224 replies to this topic

[PiotrM]

Denoise, upscale, automate, align with derotation. Those are the core features nowadays I would say. There is a lot of sharpening options.

[925chicago]

First thank you for an excellent program! I appreciate the time and effort you have put into this over the years.

 

Regarding any "wants", would it be possible when saving schemes to include the functions like RGB alignment, rotation, etc. I find myself continually repeating these manually for the multiple stacks.

[Foehammer]

Corr, 

 

First of all, THANK YOU for the excellent work you have done with Registax. It has been a fundamental software for planetary imaging. I remember using previous versions (was it 3? or 4?) on my Celeron, 256mb ram, square-screened laptop to process toucam images back in the day. Things have certainly come a long way since but Registax is still the #1 option for sharpening stacks!

 

The number one function I would absolutely LOVE to see implemented is - as others have stated - batch processing: A way to apply wavelet sharpening (primarily) and stretching / color balance to the multitude of stacks we produce. I now have to apply wavelets to 124 consecutive stacks from last night's imaging run. Schemes work great in applying the same wavelet settings, but having to do this 124 times (sometimes more!) is tedious at the least. Please please please automate this process! I would gladly support a paid version of registax which includes this feature as it would save me so much time! 

 

Thank you for considering it, and thank you for this great software! 

 

Agapios

[theaberrator]

Corr, 

 

First of all, THANK YOU for the excellent work you have done with Registax. It has been a fundamental software for planetary imaging. I remember using previous versions (was it 3? or 4?) on my Celeron, 256mb ram, square-screened laptop to process toucam images back in the day. Things have certainly come a long way since but Registax is still the #1 option for sharpening stacks!

 

The number one function I would absolutely LOVE to see implemented is - as others have stated - batch processing: A way to apply wavelet sharpening (primarily) and stretching / color balance to the multitude of stacks we produce. I now have to apply wavelets to 124 consecutive stacks from last night's imaging run. Schemes work great in applying the same wavelet settings, but having to do this 124 times (sometimes more!) is tedious at the least. Please please please automate this process! I would gladly support a paid version of registax which includes this feature as it would save me so much time! 

 

Thank you for considering it, and thank you for this great software! 

 

Agapios

Hi,

 

I will be looking at an fully automated way of applying a set of operations on an image and also allow the user to process a set of images that all need the same operation done.

Be aware that this is NOT going to be another RegiStax, its a separate application purely aiming at sharpening and other default image-processing tasks people currently mainly do in RegiStax.

It would be nice if someone in the forum would take up the task to get a clear list of features that need to stay, need to go (no use) and need to be added. As I am not much into imaging myself anymore

it would be good to have a clear view also on the image-formats people want to use. As we are going to deal with stacked-images mainly (thats the core of this application) its my guess these are in normal

formats like PNG, FITS. But maybe we nowadays also have other formats that allow 16bit/channel storage.

 

Anybody that wants to take this up ?

I am currently allready having a first running basic setup with 3 very basic wavelet-layers and only in BW. Since I am planning to develop for Linux, Windows and Mac the first thing
I want to do is test if a simple application will indeed work on all these platforms as expected. Please dont read this as "the project is ready tomorrow"

 

cheers

Cor

[Borodog]

I'll take a stab.

 

Version 1:

 

Stay

Wavelets

Denoise

RGB Align (by the way, please enhance this to sub-pixel alignment instead of only integer pixel alignment)

RGB Balance

Histogram (black point and white point) adjustment

Gamma/Curves

Color mixing, B&W/luminance extraction

Rotation, flipping, resizing

Saturation adjustment

View Zoomed

Line Graph

 

Go (subject to other user's feedback; does anyone use these?)

Stacking

Denoise/Deringing - For me the denoise on the layers is sufficient, and I try not to sharpen to the point of ringing if possible. Plus I never had much luck with the deringing when I did try it. Anybody use it?

Wavelet Filter Panel - Anybody actually use these to change the default wavelet filter?

Masking - Anybody use it? This accomplishes something like what multiscale sharpening does automatically.

Compare Panel - Anybody use it?

 

Add

Batch processing

Scripted processing

Command line processing

Vibrance

Channel by channel histogram adjustment - i.e. set black and white points channel by channel; same with gamma/curves

Non live update option (i.e. "execute" button does all processing)

Rotational alignment - I believe what PiotrM is after is a way to automatically align planetary images to remove field rotation automatically without the full hassle of WinJUPOS. This would probably involve calculating the first and second moments of inertia of the image to determine orientation and then rotating the image to some orientation or perhaps to match a specified reference image. Piotr, please correct me if I am wrong here.

Edit: A better way to limit wavelets near hard edges (opposite of multiscale sharpening)

 

I am not the arbiter of what stays or gos or gets added, so please provide some feedback if I've missed something or included something in the Go list that you actually use.

