• Please use real names.

    Greetings to all who have registered to OPF and those guests taking a look around. Please use real names. Registrations with fictitious names will not be processed. REAL NAMES ONLY will be processed

    Firstname Lastname

    Register

    We are a courteous and supportive community. No need to hide behind an alia. If you have a genuine need for privacy/secrecy then let me know!
  • Welcome to the new site. Here's a thread about the update where you can post your feedback, ask questions or spot those nasty bugs!

Understanding the USM

I am in the process of writing a review for Chapter 6 of the new edition of Dan Margulis book, Professional Photoshop, Fifth Edition. The title of the chapter is "Sharpening with a Stiletto".
While describing his techniques Dan made a huge point that sharpening, in fact, is based on blurring...

I felt compelled to figure that out for myself.
The result were shocking - I could not believe I did this:

118514556-L.jpg


If you're interested to see how it's done - check my tutorial:
http://nik.smugmug.com/gallery/2269143

It will also become a part of my future chapter review, but I wanted to share this particular piece earlier. Consider it a Christmas present :)

HTH
 

Andrew Rodney

New member
I'm a bit confused by what this is supposed to tell us.

With respect to entering the correct values for sharpening using any technique, how are we supposed to know what values to enter? How does this apply to all the various output devices that require differing amounts of sharpening based on how they take pixels and produce dots?
 

Asher Kelman

OPF Owner/Editor-in-Chief
Sharpening is not undoing a blurred high contrast junction.

Sharpening is perceptually making natural soft changes in tonality appear sharp. It is something like taking the first derivative to find areas of switching from one piece of a pattern to another.

According to the richness in detail, the nature of tonal gradients, fidelity of the existing digtial capture and the distance in viewing the final sets of pixels, the eye will or wont be successfully tricked into believing what one is trying to create.

Margolis is an experienced writer, for sure. You are a good friend and talented too. Still, I'm not clear what the message is here.

The presentation is, however, fascinating. :)

I'm still not sure how this could guide me in what I do with any file!

Asher
 
Andrew,

Andrew Rodney said:
I'm a bit confused by what this is supposed to tell us.

With respect to entering the correct values for sharpening using any technique, how are we supposed to know what values to enter? How does this apply to all the various output devices that require differing amounts of sharpening based on how they take pixels and produce dots?

If you ever tried to tell a n00b (and I believe that you actually did:) that to sharpen one needs to blur, the reaction was probably the wide open eys at least.

This little "tute" is simply a rather academic exercise to prove this point. In our R&D dept we usually call this sort of things "a spherical frictionless chicken" :)

Merry Christmas!
 
Asher,

Asher Kelman said:
Sharpening is not undoing a blurred high contrast junction.

Sharpening is perceptually making natural soft changes in tonality appear sharp. It is something like taking the first derivative to find areas of switching from one piece of a pattern to another.

According to the richness in detail, the nature of tonal gradients, fidelity of the existing digtial capture and the distance in viewing the final sets of pixels, the eye will or wont be successfully tricked into believing what one is trying to create.

Margolis is an experienced writer, for sure. You are a good friend and talented too. Still, I'm not clear what the message is here.

The presentation is, however, fascinating. :)

I'm still not sure how this could guide me in what I do with any file!

Asher

Sharpening can be done in many ways. One could be carefully selcting the "areas of color/lightness/pattern shift" and manually applied a contrats boost, thus reaching those "first derivative" increase.

However, the point of this message, this tutorial and the whole chapter in question is how to make image look sharper WITHOUT necessarily trying to select every hair strand manually.

You have asked if this message has any practical meaning. Wel, I can honestly tell you: yes, it has.

Below is an example of application of the said technique to a rather soft shot I took recently. I didn't deviate much from the procedure, except choosing a higher (less contrasty) value in Levels to avoid quite undesirable white spots on the skin (60 instead of 15). Other than that - it's PRECISELY the steps I described. If made into action, it would take a few seconds. Mind you - no selection whatsoever.

Here's the small version:

118656340-L.jpg


If you click here you'll get to the (much larger) full size image, so you can decide whether this approach works at all for you.

Merry Christmas!
 

JohanElzenga

New member
Quite frankly, I think the result is horrible. It looks like somebody spray painted her face with some kind of plastic. No, that certainly doesn't work for me!
 

Asher Kelman

OPF Owner/Editor-in-Chief
Nikolai,

I commend your investigation of the applicability of your process. However, this is no free lunch. The are too many foci of lost of data so there is a file that can no longer be printed well at high resolution and even as is the white artifacts, say around the eyebrows.

Don't be discourage by the criticism and I know you do a lot of good in your tutorials. This, I fear might be, as yet, unready for prime time :)

Asher
 
Johan,

JohanElzenga said:
Quite frankly, I think the result is horrible. It looks like somebody spray painted her face with some kind of plastic. No, that certainly doesn't work for me!
The result you see is the result of one particular way of sharpening. First, there are many, and second, nobody ever said theire is one universal procedure how to make any image look better.
My only point was: sharpening can be achieved via blurring. The rest are side effects and, as such, should be ignored in the light of the primary target.

