New script for denoising linear monochannel images

[joelshort]

Immediately after ImageIntegration, copy the info from the Process Console that ImageIntegration created.  Then open the Script Editor tab and click the "T" to create a new text file.  Paste the copied info in the right hand panel and then save it somewhere.  Very easy, even I figured it out. 

[joelshort]

A couple more screenshots

[pfile]

on osx, if you are comfortable with the terminal, you can highlight and copy what joel has shown above and then in the shell, issue the command: pbpaste > integration.txt

rob

[mschuster]

Joel and mmirot, thank you very much, Rob thanks also. Joel, I will implement your suggestion. mmirot, your suggestion remains a wish list item.

Regards,
Mike

[mmirot]

Thanks Joel and Mike.

[dmcclain]

I know this isn't kosher from the math perspective... but I have had tremendous success going back over old images taken with an ATIK 490 OSC, already deBayered, registered, and integrated. Out of more than 100 such images, only one failed to show drastic improvement applying MURE denoising against the integrated stack image.

For parameters, I'm just putting in the measured gain and read noise of the sensor, treated as a mono sensor. So I'm probably not even using the correct noise setting, since you indicate earlier in the post that it should be the quadrature sum of read noise and dark current noise.

[mschuster]

Ok, thanks, I recommend checking carefully for artifacts and loss of detail.

Also, version 1.15 just released as an auto update. Load button loads combination count also (thanks Joel). New note that flatfield must be bias or dark-subtracted. No other changes.

Regards,
Mike

[steveyzh]

Hello Mike,
I got this error msg, the script console shew running but no result after 15 mins.

Processing script file: C:/Users/Administrator/Desktop/MureDenoise.1.15/MureDenoise.1.15/MureDenoise/MureDenoise.js
*** Error [022]: C:/Users/Administrator/Desktop/MureDenoise.1.15/MureDenoise.1.15/MureDenoise/MainViewController.js, line 568: TypeError: console.beginLog is not a function

Steve

[mschuster]

Hello Steve,

Sorry for the trouble.

This error likely is due to an attempt to run the script in an older version of PixInsight that does not support the console logging functions (beginLog/endLog/logText introduced in version 1115).

Please double check that you are running the latest version (1195 on Mac/Win, 1198 on Linux/FreeBSD). If you are running the latest version, then I am not sure what to suggest, other than to try a system reboot and a PixInsight software reinstall.

Thanks,
Mike

[gnomus]

This is a great script.  Thank you.  A friend suggested some settings for my KAF8300 chip, and these work well.  I now have a QSI 690.  Could anyone recommend some starting settings for that camera?

[mschuster]

Thank you gnomus. The QSI spec here gives typical high gain 0.16 e-/DN (default), low gain 0.36 e-/DN, and read noise 3 e- RMS in high quality mode. For the script's gain value, specify one of these. For Gaussian noise in DN units, convert read noise in e- to DN by division by gain, eg either 3 / 0.16 = 18.8 DN or 3 / 0.36 = 8.3 DN.

To find a more accurate gain value for your camera, open a frame, choose File > FITS Header, choose the frame's view, and scroll through the list to find the EGAIN keyword. Use this value for the script's gain, and again divide 3 by this value to get Gaussian noise in DN.

To find a more accurate Gaussian noise value, open two good bias frames and run the script DarkBiasNoiseEstimator. Select the two views and click Estimate. Use the temporal noise value which is in DN.

As a sanity check, both EGAIN and the temporal noise value from DarkBiasNoiseEstimator should be roughly equal to the spec values (noise after division).

Best,
Mike

PS: Some vendor drivers do not include the EGAIN keyword, or the value may be generic and not specific for the camera. In this case, to find a more accurate gain value for your camera, open two well-exposed uncalibrated flat frames and one good bias (or flat-dark) frame and run the script FlatSNREstimator. Select the three views appropriately, select 20% as the foreground quantile, and click Estimate. Use the gain value which is in e-/DN units.

[mschuster]

MureDenoise Version 1.18 soon to be released as an update with small denoising quality improvements.

• Exploits the decorrelation of noise at coarse wavelet scales.
• Refines the load variance scale calculation in some cases.

Thanks,
Mike

[Duncan]

This looks great - but, dumb question: how do I install the script?

I've been searching the forum trying to figure this out but without any success. Any help much appreciated...

[mschuster]

The script is part of the standard install, with the new version released in today's auto update.

In PI's menu bar Script > Noise Reduction > MureDenoise.

Thanks,
Mike

[Duncan]

Thanks for the super quick response! I guess I need to update my install...