Edited by Borodog, 28 October 2022 - 10:52 AM.

[Sunspot]

Borodog,

 

That is a great start. Perhaps we could make this a new thread. I'm glad you added the non-live update.

 

Paul M.

[theaberrator]

Borodog,

 

That is a great start. Perhaps we could make this a new thread. I'm glad you added the non-live update.

 

Paul M.

 

 

I'll take a stab.

 

Version 1:

 

Stay

Wavelets

Denoise

RGB Align (by the way, please enhance this to sub-pixel alignment instead of only integer pixel alignment)

RGB Balance

Histogram (black point and white point) adjustment

Gamma/Curves

Color mixing, B&W/luminance extraction

Rotation, flipping, resizing

Saturation adjustment

View Zoomed

Line Graph

 

Go (subject to other user's feedback; does anyone use these?)

Stacking

Denoise/Deringing - For me the denoise on the layers is sufficient, and I try not to sharpen to the point of ringing if possible. Plus I never had much luck with the deringing when I did try it. Anybody use it?

Wavelet Filter Panel - Anybody actually use these to change the default wavelet filter?

Masking - Anybody use it? This accomplishes something like what multiscale sharpening does automatically.

Compare Panel - Anybody use it?

 

Add

Batch processing

Scripted processing

Vibrance

Channel by channel histogram adjustment - i.e. set black and white points channel by channel; same with gamma/curves

Non live update option (i.e. "execute" button does all processing)

Rotational alignment - I believe what PiotrM is after is a way to automatically align planetary images to remove field rotation automatically without the full hassle of WinJUPOS. This would probably involve calculating the first and second moments of inertia of the image to determine orientation and then rotating the image to some orientation or perhaps to match a specified reference image. Piotr, please correct me if I am wrong here.

 

I am not the arbiter of what stays or gos or gets added, so please provide some feedback if I've missed something or included something in the Go list that you actually use.

Hi Borodog,

 

Nice that you have taken up the gauntlet. For the more exotic wishes like "vibrance" and derotation I would be please if someone could at least give me some pointers to actual code that can do this. Can be in most languages as I do speak pascal, C++, Java and python but not all of it fluently

 

To add a few I allready have on my list, "non colour-changing" sharpening. As you know when we sharpen colour images in RegiStax all the channels use the same set of gaussian kernels. But we can do this also in a different fashion. We can simply take the luminance of the image (any mixture of the channels) and use that to reconstruct the features and merge that back with the original colourset. That way colours get less

distorted by sharpening. And under the hood I also have added a new sharpening module that uses bilateral sharpening, this is WIP (work in progress) but potentially should allow crisper edges with less ghosting/edging of hard contrast parts of the images.

 

Thats it for today, keep the ideas coming !

 

cheers

Cor

[RedLionNJ]

Since Registax has been such an important piece of astro software over the years, I do think this merits its own (pinned?) thread.  Borodog, would you like to start one and I'll pin it (if we think that's a collectively good idea)?

 

When (if?) this proposed development comes to fruition, I'm sure it will be just as important a contribution to both the hobby aspect and the science aspect of planetary imaging (particularly Jupiter) as Registax has been.

 

Thank you, Cor & Borodog!

[Borodog]

We could do that. Not sure why we need a thread other than this one though?

 

For the more exotic wishes like "vibrance" and derotation I would be please if someone could at least give me some pointers to actual code that can do this.

 

theaberrator,

 

I went poking around and to my surprise discovered there is no standard definition of "vibrance" in color images. There seem to be two main definitions and they are definitely not the same. In one of them, enhancing the "vibrance" of an image is focused on bringing the saturation of the mid tones up specifically, rather than the highlights or the shadows. This seems like it would be particularly nice for astrophotos because it tends to leave the highlights alone and also not bring up color noise in the shadows. In the other approach, increasing "vibrance" simply targets increased saturation for areas with lower saturation; in other words it's like a saturation transform that enhances output saturation as a function of input saturation, without regard for value. I'm not sure how that would behave in astrophotography. I believe the "enhance vibrance" tool I am used to, the one in Astra Image, works like the former. I can tell you how *I* would implement such a tool, but I can't promise it would be how any other tool in particular might do it.

[Baikalic]

Amazing the contributors there are in this community, decade old software doesn't usually get a second life. Looking forward to watching any progress.

 

One feature seems like someone hasn't mentioned (maybe I'm the idiot here and who reads software manuals right?? ) is that I've missed a quick zoom feature on the main image panel. I understand I can use the view zoom button and look at a magnified portion of my image but it seems the ability to zoom in via hotkey-modifiers like you can do in PS and GIMP (e.g., Ctrl+mouse middle scroll up/down) would be a logical next qol improvement.

