PCL: CosmeticCorrection module

[DarrenS]

Thanks Nikolay,
real time preview is a nice addition. 

[NKV]

Module updated to ver 1.0.3.0002.
Not important changes. Log in post #1.

Best regards,
Nikolay.

[jtheios]

Thanks Nikolay,
real time preview is a nice addition. 

Absolutely. A very nice piece of work. Thanks!

[oldwexi]

Hi Nikolay!
I dont have darks with my QHY8. It has only a few hotpixels so no dark processing
needed only hotpixel removal.
Your routine looks great but does not allow me to work without darks.
I created in PI  a fake dark (black image). Now it runs,  but does
only softening of the image -  CFA checked or unchecked.

Is this again a user error or is working with no dark or fake dark a nogo?

Aloha
Gerald

[NKV]

Hello Gerald,
It's because absolute zero pixel (value 0.0) interpreted as Bad pixel.
Just add some pedestal in each pixel of the "fake dark". You can do it via PixelMath $T+0.1.
Don't forget to disable "Rescale result".

Also, for "fake dark" there are special tool "DefectMap".

Best regards,
Nikolay.

[oldwexi]

Thanks Nikolay!
Thanks for a great tool and your fast help!
Aloha
Gerald

[Philippe B.]

Dear PI develop team

This PCL cosmetic correction module is better than the script included in actual PI release  because we can see the effect by moving the cursors in preview mode.
I tested it on my win 32 bits computer but at home, all my process are done under OSX 64 bits PI version.
Is it possible to include this module in the next release version (all platform) or sooner by an update ?

Thank you very much

Cheers
Philippe
 

[NKV]

v.1.1.0
+Auto Detect hot/cold defective pixel without MasterDark. PixelMath RemoveBadPixels.xpsm in attachment.
+Show DefectivMap in Real-Time Preview. And Show Inverted DefectivMap.
+Snapshot of Real-Time Preview in real size of sources image or defined Preview. Useful for generate DefectiveMap file.

Note: you can use both method ( with and without MasterDark ) in one pass.
So, if you have fresh MasterDark - use MasterDark.
If you have old MasterDark and you think in your CCD there are new defective pixels, which not so bright in the old MasterDark  - use the MasterDark plus Auto Detect method.

Anyway, you can control all process in Real-Time Preview and via Snapshot.

Best regards,
Nikolay.

[NKV]

I found much better PixelMath method. See attachment.

[Philippe B.]

I found much better PixelMath method. See attachment.

Thanks !!!

Made some tests but on small stars (fwhm = 1.0 to 1.1 pixels), the PixelMath removes center pixel 

Best

[NKV]

Made some tests but on small stars (fwhm = 1.0 to 1.1 pixels), the PixelMath removes center pixel 

You are right. A under-sampled images is big problem. Some times very difficult to understand which pixel are star and which are hot pixel.

Anyway, I write more complex method to sort star/hot pixels. The method is very slowly. Sorry for that.
The method avoids to search a bad pixels in center of stellar object.

Please test attached PixelMach with your images.

Best regards,
Nikolay.

[Philippe B.]

Thanks Nikolay
Result is much better but less pixel defect are corrected

In fact, I really like your PCL version of CosmeticCorrection
There is source code but as I'm not programmer, it is difficult for me to include it for OSX 64 version

Maybe we could ask Juan to include it in the next release ?  Juan, if you listen to us 

[NKV]

Maybe we could ask Juan to include it in the next release ?  Juan, if you listen to us 

First, I should solve a under-sampled images problem. After that, I will ask Juan to test the code.

But current version is not ready to be in standard PI distribution. Sorry for that, and please wait.

Best regards,
Nikolay.

[NKV]

Next one.
v 1.1.0. build 2
+ STF Inverted together with image. Explanation in the thread PCL: STF in RTP
+ Calculate Hot/Cold in RTP area.
+ AutoDetection threshold slider in Average Deviation.

RTP and main Engine rewritten. Now Auto Detection method more stable detect hot/cold pixels, especially in under sampled images.
Of course  much slowly... because Auto Detection calculate local background, average neighbor and median neighbor data for every pixel. All calculation doing via MorphologicalTransformation with BitmapStructure.

Best regards,
Nikolay.

[Philippe B.]

Nikolay
I just tested your new x86/x64 version !
I really love it 

In fact, "autodetect defective pixels" only selected with sigma of 3.0 does the job at 99% 
RTP is a "plus" to see what are the best settings
I will try to do more tests on other images (but I really prefer working under OSX64  )

Very nice work, thank you very much !