Merry Christmas!
 
Asher,

Asher Kelman said:
Nikolai,

I commend your investigation of the applicability of your process. However, this is no free lunch. The are too many foci of lost of data so there is a file that can no longer be printed well at high resolution and even as is the white artifacts, say around the eyebrows.

Don't be discourage by the criticism and I know you do a lot of good in your tutorials. This, I fear might be, as yet, unready for prime time :)

Asher

I hope you already know me well enough to suspect I would recommend any type of panacea, be it sharpening, contrast boost or anything else.
Every image is unique, and as such, requires a variety of respectively unique steps to make it look better. However, knowing what steps are, in fact, applicable, in what sequence and what their results could be is universally important.

As I mentioned earlier in my answer to Andrew this tutorial was done primarily for academic purposes, to prove the point that sharpening effect can be, in fact, delivered by blurring. In my answer to you, I hope, I have shown that this method can be applied to the real life images, not only to the black-and-white test tube pixel collections.
The applicability of this effect is, by all means, questionable, but I don't think you, or anybody else for that matter, would deny that it does what it claimed to do.

The reason I've chosen this particular image for the demonstration, while not necessarily wise, is very simple: I don't get too many OOF images:), and even if I occasionally do, I part with them ruthlessly as soon as see them. This was the softest image I could recall I have recently, so I've picked it. I agree, sharpening the young woman's face is normally the last thing you would do.

Having said all that, I would also like to say that I always appreciate a constructive feedback, be it positive or negative:) That's the way we learn.:)

Cheers and Merry Christmas!
 

Andrew Rodney

New member
It's easy to control what gets sharpened and not by using the image itself to build a complex mask in which to control what gets sharpened and what doesn't to a huge degree. The image has all the necessary info to build such a mask to protect shadows and avoid sharpening noise (you never want that) and to protect highlight detail which could blow out since we're affecting contrast here.

I've seen some of Dan's work on sharpening and I don't see how useful it is when you consider that:

1. You can't sharpen visually! What looks "good" (quotes on purpose since this isn't a definable matrix) on an LCD looks soft on a CRT. Zoom ratio plays a huge role in what you see. The old 'sharpen at 100%' isn't effective for all output uses depending on the number of dots produced.

2. Every output device needs it's own set of output values for USM. You can't sharpen an 8x10 print for an Epson 3800 like you would an 8x10 going out to press. Major difference!

3. Sharpening is subject, resolution and output dependant. A 50mb file of a portrait doesn't get the same treatment for value as a 5mb portrait. Does the file have a lot of high frequency or low frequency values?

There are ton's of tutorials that discuss how to make a mask out of an image for sharpening (I have but one example in a PDF on my site). But the big questions are, what values do you enter to blur the mask? What values for USM for any capture or output device? Does Dan address any of this? If not, you're flying and sharpening blind.
 

Andrew Rodney

New member
Luiz Vasconcellos said:
SoftWhile CrispImage is interesting too as it can sharpen
the lightness channel without the need to switch to LAB mode.

Use the Fade->Luminosity along with the opacity slider (optional) and you'll get the advantages of such a LAB conversion without having to jump though all the hoops and toss away a pretty sizable amount of data (depending on working space). More control too.
 

Asher Kelman

OPF Owner/Editor-in-Chief
Andrew Rodney said:
Use the Fade->Luminosity along with the opacity slider (optional) and you'll get the advantages of such a LAB conversion without having to jump though all the hoops and toss away a pretty sizable amount of data (depending on working space). More control too.

Good points! But I wonder how many people every use the opacity slider?

Asher
 

JohanElzenga

New member
Nikolai Sklobovsky said:
The result you see is the result of one particular way of sharpening. First, there are many, and second, nobody ever said theire is one universal procedure how to make any image look better.
My only point was: sharpening can be achieved via blurring. The rest are side effects and, as such, should be ignored in the light of the primary target.

Point taken, but then you shouldn't have added "If you click here you'll get to the (much larger) full size image, so you can decide whether this approach works at all for you". That implies that this is much more than just an academic exercise to prove something. It invites me to try it to see if it works for me. So I did what you suggested and decided that this method does NOT work for me.
 
Johan,

JohanElzenga said:
Point taken, but then you shouldn't have added "If you click here you'll get to the (much larger) full size image, so you can decide whether this approach works at all for you". That implies that this is much more than just an academic exercise to prove something. It invites me to try it to see if it works for me. So I did what you suggested and decided that this method does NOT work for me.

As you know, sometimes even the simplest methods works. However, much more often they don't. Besides, as Andrew pointed out, sharpening is particularly peculiar process. much depending on the image, its target size and even its target media.

It was an exercise, a part of a learning curve. Yes, it works (sometimes for somebody), but I would not suspect it to work everytime for everybody.

Cheers!
 
Top