 

Thanks again for all your work (previous and future)!

[Borodog]

We could do that. Not sure why we need a thread other than this one though?

 

 

 

 

theaberrator,

 

I went poking around and to my surprise discovered there is no standard definition of "vibrance" in color images. There seem to be two main definitions and they are definitely not the same. In one of them, enhancing the "vibrance" of an image is focused on bringing the saturation of the mid tones up specifically, rather than the highlights or the shadows. This seems like it would be particularly nice for astrophotos because it tends to leave the highlights alone and also not bring up color noise in the shadows. In the other approach, increasing "vibrance" simply targets increased saturation for areas with lower saturation; in other words it's like a saturation transform that enhances output saturation as a function of input saturation, without regard for value. I'm not sure how that would behave in astrophotography. I believe the "enhance vibrance" tool I am used to, the one in Astra Image, works like the former. I can tell you how *I* would implement such a tool, but I can't promise it would be how any other tool in particular might do it.

Probably best to skip it if nobody but me is interested. It's only subtly different from saturation I guess.

[cytan299]

Hi Cor,

   Kudos for writing Registax, which despite its age, it is still the de facto standard for processing planetary images. I'm glad that you'd be adding a Mac version because right now, I'm running Registax under Wine which is not the most ideal situation.

 

I'll be looking forward to testing a new native version of Registax on my Mac.

 

cytan

[otoien]

Cor, your software is amazing and I use it every time. I like the hold wavelets function but I need to reset the histogram and white balance for each image in a batch. I do use another program (Astra Image) for deconvolution. But if nothing changed in Registax I'd continue to use it for years to come!

+1 to the amazing software, and yes, it would have been a big plus if also all the other settings (color adjustment, contrast, levels etc.) could be saved and recalled in a similar way to the wavelet settings, preferably together with the wavelet settings as an option so that all settings to process a similar series of frames could be easily recalled with just one settings file. 

Edited by otoien, 27 October 2022 - 09:45 PM.

[RedLionNJ]

 

I am not the arbiter of what stays or gos or gets added, so please provide some feedback if I've missed something or included something in the Go list that you actually use.

The desire for some way to tame (or not enhance so much) the "Mars rind" was included earlier in this thread, but appears to have been dropped from the list?  This could potentially be as simple as an automatic reduction in the wavelet intensity near a severe, high-contrast gradient, like the planet's limb. Maybe a way to alter the intensity of the effect, scale of 0-10.

[theaberrator]

Hi,

 

Ive made some progress (see attached image), this is the current setup (pure demo in terms of layout etc just to get default functionality working). It can load images (currently only PNG), can set wavelet-filters both sharpening and blurring. Linking layers is active.  New option is filterweight, this balances the wavelet filter a bit towards the "original" more but most of the time has little effect. Ive added this as an option not to break compatibility with the current setup.

You can set the working order to manual or automatic. In Manual you will need to press refresh to recalculate. And ive added "highlights", this a toggle option that will show which areas are currently too dark or too bright to fit into a 16bit dataspace of a png-image. They thus show where things are getting overstretched probably.  The internal datasets are as in RegiStax not limited to 16bit integers.

This all works fine under linux and the program also runs under windows 64bit but ... in windows 64 I cannot get a 16bit png file into the memory as I wish. So I wont release a trial DEMO until that also works as planned.

For Mac I even cant test myself but I am sure someone will jump in.

 

For those on linux the tester is at https://github.com/CorBer/waveSharp (project1.zip)

 

cheers

Cor

Attached Thumbnails

• 

Edited by theaberrator, 28 October 2022 - 06:08 AM.

[camman]

Hi Cor - awesome to see this post from you!

 

Registax has been (and continues to be) a wonderful piece of software for sharpening and is perfectly adequate for color adjustments and mild denoising.  There are many times (admittedly with a good stack out of AutoStakkert) where I have needed no software other than Registax to complete my planetary processing.

 

One wavelet-related feature which would be nice to have is a way to "tame the rind" (the hard, false edge which often occurs on the sunlit limb of Mars, in particular) while leaving all other detail visible. When processing Mars, this is the only step I look to other applications to accomplish.

 

My very best wishes and a thousand thanks for your work,

 

Grant

I agree if a circle could be drawn around mars, and less sharpening was applied in that region. You could sharpen what is detail, and leave the nasty false edge smooth. As wavelets are applied the rind starts appearing, and the edge starts to go over bright. So of course, if a circle could be put around the the edge in question. where minimal sharpening is applied. The edge would look far better and more natural. This could be useful for jupiter too. Though to a less degree. 

