Next step: sharpening

[Nocturnal]

Hi,

I think I have a decent handle on curves, histograms and noise reduction. Lots left to learn but I'm generally happy with the results I'm getting. I think it's time to add another tool to my arsenal and I think that sharpening is something I need to learn how to do. Bring out contrast in nebulae and galaxies. There was a recent comparison with various processes applied to moon images but it says they don't apply to DSOs.

Which processes should I start experimenting with?

Some of my images are here, most recent first:

http://gallery.tungstentech.com/main.php?g2_itemId=413

I have several more that are waiting to be processed. I don't expect a lot of hand holding here, just some pointers to get under way. As I understand it deconvolution should take place on raw data, before any stretches, not even a linear one. Is that correct?

Thanks,

  Sander

[Jack Harvey]

One trick I use in final sharpening is to take the image and extract the luminance using channel extraction.  Now deconvolve the lum 10-20 iterations and replace it using channel combine.  Very quick and simple.

Another method is to mask the stars and use the atrous wavelets

Good luck

[Juan Conejero]

Hi Sander,

Which processes should I start experimenting with?

I strongly recommend ATrousWaveletTransform. This example will show you many important things about this essential tool:

http://pixinsight.com/examples/wavelets/NGC7000/en.html

Specifically, the Step 3 of the above document is what you're looking for.

One trick that may save you a lot of work: use HDRWaveletTransform before StarMask. In this thread I put an example of this technique and also a link to a wavelet-based technique:

http://pixinsight.com/forum/viewtopic.php?p=3582#3582

I insist on this topic because star masks are crucial in processing DSO images, as I'm sure you know well.

Another recommendation about ATrousWaveletTransform: the "significant structure protection" thing is experimental and I don't recommend using it. It probably will be removed/substituted in a future version of this tool. With this feature I wanted to make available something that still isn't mature, which was an error.

Summarizing, you have three main options for edge and local contrast enhancement:

- As a general image restoration and edge enhancement tool, ATrousWaveletTransform. This tool is powerful, controllable and very fast.

- If you have very high signal-to-noise ratios, constrained least-squares deconvolution. Mainly for daylight, lunar and planetary images. For deep-sky only when you have a lot of signal (or limiting its action to masked regions that have high signal - you know, the zone system). There's no easy way to control ringing around stars. Also an excellent option to quickly search for a PSF that can be used with regularized deconvolution (see below).

- If you have a reasonable SNR, Deconvolution. Specifically, the regularized Richradson-Lucy algorithm is the best option for DS images. However deconvolution is difficult and requires a lot of work and accumulated experience to achieve consistent results, except in special (easy by nature) cases.

Don't forget HDRWaveletTransform. This is a fundamental tool that must always be taken into account. Many processing problems are actually high dynamic range problems, and this algorithm simply solves them.

I hope to add more solutions than confusion with this list

[Nocturnal]

Many thanks Jack and Juan! I will give all this a try and let you know how things work out.