Cosmetic correction bugs

[georg.viehoever]

I played around with cosmetic correction and found a couple of problems. The version is the one distributed via the update system.

- By default, it should only set a postfix, not a prefix for filenames. Postfix seems to become the standard for PI (see calibration script for instance)

- with the settings shown in the second screenshot, the module will process nothing. I did not hit cancel or something. Output at console:

Code: [Select]

CosmeticCorrection: Global context
Loading MasterDark image:
C:/Users/Public/Pictures/Astro/2011_03_08_orion_85mm/fits/master/dark-BINNING_1-EXPTIME_120_mirrored.fit
Prepare MasterDark: Hot map. done.
Prepare thread TargetList.

CosmeticCorrection of 2 target frames:
* Using 2 worker threads

File 1 of 2
* Skipping disabled target

===== CosmeticCorrection: 0 succeeded, 1 skipped, 2 canceled =====
3.250 s
- I somehow managed to realtime preview into a state where it was busy forever, and I could not make it go away. I had to restart PI to get rid of it. I believe I saw a message about incompatible dimentsions. I could not reproduce the problem.

- The attached first screenshot shows the result of a cosmetic correction. As soon as I activate Cold Pixel Threshold with the settings in the screenshot, the result becomes bad. The GUI claims it finds no cold pixels in the data. Screenshot from left to right:
-- original light. Clearly has hot pixel in upper half
-- master dark used for cosmetic correction. Hot pixel clearly visible
-- corrected light, only Hot pixels enables
-- corrected light, hot and cold pixels enables (as shown in screenshot). Image gets worse.

Anyway: very useful tool!
Georg
 

[Philippe B.]

I tested intensively this module since several days/weeks it is on the forum. I use it for b&w CCD images.
You must be use "cold" carefuly. Also "hot" too !

The best result I get is when I use the "use autodark" setting with "hot pixel threshold level qty" low (around 200-400)  AND  the "use auto detect" option with HOT and COLD set to a level around 3 sigma.
If you increase "use autodark" levels, or use cold pixels enable here, result can be hazardous 
For sure, not all the hot spots will be corrected with "normal settings". You must tweak a little.
Also, this version works very fine with stars with 1.1-1.5 pixels FWHM

This version is MUCH more better than the script version. 99% of my hot spots are gone (maybe 50% with script version)

Now, you use CFA. Maybe there is more difficulties because of this.

cheers 

[NKV]

- By default, it should only set a postfix, not a prefix for filenames.

Fixed.

- with the settings shown in the second screenshot, the module will process nothing.

Fixed.

I believe I saw a message about incompatible dimentsions.

"incompatible dimensions" i.e. size of MasterDark not equal to size Light. Please upload your files for testing. Which OS you use and how much memory in your system?

-- corrected light, hot and cold pixels enables (as shown in screenshot). Image gets worse.

Please upload the files for testing. One MasterDark and one Light.

Best regards,
Nikolay.

[georg.viehoever]

"incompatible dimensions" i.e. size of MasterDark not equal to size Light. Please upload your files for testing. Which OS you use and how much memory in your system?

This is Win7-x64, 4 GB of RAM. Problem happened with Preview of images that I upload for the second problem. But as I said, I dont have a fixed sequence of actions to reproduce it. I may have changed the main view to other images while working with this.

Please upload the files for testing. One MasterDark and one Light.

The images are available on http://dl.dropbox.com/u/31016791/TestCC.zip (114MB zip). One calibrated light milkywaynorth9_c.fit, the master dark dark-BINNING_1-EXPTIME_20.fit, the light cc'ed with HotPixelThreshold only milkywaynorth9_c_cc.fit, and the light cc'ed with Cold Pixel Threshold as well milkywaynorth9_c_cc_2.fit which shows the problem , all CFA. To reproduce the problem, you have to move the Cold Pixel Threshold Slider from left to right and back before apply.

Two additional issues:
-Occasionally I get a memory access problem when I activate/deactivate CFA. Sometimes, CFA cannot be deactivated after this.
-The hot/cold pixel sliders may be too limited in their range, I think Harry reported this as well. Maybe provide log-scale sliders here?

Thanks for this fantastic tool!

Georg

[NKV]

To reproduce the problem, you have to move the Cold Pixel Threshold Slider from left to right and back before apply.

Thank you! The bug found and fixed.

-The hot/cold pixel sliders may be too limited in their range, I think Harry reported this as well.

This is same bug as above. Fixed.

Latest version 1.2.0.5