[theaberrator]

I agree if a circle could be drawn around mars, and less sharpening was applied in that region. You could sharpen what is detail, and leave the nasty false edge smooth. As wavelets are applied the rind starts appearing, and the edge starts to go over bright. So of course, if a circle could be put around the the edge in question. where minimal sharpening is applied. The edge would look far better and more natural. This could be useful for jupiter too. Though to a less degree. 

I think it will be easier to get the software simple detect the planetedges, there are many options to do so. Based on that you could make a drop-off function that gradually decreases the effect of the wavelets. The masking feature in R6 was created with that in mind. But I am sure we can do better now. In 10yrs the no of algorithms for image processing has also increased further.

 

Cor
 

[Great Attractor]

Cor,

 

Will you be also releasing the source code? I strongly encourage it for the reasons: contributions and long term maintenance.

 

Like everyone here, I'm joining the big "thank you" for R6. I had been using it for a few years, but then started doing more time lapses and the lack of convenient batch processing was the main reason I wrote ImPPG (also, some already mentioned QoL improvements, like smooth zooming and scrolling, more tone curve editability). Now, were RegiStax open-source back in 2015, I'd have simply (with your approval) added all this to RegiStax instead.

 

And it's great that you're going cross-platform (I've checked recently that R6's installer isn't working under Linux+Wine, though the program itself works once copied over from Windows).

Edited by Great Attractor, 28 October 2022 - 08:47 AM.

[cytan299]

Hi,

 

Ive made some progress (see attached image), this is the current setup (pure demo in terms of layout etc just to get default functionality working). It can load images (currently only PNG), can set wavelet-filters both sharpening and blurring. Linking layers is active.  New option is filterweight, this balances the wavelet filter a bit towards the "original" more but most of the time has little effect. Ive added this as an option not to break compatibility with the current setup.

You can set the working order to manual or automatic. In Manual you will need to press refresh to recalculate. And ive added "highlights", this a toggle option that will show which areas are currently too dark or too bright to fit into a 16bit dataspace of a png-image. They thus show where things are getting overstretched probably.  The internal datasets are as in RegiStax not limited to 16bit integers.

This all works fine under linux and the program also runs under windows 64bit but ... in windows 64 I cannot get a 16bit png file into the memory as I wish. So I wont release a trial DEMO until that also works as planned.

For Mac I even cant test myself but I am sure someone will jump in.

 

For those on linux the tester is at https://github.com/CorBer/waveSharp (project1.zip)

 

cheers

Cor

If there is source code, I can compile on my Mac.

 

cytan

[leoyasu]

+1 on batch processing. I know RS6 has a workaround for this but for some reason it add some noise. +1 on also being a key piece of soft still for modern planetary imaging and overall all I'm good with all the other features.

Edited by leoyasu, 28 October 2022 - 07:38 AM.

[theaberrator]

Cor,

 

Will you be also releasing the source code? I strongly encourage it for the reasons: contributions and long term maintenance.

 

Like everyone here, I'm joining the big "thank you" for R6. I had been using it for a few years, but then started doing more time lapses and the lack of convenient batch processing was the main reason I wrote ImPPG (also, some already mentioned QoL improvements, like smooth zooming and scrolling, more tone curve editability). Now, were RegiStax open-source back in 2015, I'd have simply (with your approval) added all this to RegiStax instead.

 

And it's great that you're going cross-platform (I've checked recently that R6's installer isn't working under Linux+Wine, though the program itself works once copied over from Windows).

Yep, I will share the project on GitHub.
 

[theaberrator]

If there is source code, I can compile on my Mac.

 

cytan

The source will be based on Freepascal and I am developing this in Lazarus. I was first opting to use C++ but developing a proper GUI in C++ seems to be a hurdle. The reasoning to choose C++ is the speed of the code, but we might still use a C++ library for a few dedicated tasks.

 

Cor

Edited by theaberrator, 28 October 2022 - 09:22 AM.

[Borodog]

I edited the list to include a better way to limit wavelets near hard edges. Note that this is the exact opposite of standard multiscale sharpening, which looks to sharpen near edges but limit sharpening in flat regions.

[RedLionNJ]

I edited the list to include a better way to limit wavelets near hard edges. Note that this is the exact opposite of standard multiscale sharpening, which looks to sharpen near edges but limit sharpening in flat regions.

Thank you!  Yes, I realize it's counter to the way sharpening is supposed to work

[RedLionNJ]

To add a little more to the desired scripting functionality (at least in my mind):

 

A command-line invocation with optional parameters such as source files (including wildcards), settings file (operations to apply) and output file type would be awesome.

 

e.g. 

 

Appname ./p25/*.tif  settings.ops  PNG

 

would execute the app for all TIF files in the p25 subdirectory, using the settings in "settings.ops" and saving the results as PNG files in that same directory.

 

I'm sure some more concerted thought could come up with a better way to do this, too.