Hi Max,
I must say your devil's advocacy is excellent
I think some exta tools such as a paint brush and layers should be included in PI.
Me too. See my previous post above. Having a paint brush available does not mean that we are going to paint our images. I explain this in my post.
"Astrophotography is documentary photography of astronomical subjects. To apply an image processing technique or algorithm selectively, pixel selections must be defined based on plausible criteria applied homogeneously to the whole image. Plausible selection criteria are based on properties of the objects (usually physical or morphological properties). Arbitrary manual selections, in general, are incompatible with this principle."
Nonsense! We selectively inhance our images non globally with masks in PI now. If I use a mask that selects a galaxy core and excludes the background and perhaps the outer spiral arms then this is non homogenous process. I can then apply sharping or other enhancemnet to a portion of image.
Homogeneity here refers to
selection criteria, not to the selection itself. Advanced processing techniques are all about building and using masks to apply transformations selectively. My paragraph refers to the "why" and the "how" of selections. You've put a good example. If your selection of the galaxy core:
- obeys to some physical properties of the object --for example, its brightness,
- or obeys to properties of the data --for example, the signal-to-noise ratio,
- and those properties are evaluated uniformly over the entire image (within the reasonable limits allowed by the tools available),
- and you don't try (purposely) to remove significant data and/or add nonexistent data (again, with the technical limits imposed by the tools and algorithms available),
- and you don't apply algorithms or techniques inappropriately, so that you only try to represent the nature of the objects (this has a lot to do with common sense and good taste),
then you are applying documentary criteria, and your procedures are honest and respectful with the data. Then I will see your image and will trust what you are communicating. The image will have the value of transmitting the essence and wildness of nature. This is the wonderful mixture of science, art and technology that astrophotography is about.
Oppositely, if you draw a lasso selection by hand to encompass what you arbitrarily decide, then you are painting and not doing astrophotography (IMO).
My point of view has more to do with the ethics of image processing and with what we want to achieve with astrophotography, both as a personal way of evolution and as a way to carry out divulgation of science and knowledge of nature.
"There are sound exceptions. For example, you can fix small artifacts or imperfections with the clone stamp or a similar tool, always with great care and knowledge of what you are removing. If in doubt, always prefer to leave the artifacts. "
Agreed.
What do have you against the paint brush tool?
It can be a lot easier to use than the clone stamp
Nothing at all. A paint brush tool can also be useful to carry out these tasks. However, I prefer the clone stamp or similar tools when it comes to act directly over the image. The reason is that the clone stamp can reproduce the distribution of noise in the vicinity of the modified areas, in a plausible and "natural" way.
Typical example : I produce a nice star mask.
It is nearly perfect but a few structures are left , such as a touch of bright nebula.
Also, perhaps a few of brightest stars were excluded because they have more structure do to diffraction spikes.
I see no major problems in using a paint brush in these particular cases, although in my opinion they are close to the limit of what I consider as acceptable. These applications are always complex and require a case-by-case analysis.
You get that touch of bright nebula or diffraction spike because the wavelet-based tools that you've used to build the mask, despite they are very accurate and powerful in their concept and implementation, are of course not perfect. Then you, with precise knowledge of these facts and limitations, apply a paint brush to remove it with great care. Close to the limit as I've said, but acceptable in this case.
I personally would prefer to fine tune the mask using purely algorithmic methods, until it can't be further improved reasonably. Only then I'd apply a paint brush to modify a mask by hand, as a last resort. And if by doing so I feel like crossing the limit, then I'd restart my processing strategy from scratch.
Better yet have the mask layed in a mode that shows that I am painting the right part of the mask.
Indeed. The new mask rendering system that comes with PI 1.5.5 fixes the current problems in this regard.
"A big blooming left in place is more honest than a nice and perfect retouch, if the latter cannot be done without destroying significant data."
It is in the original image but there is no significant data here since it is fully saturated.
Unfortunately, the things are usually not so easy; that's why good blooming removal tools are difficult to write. To remove a blooming artifact giving its surrounding areas a natural aspect, the tool must inevitably modify not only the blooming, but also some of its neighbor pixels, providing some "feathering" effect. This is what one does manually with the clone stamp. There is always the risk to damage some nearby image structures.
