Sharpening – the true story

[byRo]

High Pass vs Unsharp Mask

Sharpening is a subject that comes up quite often in the threads.
There are those that defend use of the Unsharp Mask (USM) and others who say that the High-Pass method is better. Let’s look into this......
[details]

Rô

[realaqu]

really in depth tut, thanks a lot

realaqu

[Gary Richardson]

As usual Ro, you make me think more about what I'm doing. Thanks.

[Duv]

Tasty yummers lil snack.

Dave

[hpycmpr]

byRo, thanks for the tutorial. The comparison between USM and HighPass is very interesting. When I sharpen an image with lots of layers, I would create a merge visible layer on top and sharpen that layer.

Sharpening can introduce halos, noise, color shifts and contrast shifts. I find that creating an edge mask before sharpening can protect the masked and unsharpened areas from these. Testing and making tradeoffs between sharpness and these undesireables is the name of the game, and is tedious. Your suggestion to toggle between two layers is a great help.

A couple of related questions:

Some suggest that an image intended for printing should be "over sharpened". Is there any merit to this, and how much "over sharpening" should be applied?

If an image can be improved with local contrast enhancement (i.e. USM with amount of ~30 and radius of ~60), should this be done before or after sharpening?

[byRo]

Originally Posted by hpycmpr

Some suggest that an image intended for printing should be "over sharpened". Is there any merit to this, and how much "over sharpening" should be applied?

Sorry, but I'm not qualified to reply - I don't print much.

Originally Posted by hpycmpr

If an image can be improved with local contrast enhancement (i.e. USM with amount of ~30 and radius of ~60), should this be done before or after sharpening?

Local Contrast Enhancement and Sharpening are exactly the same thing - only done at different radii. I would always go from the higher radius to the lower. So enhancement first, sharpening later.

Sometimes I use "rounding" (my name for it), which is an intermediate radius band (see the 10 minute Toolbox).

So the full sequence is:
1) High Radius: Local Contrast Enhancement, or more often its inverse which is just a mixed in High-Pass High-Radius layer;
2) Rounding: Mid-range, makes people look better! (Especially the ladies) ;
3) Detail / Edge Sharpening: USM for edges or High-Pass (overlay mode) for texture.

Hope this helps,

Rô

[hpycmpr]

Thanks again byRo.

If possible, can you provide a more detailed description and user guide for your Toolbox. In particular, more on "rounding".

[byRo]

Originally Posted by hpycmpr

Thanks again byRo.

If possible, can you provide a more detailed description and user guide for your Toolbox. In particular, more on "rounding".

OK, here you go!

Rô

[rari]

byRo, i can't see your "10 minute toolbox", the link don't work.
I'd love to see it!

Please, where is this?


thanks,
rari

[byRo]

Hi there, rari,

Sorry about that. The "toolbox" was in the Resources section which, unfortunately, has been discontinued.

I had promised to post a new version (with tutorial ) but got sidetracked.
Hope I can get round to it soon.

Rô

[prizo]

Good article, but there is one error in the tutorial in regards to multiple passes.

A 500% USM is not equal to 10x50%.. its only equal to 4x50%. This is why the 10x50 looks so extreme and so bad.

The math is as follows:

100x50% = 150%
150x50% = 225%
225x50% = 337.5%
337.5x50% = 506.25%

As you can see you get over 500% by the fourth iteration.

[byRo]

Originally Posted by prizo

Good article, but there is one error in the tutorial in regards to multiple passes.....As you can see you get over 500% by the fourth iteration.

Thanks for the comment, prizo.
..and welcome to RetouchPRO.

You're probably quite right, as soon as I can get me some free time I'll look into it.

Rô

[Dm_Cher]

Originally Posted by byRo

High Pass vs Unsharp Mask

Sharpening is a subject that comes up quite often in the threads.
There are those that defend use of the Unsharp Mask (USM) and others who say that the High-Pass method is better. Let’s look into this......
[details]

Rô

USM and High Pass are very simular techniques, I think there are many exapmles where one of two works better and visa versa.

Both have simular defects: halo, noise increasing, different sensitivity in different illumination area etc.


Dmitry
ImageSkill-software for Digital Image Processing
http://www.imageskill.com