terms of the data. Nothing arbitrary about it except for selecting the right masking and process parameters but they apply to the entire image.

So to summarize I don't think the core of the issue is layers. The issue is arbitrary process application by the user. To take the saturation brush and boost the saturation in one arm of a galaxy while leaving the other out, to provide an extreme example. You won't ever be able to describe that step in terms of the whole image, if you're honest. You can use layers in PS in a completely documentary fashion, just like in PI. But it's not usually the way the tools get used, it seems.

As far as PI's future goes, I feel I already got an excellent value for my money. If Juan jumped to 2.0 next month and wanted an upgrade fee I'd pay it the same day. Even if I don't always get the things I want in PI or if I get things I don't care for   There is about 0% chance of us affecting the future of PS for our (evil) purposes. We have an opportunity here (as with other AP dedicated software, I'll admit) to participate in the future development of the software. Many of the things PI does in the framework and modules border on magic. I'm a pretty good software developer myself and my mind boggles thinking of how all this works. So yes, sometimes Juan goes of the deep end and completely overdesigns something that would work perfectly well if it was much simpler. But you can't deny the PI keeps moving forward and upward and I am thankful for that.

[Niall Saunders]

Hear, hear Sander - when Layers comes down the PI line, we'll be ready for it. It isn't 'that important', but it will be nice to have when it arrives.

Cheers,

[RBA]

[...] I see as a significant difference between the 'documentary school' style and 'old school' style.

I don't really agree with this classification of old vs documentary.

On one hand I agree that PI in particular offers tools that allow for novel processing techniques, and as such, more "customary" techniques may be disregarded as "old". I don't think there's disagreement here.

However, there area quite a number of imagers who are certainly "old school" who believe that anything besides a histogram stretch and maybe a light noise reduction (if at all) is considered "overprocessing". For them PI alone is as evil as PS (maybe these should be referred as "very old school"? ). Then, there's also a considerable amount of imagers who are anything but old school and who apply purely artistic techniques (some old, some invented yesterday) to their images, regardless of what the data "tells them".

And on the other side of the coin, I disagree with the concept that an astroimage only has documentary value if it has been processed only using the techniques the DSA states as valid. This enters the realms of opinion, though, and perhaps it is sufficient as long as we know what we're talking about, but I thought I'd share my opinion.

And of course, this has nothing to do with layers  8)

[Nocturnal]

Right, we could argue what constitutes 'old school' I probably shouldn't have used that term as it distracts from the differences that matter.