Comparison of sharpening methods for Moon images

Started 3 months ago | Discussions
DavidWright2010 Senior Member • Posts: 2,747
Comparison of sharpening methods for Moon images
4

The wavelet sharpening in RegiStax 6 (nfi) is pretty much the standard technique for sharpening lunar images. But Topaz Sharpen AI (also, nfi) can do pretty impressive things, and has very few user-controls to confuse folks. (I admit I am not an expert in the use of the wavelet technique even tho I've read the tutorials on the subject.)

But we know the 'ground truth' of what we should see in a lunar photo. Here I compare an image from the Lunar Reconnaisse Orbiter (LRO) in the center, with wavelet sharpening (LHS) and Topaz (RHS) I did a little further adjustment in both of my photos.

This region of the Moon shows Hadley Rille (slightly right and down from center), which is that sinuous feature in a small valley and has a small crater touching it. This area was visited by the Apollo 15 astronauts.

The LHS and RHS images are shown at abou 600%, to match the scale of the LRO image. The match is not exact because the LRO image was (probably) taken from directly overhead, while we see this region at an angle.

Topaz makes what seems to be a sharper image, but it does so by 'eroding' features. Look at that chevron-shaped feture at the top, slightly left. Topaz (RHS) thins it out and it looks nothing like the real thing. Wavelets do much better.

Likewise, the small crater that touches the Rille is practically invisible after Topaz did it's thing, but is clearly visible after RegiStax wavelet processing.

You may be aware that Topaz has three modes of sharpening - focus, stabilize, and sharpen. The one that I showed in the comparsion above was focus, because I thought that did the best in this situation. Here are all three - focus stabilize and sharpen.

Stabilize is inventing diagonal features here (they all do, to some extent). Sharpen is different than focus, but IMO not any better.

So my opinion, wavelet processing may seem not as sharp as Topaz, but it gives a more accurate result.

David

 DavidWright2010's gear list:DavidWright2010's gear list
Sigma DP1 Merrill Sigma DP2 Merrill Sigma DP3 Merrill Sigma SD1 Merrill Pentax K-1 +1 more
nighthiker Regular Member • Posts: 485
Re: Comparison of sharpening methods for Moon images
1

Hi!

Thanks for the comparison, but I would have expected something like that. Topaz can be great, but it's a 'black box' and I wouldn't use it in this case. I'm not an expert on wavelet sharpening, but I understand the principles. It's only data driven (by the data of the actual image) and can only enhance structures which are already in that image.

What I've tried once was to put the image after wavelet sharpening into Topaz with low values for sharpening, but - I think - there was no big gain.

tradesmith45 Senior Member • Posts: 2,142
Settings Please?

Both these programs have settings that influence results.  Comparisons like this are a favorite subject on the 'net.  But they are not helpful if they do not describe the settings.

I've used Wavelets features in RT5.8 and they make a large difference in results, same for Topaz.  Topaz AI in auto modes applies NR in addition to sharpening.  Was that used here?

We neeed to know more.

 tradesmith45's gear list:tradesmith45's gear list
Fujifilm X-T2 Olympus E-M1 II Fujifilm X-T20 Fujifilm X-T100 Fujifilm X-T3 +13 more
OP DavidWright2010 Senior Member • Posts: 2,747
Re: Settings Please?

tradesmith45 wrote:

Both these programs have settings that influence results. Comparisons like this are a favorite subject on the 'net. But they are not helpful if they do not describe the settings.

I've used Wavelets features in RT5.8 and they make a large difference in results, same for Topaz. Topaz AI in auto modes applies NR in addition to sharpening. Was that used here?

We neeed to know more.

Sure.

You'll notice that I showed all the modes available in Topaz - I used the recommended settings and turned on 'do additional NR'. I thought 'focus' mode did the best on this image, but it seemed a little too 'sparkly' so in PS Elements I applied 0,3 pixel Gaussian broadening, and that attenuated the 'hot' or 'sparkly' pixels.

The wavelet processing in RegiStax was one level only, settings: NR 0.20, SH 0.120,  11.6. Afterwards in PSE I did some chroma NR (10, 50 and 80 for the 3 values in the NR edit box). Then I applied 100% sharpening 0.5 px radius, and unsharp masking 10% 4 pixels.

Any more questions?

BTW, I'm hoping that someone else will describe their wavelet technique. I've seen that tutorial where they claim a great improvement with several levels of wavelet processing, and talk about how the noise can be 'trapped' in the first level. But I never saw any improvement in using more than one leve.

David

 DavidWright2010's gear list:DavidWright2010's gear list
Sigma DP1 Merrill Sigma DP2 Merrill Sigma DP3 Merrill Sigma SD1 Merrill Pentax K-1 +1 more
tradesmith45 Senior Member • Posts: 2,142
Coparing Sharpening & Working w/ Wavelets

DavidWright2010 wrote:

tradesmith45 wrote:

Both these programs have settings that influence results. Comparisons like this are a favorite subject on the 'net. But they are not helpful if they do not describe the settings.

I've used Wavelets features in RT5.8 and they make a large difference in results, same for Topaz. Topaz AI in auto modes applies NR in addition to sharpening. Was that used here?

We neeed to know more.

Sure.

You'll notice that I showed all the modes available in Topaz - I used the recommended settings and turned on 'do additional NR'. I thought 'focus' mode did the best on this image, but it seemed a little too 'sparkly' so in PS Elements I applied 0,3 pixel Gaussian broadening, and that attenuated the 'hot' or 'sparkly' pixels.

The wavelet processing in RegiStax was one level only, settings: NR 0.20, SH 0.120, 11.6. Afterwards in PSE I did some chroma NR (10, 50 and 80 for the 3 values in the NR edit box). Then I applied 100% sharpening 0.5 px radius, and unsharp masking 10% 4 pixels.

Any more questions?

BTW, I'm hoping that someone else will describe their wavelet technique. I've seen that tutorial where they claim a great improvement with several levels of wavelet processing, and talk about how the noise can be 'trapped' in the first level. But I never saw any improvement in using more than one leve.

David

Thanks David for the additional background.  I recommend doing this valuable comparison a different way.  What you have done seems to use different software for both NR & sharpening (RegiStaxs & Sharpen AI) so we really can't say which caused the visible differences.  I'd either turn off NR in both and just sharpen or apply NR to all the individual images you submit to RegiStax for alignment and then  only sharpen by Wavelets & Topaz w/ no NR.

FWIW My experience working w/ wavelets in Rawtherapee (which may be completely irrlevant to working w/ RegiStax-:) showed me how hard it is to figure out optimum settings - number of frequency domains, how strong an adjustment to apply to each. Just too many variables.  BTW RT allows you to perform NR, contrast & sharpening to each of the Waves.   have tried it on mainly high ISO astro images.  In those, the idea of "trapping the noise" in the smallest wave is a joke.

Keep up the good work.

 tradesmith45's gear list:tradesmith45's gear list
Fujifilm X-T2 Olympus E-M1 II Fujifilm X-T20 Fujifilm X-T100 Fujifilm X-T3 +13 more
calson Forum Pro • Posts: 10,521
Re: Coparing Sharpening & Working w/ Wavelets

I have found that I get the best results when first adjusting Levels and then Contrast and after any resizing I use a sharpen application. I always get the best results with a manual adjustment of the sharpening tool but it does take a lot more time than an auto processing of the image.

-- hide signature --

"It is horrifying that we have to fight our own government to save the environment." Ansel Adams

 calson's gear list:calson's gear list
Nikon D5 Nikon D850
Keyboard shortcuts:
FForum MMy threads