Octave Sharpening

[stepanekos]

Today I stumbled upon a new (at least for me) sharpening technique called "octave sharpening".

You can read about it in the new Fotoespresso 4/09 (in German) http://www.fotoespresso.de/, pp 15, or may be better in a (PDF-) tut of its developer, Lee Varis (http://www.varis.com/StepByStep/sharpen/Sharpen.html), pp 16.

In brief: copy the original layer 4 times, set mode to value, apply USM with an amount of 5 (!!) and an increasing radius (from bottom to top) while simultaneously decreasing the opacity (100%, 50%, 25%, 12,5%).

He works with a layer group, so may be this is easier to handle in Gimp 2.8. Or may be it is worth scripting (but I'm not a script wizzard  )

Ok, does anyone here know or even use this technique? What do you think?

[monoceros84]

Thanks for sharing this idea!
Your second link is not working...

What exactly is a value of 5? I know that e.g. Photoshop has a different scale here than GIMP and the standard USM in GIMP has a different one than the UnsharpMask-2 you can get by a plug-in...

[stepanekos]

Thanks, I've corrected the Lee Varis-link in my first posting, which should work now.

For USM amount, PS uses percent while Gimp uses standard decimal numbers. In the old Grokking the Gimp, Ch. 6.4.3, last paragraph, the author uses percent while the screenshot shows decimal numbers, see http://home.arcor.de/ulile/node63.html#SECTION001443000000000000000. So I think (or believe), this is equivalent.

P.S.: On Lee Varis' site there are some other interesting (sounding?) papers http://www.varis.com/Navigation/Steps.html

[dogwatcher]

I read the same article.. quite interesting.

I think the abscence of grouping in the (actual) GIMP won't be a showstopper here.... it's more a convenience issue than anything. If I'm not mistaken, it should be possible without groups.

[Rolf]

I think this is very interesting. Much easier with layer groups, because one can control the overall intensity.

But up to now I haven't understood the basic principle behind this. Have to read the second article and try it.

[stepanekos]

I've made a little test with an unsharp pic of a feather on the pavement. Four layer copies, all set to mode "value", USM-amount is always 5 (in words: five!), radius starts with 0.5 and doubles, opacity halves with every layer upwards.

Looking around in the internet, this seems to be the standard procedure; up to now I didn't find any discussions on the radius and so on. The fine tuning is done by masking and varying the opacity of the layer group.  I didn't use a mask but I merged down the USM-layers, so (instead of a layer group) I have one sharpening layer to play with.

Not bad at all. You can even see little drops of water on the feather (it was a rainy day). But time consuming!

[Rolf]

The result is impressive. The time can be reduced by turning this into a script. I'll try it - doing and understanding..... ;-)

[monoceros84]

But up to now I haven't understood the basic principle behind this. Have to read the second article and try it.

I think the idea is very basic: By using different radii you will grasp smaller and bigger structures to be sharpened.
I just don't get why it need to be a value of 5. It could be any other value as well - you will anyway adjust the overall opacity of the sharpening layers. It might even by different values in the different layers depending on your image. Do you want to get more sharpness in bigger structures, increase to value of the big radius sharpening. If you want to avoid noise, decrease the value at small radii...

BTW: I am not that euphoric about the result. It's very over-sharpened and the noise seems to be much more sharpened than the actual subject.

[stepanekos]

Thanks you both.

I put the opacity to 100% and USM-threshold to 0 which results in some oversharpening and noise; further on  I didn't use a mask. Btw, Varis himself calls this technique "a special technique for intense sharpening".

What's indeed confusing me are these "standard values", so I will do some tests with smaller starting radii.

[Rolf]

Just recorded the show about this - and understood how it is working while I did it. ;-)

[stepanekos]

Ok, I'm looking forward to this!

[Rolf]

I think it will be ready tomorrow evening. I have still to edit it.

[di98jgu]

This works well and can even be improved with another technique that I have seen somewhere. I don't remember where but I think it was here in the forum or in one of Rolf's shows.

After you have copied visible layers and pasted you double the new layer and set one to darken only and the other to lighten only. That way you can balance the effect. Typically dark halos looks better then white ones.

[monoceros84]

This works well and can even be improved with another technique that I have seen somewhere. I don't remember where but I think it was here in the forum or in one of Rolf's shows.

After you have copied visible layers and pasted you double the new layer and set one to darken only and the other to lighten only. That way you can balance the effect. Typically dark halos looks better then white ones.

It was in the forum but I don't remember where
Alternatively, you can use UnsharpMask2 where you are able to set dark and bright halos independently.

[littletank]

What did we do in the old days when sharpening an image was virtually unheard of? If the negative was not sharp we threw it away. What could cause the image to be unsharp? Generally either camera shake, poor focusing or poor optics. Now we have cameras with devices which counteract camera shake, we have automatic focusing we have good optics so why all this hype about sharpening the image?

Before someone leaps in to put me right I know the unsharp mask method of sharpening was developed in the darkroom as were many of the techniques we try to emulate digitally today. However, most amateurs then had either never heard of sharpening or did not have the facilities or the money needed to buy the equipment required.