I am having some troubles using the Asinh Transformation in the latest version of Siril. The transformation does not appear to be initializing properly, most especially with regards to setting the Black Point.
The first time I use the transformation, if I set any Black Point, then the black point in the stacked image seems to be set all of the way to “right”, ---- in terms of looking at it from a Levels perspective, and the histogram goes berserk.
What I have found that does work, when first using of the transformation, is to set the Stretch Factor to 10, leave the Black Point setting untouched, and click the Apply Button. After that the transformation works correctly, allowing any setting for the Stretch Factor and Black Point.
Unfortunately, trying to apply the transformation with just the initial Stretch Factor and Black Point settings still causes the stacked image’s black point problems, to get it to work ‘properly’ i have to change the initial Stretch Factor and leave the Black Point untouched.
i am using Siril on a 2018 Mac Mini, 64GB with an external Radeon RX5700 graphics card.
Thank you for all of you hard work on bringing this software out for us!
For me this is not a bug but the normal behavior.
Once you’ve applied a first stretch, the range for the black point become wider so it is easier to play with it.
Okay, but this is not how it has worked in the past.
Before this release, it was possible to set a Stretch Factor, click the return key, watch the image brighten, then set a Black Point, watch the background darken, then click the Apply button and be able to repeat the process or move on to the next image processing technique, with the histogram widened depending on how much of a change you put into the two options.
Now, doing that really messes up the image. In fact, as I mentioned, just accepting the default 1.0 Stretch Factor and 0.00000 Black Point (click the Apply Button) will wildly mess up the image.
But we haven’t changed anything in this tool.
Are you sure that you were doing that on linear image? Because on already stretched image yes you can do what you describe.
This is the procedure I have been using, for quite a while:
After converting files to fit format, preprocessing with flat, flat darks, darks and bias files as need be, registering and stacking, then:
1.) Crop.
2.) Background Extraction (use auto stretch in the viewer is fine to be able to see the selection areas).
3.) Color Calibration Note: Use manual version for background and then white point.
4.) Remove Green Noise.
5.) Photometric Color Calibration.
6.) Cosmetic correction.
7.) Deconvolution (if needed).
Save a copy as fit file, do not overwrite the original.
8.) Stretching:
Set view mode to Asinh, then:
Asinh Transformation – multiple small steps (I outlined in the previous message how this used to work)
Histogram Transformation
Then save as TIFF for external graphics processing, such as in GIMP or Affinity Photo or whichever.
I am not really in a position right at this moment to illustrate the issue with screen grabs, (it is Friday Night where I am living) but tomorrow I can do so along with screen grabs running the exact same processing on the RC-1 release or as far back as the 0.99.8.1 release – i did not keep any older copies. This RC-2 release is the first where the problems occurred.
I loaded a stacked image, it was converted from raw files, preprocessed with flats and darks, registered and stacked. No other image processing has been done to it (to eliminate the possibility of oddness).
This is why it seems like the Asinh Transformation is not being initialized properly. From what i am seeing, the Stretch Factor must be changed on the first try, with the Black Point being left alone. After that, the command works fine. Certainly, there may be something in the setup to this point that is causing the issue. but as far as I recall, the only thing different in the earlier process is I tried the different registration methods, using Shift Registration for this one.
Thank you again for taking the time to look at this.
Ok, this is what I was saying.
Your first stretch changes everything. As the levels are streched black point has now a wider range.
I don’t think it is a bug.
If that were true, it would be okay in the first stretch, where I set the Stretch Factor to 100, hit the return key, and then changed the Black Point to 0.0010, then clicked the Apply button. However, doing that caused the issue.
In the past, we have been able to set the Stretch Factor, adjust the Black Point, and see the effects before clicking the Apply button.
That is different that what you are saying need to be done now. Perhaps the Black Point adjustment needs to be disabled on the first go.
But that also misses what I mentioned earlier, if I bring up the Asinh transformation, then on the first pass use the default Stretch Factor setting 1.0 and default Black Point setting 0.0000, just click the Apply button, this also messes up the image. The SF must be set to something else.
This is because black point is computed from the input image. I will see if it is normal (but I think yes) but when you say it was different before it is difficult to believe, as we never changed its behavior and I reproduce it with previous version.
When I first brought the issue up and demonstrated the initial settings caused a problem, along with the notion this is a change from how the previous versions worked, you replied:
After I showed how the previous versions worked, you are saying the old way is also how the new way works.
Somewhere along the line something is messing things up and it becomes apparent when applying the Asinh Transformation.
I always said that in my opinion nothing has changed in the siril behavior as we did not change the asinh code.
When I compare 0.99.10 and 1.0.0 I have NO difference.
The screenshots I sent you are done with 1.0.0 and as you can see, behavior is the same as yours in the previous post with 0.99.10.1
