HDRComposition

[gianpri]

I tried to apply HDRComposition to the globular cluster M5. I have resumed 4 short sequences (20 ", 40", 60 "and 80") and a main sequence (300 "). But when I apply the process I always get this error"Error: Inconsistent HDR composition detected (bad linear fit, channel 0 ) <* failed *> ". Maybe HDRComposition does not work with globular clusters?
Thank you
John

[drmikevt]

It works with globulars, just not with open clusters.  No, I'm kidding of course - the subject matter does not matter.

You did not describe your process.  All of the frames that you are putting into HDRC should be fully calibrated,  integrated to a master, and linear (not stretched) - is that the case?

Mike

[gianpri]

Thanks for the answer Mike. All files are calibrated, aligned and still in the linear phase. If you want you can find them here:

https://www.dropbox.com/sh/lqsi1rc5v0kpdqa/AABqrRBZYvwITOz2zyLTfsMza?dl=0

Giovanni

[drmikevt]

So, I did try your data and I can not say I fully understand what is happening or why.  Smarter people then me will have to answer. 

However, I was able to create an HDR image.  It simply would not work with all of your masters, as you said.  I was able to successfully combine the 20,40 and 60s sec frames into an HDR image.  If I tried to include the 80s, it would not work.  I also defined the area of interest, which may be important in this case.  After creating the HDR from the 3 short subs, I used that HDR image and combined it with the 300s image in HDRC, creating another HDR image.  This also worked well.  Afterward, the new HDR image was much improved over the 300s by itself, I think.  Try it and see what you think.  If I added the 20,40,60,300 frames all at once, it also did not work.

Mike

[pfile]

experiment with changing the binarizing threshold

rob

[drmikevt]

Rob - I did try that, didn't seem to make a difference, unless it had to be very low - I only went down to .5

Also, I forgot to mention that I applied LinearFit to all the frames before doing the HDR, but I'm not positive that was required. 

[pfile]

are these RGB images? there is only one binarizing threshold applied to each image, so if one of the channels is out of whack with the others, the binarizing threshold may not "work".

one strategy would be to split the images into R G B components and try doing the HDRComposition 3 times

rob

[gianpri]

I also tried with the strategy proposed by Rob, also changing the binarizing parameter,: it does not work. By now I gave up ...
John

[Andres.Pozo]

I also tried with the strategy proposed by Rob, also changing the binarizing parameter,: it does not work. By now I gave up ...
John

HDRComposition is not necessary for these images. The longest image (300 seconds) is far for being saturated, so HDRComposition is not going to make any difference.

In any case, HDRComposition works if you deactivate the "Automatic exposure evaluation" option and sort the images by decreasing exposure.

Code: [Select]

HDRComposition: Global context

Opening files:
C:/HDR/integration_300.xisf
Loading image: w=3603 h=2340 n=3 RGB Float32
C:/HDR/integration_80.xisf
Loading image: w=3603 h=2340 n=3 RGB Float32
C:/HDR/integration_60.xisf
Loading image: w=3603 h=2340 n=3 RGB Float32
C:/HDR/integration_40.xisf
Loading image: w=3603 h=2340 n=3 RGB Float32
C:/HDR/integration_20.xisf
Loading image: w=3603 h=2340 n=3 RGB Float32

HDR composition of 5 images:
 1 : C:/HDR/integration_300.xisf
 2 : C:/HDR/integration_80.xisf
 3 : C:/HDR/integration_60.xisf
 4 : C:/HDR/integration_40.xisf
 5 : C:/HDR/integration_20.xisf

* HDR image component 1 of 5

* Initial quantization:
q0 = 1.50537e+05 (17.20 bits)
q1 = 1.52092e+05 (17.21 bits)
q2 = 1.52807e+05 (17.22 bits)

* Integrating HDR image component 2 of 5
Building HDR composition mask: done
Fitting linear HDR image: done
Linear fit functions:
y0 = +0.004129 + 0.399718 * x0
y1 = +0.004150 + 0.374567 * x1
y2 = +0.004720 + 0.239484 * x2
Performing HDR composition: done

* Integrating HDR image component 3 of 5
Building HDR composition mask: done

** This HDR composition cannot integrate further images - aborting process.

* Dynamic range estimates:
q0 = 3.76608e+05 (18.52 bits)
q1 = 4.06047e+05 (18.63 bits)
q2 = 6.38069e+05 (19.28 bits)
23.647 s
It only integrates the two deepest images because the rest wouldn't add anything. Even the second only desaturates a couple of stars near the edges of the image.

[gianpri]

Thanks Andres for the clear and exhaustive explanation. Now I get it.
John