Article: Sharpening Images using Deconvolution

[iceman]

Hi all

Stuart Hodgson (wallabystu) has written an article describing the process of Sharpening Images using Deconvolution.

You can read the article on the IceInSpace Projects & Articles (http://www.iceinspace.com.au/?projects) page, or directly by clicking on the link below:

Sharpening Images using Deconvolution (http://www.iceinspace.com.au/index.php?id=63,455,0,0,1,0)

Thanks to Stuart for writing the article. If you'd like to submit an article or review to IceInSpace, please contact me (http://www.iceinspace.com.au/?contact).

[iceman]

Article uploaded.

[jase]

Good article.

Not sure if anyone has made frequency measurements between the difference of Richardson-Lucy (http://en.wikipedia.org/wiki/Richardson-Lucy_deconvolution) or Max Entropy deconvolution formulas. It is believed that Max Entropy works best on lower frequencies. Have not had the time for pixel level of analysis of late. Either can perform wonders on an image if you've got good data.

Deconvolution can wreak havoc on faint wisps of nebulosity - some disappear, thus you lose valuable data. If you're trying to maximise nebulosity I recommend staying away from heavy deconvolution. Keep the iterations light if you must - a minimalistic approach. A better approach is to layer a deconvoluted image over the background (non deconvoluted image), then blend the two in PS. That way you get the best attributes of both images - Faint wisps from the non deconvoluted image and the detail and structure of the deconvoluted image which you use a layer mask to highlight details for. This works reasonably well.

A new image processing function has recently come to light. Developed by Ken Crawford I believe. It involves a few identical images that have been through different stages of deconvolution. These images are then combined in PS which displays an incredible detail (assuming good data). I had a URL to it somewhere. Will see if I can find it.

Again, well done. The more information to share around the better.

[jase]

Ahh found it. This is the development I previously mentioned from Ken Crawford. The image processing function is called Multi Strength Deconvolution Layer Blend (MSDB). An example of this function is provided on this website. Impressive. http://www.rdelsol.com/Presentations.html

Would recommend others check out his other tutorials especially the clipping mask. Very flexible solution for narrowband colour management. Enjoy.

[bird]

I use deconvolution to "help" with planetary data, through trial and error I've found that the right settings can greatly help to both sharpen details about a certain size and reduce noise below that size.

But there's no hard and fast settings or recipes - it takes trial and error cause it depends a lot on the signal/noise in your image.

cheers, Bird

[bird]

btw Stuart I'm very happy to see an MDI version of Astra Image is coming soon - I was very sad to see that interface missed out in the Astra Image 3 release, it was one of the outstanding workflow features of the that really set AI apart from similar packages.

cheers, Bird

[kljucd1]

Hi,

I just gave the trial a go on some Moon pics...WOW!!! What a difference!!!

Regards

Daniel...

[Ric]

Thanks Stuart, a big cloud of mystery has just been lifted. I've always been reluctant to use deconvolution as I really just didn't understand what it did.

I now feel more confident to go and have a more serious play with this function.

Cheers