New Tutorial: Deconvolution and HDRWaveletTransform with NGC 5189 Gemini South Image

[Juan Conejero]

What I'm trying to do is substitute every hot pixel with an average of its surrounding pixels. How do you think this (untested) method would compare to using the CloneStamp tool?

This could indeed work. Any automated solution, once it has been proved to be robust, is preferable to manual CloneStamp work. I encourage you to test this method and implement it as a script.

To average nine neighbor pixels, the obvious choice is:

1 1 1
1 1 1
1 1 1

but the kernel you are proposing:

1 1 1
1 0 1
1 1 1

also works, and is more oriented to the problem that you're trying to solve. I look forward to see that script  

Or, alternatively, in an oversampled image like that, why not just remove the first wavelet layer?

Despite this may work for some images, it isn't advisable as a general procedure. The problem is that by removing the first wavelet layer, we could also throw away high-frequency components that support sharp edges, even in oversampled images. This is especially true for oversampled images acquired under excellent seeing conditions.

[Juan Conejero]

We are new in this forum. These type of tutorials makes of PixInsight the best tool for the astronomical image processing

Hi Oriol and Ivette. Thank you, and welcome to PixInsight Forum  

[David Serrano]

I look forward to see that script  

Carlos Milovic told me in a private conversation about two possible pitfalls for a script like this:

- Hot pixels don't have to have a value of exactly 1.
- It would only work for not debayerized images.

To avoid the first problem, the script could use a lower value (0.95?) and regard a pixel as "hot" only if just 1 or 2 channels were saturated. This way, we avoid fiddling with stars. It would also help to explore the surroundings of the pixel.

The second problem seems a bit harder...

[avastro]

Hi Juan,
Congratulations for this very good tutorial.
I like it very much. It gave me a good example of a work flow to process astronomical images using a bunch of Pixinsight tools.
Also I finally understood why it's so important to work with linear images , the deconvolutions part of the tutorial gave me a detailed explanation of how that work, it's a must to be read.
The practical part using PixelMath is a very good complement to the PixelMath reference Tutorial.
You put a great effort to release this tutorial at the right time.
I'll need some time to digest all this stuff, but I'm now convinced that PixInsight is the more versatile soft available to process astro images.

Antoine

[Juan Conejero]

Hi Antoine,

I'm happy to hear that, and glad that you like the tutorial. For me, it has been a very special work for many reasons.

Thank you; I just hope to be able to fulfill your expectations.

[LD]

I'll add my own voice to the chorus of praise for your latest tutorials. What makes them so valuable is that you explain to non-math people like me the why and not just the how. It makes for a very rich learning experience. I have always thought that Berry and Burnell's Handbook of Astronomical Image Processing was the most valuable beginner text for the same reasons--it is far more than a manual for software but an education in processing theory and application. I can't wait for your full documentation.

Regards,
Larry Durst