New Techniques Page: Guide to Image Sharpening

[rick55]

ACR's masking setting should also be set with noise levels in mind. I should reword that to:
"Capture sharpening aims to address any blurring caused by your image's source, while also accounting for image noise and detail."

I see what you mean, and the new sentence is clear (to me, at least).

Thanks for another great tutorial!

Cheers,
Rick

[Steve S]

Yes you can. Maybe not the history bursh, but you can use the brush and a mask to paint in what you want to show through or hide. You can also make any kind of mask with elements, all you have to do is make a mask on a seperate layer and copy and paste it into the mask box.

[David]

Sean - Another excellent piece of work. One slight (?) point refers to paragraph 5 of capture sharpening when you talk about looking at the image at 100%. I might have emphasised this a bit more as to why 100% view is important in sharpening. Maybe you have elsewhere and I've missed it. It is a point that I had not appreciated until I read the Fraser & Schewe book.

Cheers

David

[RonH]

Thanks also from me for this excellent article.
With ref to:

Output sharpening therefore often requires a big leap of faith, since it's nearly impossible to judge whether an image is appropriately sharpened for a given print just by viewing it on your computer screen ...

... it seems to me that maybe one can use the 'eye' to judge print sharpness by screen viewing a camera pre-sharpened jpeg and comparing it with the picture you are sharpening? In my limited experience, most camera jpeg's are sufficiently sharp for the production of 'normal' prints eg those 10x15, 15x20cm prints often used in the production of family albums etc. But I guess that its a different ball-game when printing at large scale poster etc.

[The Blue Boy]

Great tutorial Sean.

Colin made me read a whole book!!!!!! :-)

Glad that you've also added the masking advice. Will point more people here to read these!

[Colin Southern]

Colin made me read a whole book!!!!!! :-)

Bet it was worth it though

[The Blue Boy]

Yeah mate, it was. :-)

[McQ]

Great tutorial Sean.

Colin made me read a whole book!!!!!! :-)

Yes, it's amazing that a book that thick can be written about a subject as seemingly simple as sharpening . Non-photographers always wonder what one could possibly fill it with. There's some aspects that I disagree with, but overall it's a great resource.

[The Blue Boy]

There's some aspects that I disagree with, but overall it's a great resource.

What did you disagree with? Now this will be interesting!

[McQ]

What did you disagree with? Now this will be interesting!

That whole silly bit about multi-step downsizing! Even their comparison was between single step (without sharpening) and multi-step (with smart sharpen), which isn't comparing apples to apples. That comparison was also surprising given a book that's otherwise technically very rigorous. In all the tests I've done, I've never seen an advantage to anything more than a single-step downsize. Yes, not all downsizing is equal, but usually this just comes down to whether you will need to add sharpening afterwards, and by how much. I think the origin of this multi-step myth came from when Photoshop used a different downsizing algorithm (pre version 6 or 7ish if I recall). In that case, there may have been a little bit of an advantage, but this certainly isn't true any more. If it were, why wouldn't photoshop just transparently downsize twice using it's internal algorithm whenever you reduce an image to less than around 50% of its original size? Who knows, maybe that's even similar to the algorithm change that they made...

[Colin Southern]

That whole silly bit about multi-step downsizing!

In a similar vein, any thoughts on multi-step up-sampling?

Personally, I just plug in whatever I need and have Photoshop do it in 1 hit - I also remember one occasion where I needed to up-sample an image and I did it in 10% increments ... with horrible results; it seemed to introduce a degree of error which was then compounded (I'll call it the "Pinocchio effect").

Just wondering if you've had any experiences with this?

[McQ]

In a similar vein, any thoughts on multi-step up-sampling?

This topic is a lot less clear cut than with downsizing. I think it really comes down to the particular algorithm that you happen to be using. With some, multi-step can probably help, but with others it might just amplify artifacts (like creating multiple sharpening halos or "ringing" along edges). Unless you're doing extreme enlargements (more than say 200-300%), I'd probably just stick with regular single step algorithms. All things being equal, I generally try to stick with the simpler solution . Even for extreme enlargements, I'd likely be considering other third-party solutions instead, such as those that use adaptive edge-based and/or fractal algorithms.

[McQ]

At the end of the "How It Works" section, the three example photos are a bit small to really see the effect: could they be enlarged?

Different, larger images have now been added.

In "Output Sharpening," I wonder if it would be better to have the original image also show the tightly cropped area that appears in the sharpened versions.

An outline of the original crop area is now shown whenever you move your mouse over any of the crops on the right.

[rick55]

Very nice, Sean. I went and read it from the beginning again. Is my memory even worse than I thought, or did you also change images at the end of "How it Works"? Wasn't that some kind of tower with filigree before?

Anyway, very well done.

Cheers,
Rick

[McQ]

Is my memory even worse than I thought, or did you also change images at the end of "How it Works"? Wasn't that some kind of tower with filigree before?

Glad you like the changes. Yes, that series of example images is also different as well. They're larger and shown at 200% to better see what's happening with the halos.

[ANSORB]

This is a little late for this thread but i think it is worth mentioning.

Light halos are almost always more objectionable than the dark ones and the are what gives an image its "oversharpened" look. The answer is to have the dark and light halos on different layers so you can control them independently. This is the mentod.

Duplicate the layeryou want to sharpen it without worrying too much if it looks oversharpened.
Change the blending mode to darken. The light halos disappear.
Duplicate the sharpened layer and change the blending mode to lighten.
This restores the oversharpened image - but with one difference. You can now control the light halos with the layer's opacity slider.

I learned this a long time ago from Dan Margulis' books. It is very simple and effective.

John

[McQ]

Light halos are almost always more objectionable than the dark ones and the are what gives an image its "oversharpened" look. The answer is to have the dark and light halos on different layers so you can control them independently.

Yes, thanks John -- that's a good point. I remember using that technique at one time, but it gradually fell out of favor because the method didn't fit in well with the typical RAW workflow. Perhaps it would work well when added to the "additional sharpening advice" section...? I think it's still worth using for particularly high contrast and/or problematic images.

[Antonio Correia]

Excuse me Sean if I am wrong but I would like to know if your statement
"Sharpening is irreversible; also save unsharpened originals whenever possible."
is still valid when using CS4 and Smart Filters on a Smart Object.

It does look like you were referring to the previous versions of CS. Am I wrong ? Sorry

Thank you

[Colin Southern]

Hi Antonio,

Smart Objects get around the problem - but - they have their own limitations (you can't do any destructive edits) which - personally - I find just doesn't suit my workflow particularly well. If in doubt I just duplicate the layer prior to sharpening.

[Antonio Correia]

Hello Colin,
Thank you