PI deconvolution tutorial

S

sym95734

Well-known member
Hi, I was wondering where I can find a good tutorial on PI de-convolution. Most precisely I would like to learn enough to make educated decisions when applying de-convolution.

I am currently working on M13 and did run PI de-convolution but I must admit it was mostly by try and error and not sure the result is correct:

http://www.stonecreekobservatory.com/cdk/m13_3_small.jpg

Some of my interest would be for example
1- How to decide if the PSF is correct for a given image
2- How to setup the correct values for PI de-convolution dialog box

I read the following material but I would like to learn more in the PI context.
http://quarktet.com/BlindDecon.html
http://support.svi.nl/wiki/HuygensDeconvolution
http://support.svi.nl/wiki/PointSpreadFunction
http://en.wikipedia.org/wiki/Point_spread_function

Thanks in advance,
Stephane

 
Hi Stephane,

A very good explanation of deconvolution is available in "The Handbook of Astronomical Imaging" (HAIP), which is supplied with the excellent AIP4WIN program.

Although it is easy, and fair, to say that PixInsight is FAR more comprehensive than AIP4WIN, it is also fair to say that HAIP provides a thorough, detailed, yet easy to follow, explanation into all of the mathematical theory behind what astro-imagers are trying to achieve.

I reckon that AIP4WIN / HAIP should be in every serious astro-imager's inventory, right alongside PixInsight and their membership of this Forum  :D

HTH
 
Thanks Niall, I do have the AIP book and completely forget to refer to it! Good reading tonight!
Stephane
 
Stephane,
in one of the processing example (don't remember which one...) there's an explanation about calculating the PSF from FWHM of your image that could be a good starting point.

andrea
 
I was wondering if there is a way to do this PI. It is easy to do in some other programs.

Max
 
Hi Max,

I was wondering if there is a way to do this PI. It is easy to do in some other programs.
Click to expand...

Indeed we lack a tool (which is relatively simple, in fact) to compute FWHM from sampled stars. This could also be automated (you know, detect a large number of stars in the image and model them as Gaussian functions).

This is a long time pending task. Added to the to-do list for version 1.6 ;)
 
Hi Andrea,

lucchett said:
in one of the processing example (don't remember which one...) there's an explanation about calculating the PSF from FWHM of your image that could be a good starting point.
Click to expand...

Yes, this is the tutorial:

http://pixinsight.com/examples/deconvolution/Gemini-NGC5189/en.html

(BTW, a large part of this tutorial is now obsolete due to improvements in the Deconvolution tool, especially the new deringing features).

Calculation of PSF parameters is described in the section entitled "The Point Spread Function", under "Step 7: Deconvolution".

In this tutorial you can also read a general description of deconvolution that I wrote. I don't know if it's clear/legible but I tried to do my best :)
 
Hi Stephan,

That M13 image is very nice!

I plan on making a video tutorial on deconvolution, at least to cover the basic procedures. The last version of the Deconvolution tool offers a lot of possibilities, especially thanks to the new deringing algorithms implemented since PI 1.5.2. With an efficient deringing, we can deconvolve images that previously were almost impossible.
 
Hi Juan,

Indeed we lack a tool (which is relatively simple, in fact) to compute FWHM from sampled stars
Click to expand...

Remember, it would be nice to be able to visualise FWHM in both the X-axis and the Y-axis, over the whole image. Think CCDInspector, if you will.

Cheers,
 
Hi Nikolay,

Juan, it's true? I am happy
Click to expand...

Yes, that's true, but more likely for 1.7 instead of 1.6. The reason is that 1.6 is becoming too "charged" of tasks now, and I don't want to postpone it too much. We'll see... anyway, the lack of a PSF modeling tool is serious and will be fixed as soon as possible.
 
You must log in or register to reply here.
Back
Top