Chapter 6, Professional Photoshop, 5th Edition
Nikolai
Registered Users Posts: 19,035 Major grins
Professional Photoshop
Fifth Edition
By Dan Margulis
Chapter 6: Sharpening With a Stiletto
Review by Nikolai Sklobovsky,
www.photosocal.com
Foreword
First of all, we need to understand that perceived sharpness has often
nothing to do with our sharpening techniques.
Some researches shows that in many cases "the number one" winner can be
simply a higher resolution image, closely followed by increased
contrast, followed by adding a small amount of noise (but not too much).
The effect of the conventional spatial sharpening (as in PS Sharpen
command) only follows in the fourth place. Nevertheless, since this
chapter is about sharpening we are going to consider various aspects of
it, however subtle its results could be on the perceived image sharpness.
Right after Mr. Margulis’ gem went out of PeachPress publishing and into
the world, Adobe released a new beta version CS3 of its flagship
product, Photoshop. Amongst numerous improvements and innovation there
was one that would definitely affect the whole chapter, if not the whole
book: I’m talking about the smart filters. Starting from this version,
an operator can apply any filter – including sharpening - as an
adjustment layer of some sort and later return to it to modify its
properties, visibility, opacity, mask and position in the layer stack.
Yes, they are not perfect and do have an effect on the PSD file size
(you can read a nice technical insight on them at John
Nack’s blog), yet I think it’s a huge step in a right direction.
However, while this opens a whole new dimension in a retoucher’s life,
most of Dan’s techniques remain valid and, naturally, very sharp.:-)
In the Beginning
As it’s easy to see from the title, the chapter 6 is mostly
devoted to the sharpening techniques.
Through the whole chapter Dan’s motto seem to be “sharpen as much as you
can get away with, and then some”. This logic based upon rather
down-to-earth fact that while many prints across all the US print houses
were rejected for being “too soft”, few, if any, were returned for being
“too sharp”.
This does not necessarily mean that over sharpening cannot hurt. However, our
generally photographically uneducated customers may not necessarily
recognize the issue and simply state that they don’t like the image
because “it makes them look bad” (or something else to that extent).
From the very start Dan suggests to use one and only one PS sharpening
tool, obscurely named USM (Unsharp Mask) and ignore all the rest, even
though “the rest” being nicely named Sharpen, Sharpen Edges, Sharpen
More, and Smart Sharpen. His reasoning, as in the rest of the book, is
simple: the level of control this tool gives you is unsurpassed by the
rest, which give you none or much smaller amount of it.
When
Before diving into the fine details of several sharpening
scenarios, Dan emphasizes a very important concept he suggests to follow
religiously: sharpening should be the very last step in the whole
retouching process, the only thing that can happen after it is printing.
And if somebody else is gonna work on this image after you done with it,
he suggests to leave the sharpening out of the picture altogether.
With the new PS CS3 it seems that you can put your sharpening into those smart
filter layers at any time and at least get an idea of what you gonna
have. Later you can move this sharpening group to the top of the layer
stack and even disable it if you need to work on something else. You
will always be able to return to it and adjust the radius, threshold,
amount and other things.
Where
Another preliminary part of the chapter deals with the question
of “where”, as in “where in the picture should we apply our sharpening”.
The answer is not always straightforward and usually image specific.
Unless you going to “take a trip to Selectionville” – the concept Mr.
Margulis quite obviously dislikes a lot – you should look very carefully
to what you’re gonna sharpen.
I personally don’t see anything criminal in “selective” sharpening, especially if one
can do that in the matter of seconds as opposed to spending half-an-hour
or so looking for particular color/channel combination. But I guess
that’s just me…:-)
However, Dan immediately provides some general ideas that can help you
to answer this important question.
First thing he suggests we need to do is to find out if there is a
particular dominant (not to be confused with primary) color we want to
sharpen – or to avoid sharpening at all costs. Green foliage of bushes
and tress, red walls of canyons, yellow sands of the desert, deep brown
of the tree bark are all typical examples of such dominant color we
would like to sharpen. Gentle pink of a female face or deep blue of the
skies are the examples where we would like to have as little sharpening
as possible, since otherwise the first would be covered with nasty pores
and the latter simply become very, very noisy.
Obviously, not all images would have such a “color boss”. Some would
have several (in different parts of it), some would have none (or simply
too many to pick one). However, in many cases we do have it, and this
information can be extremely helpful to our goal, which is to make the
sharpest image possible without the client noticing that we “cheated”… :-)
The next round of solution is based on the color space you chose to work
with. Dan’s mantra here: “avoid sharpening in RGB if you can”. Yes, you
can always Fade it to Luminosity only, or, with the new smart filter in
CS3 simply have it to affect luminosity only, but all that being said,
he suggests to perform sharpening in other color spaces, at least if you
plan to use them at all.
Speaking of the other spaces…
LAB provides a single logical winner, its L (Luminosity) channel. There
is no point, says Dan (and it’s hard to argue with him on this), to
sharpen either A or B channels since the sharpening is the game of light
and dark while both of those channels carry the color information only.
However, if your path leads you to CMYK and, especially, if the image
has only one important color, Dan suggests that one should forget about
both LAB and, of course, RGB and perform the sharpening in CMYK. And in
many cases, he says, you can sharpen K only and be done with it.
OK, so far we have covered the basics. Sharpening should be postponed
until the end of the whole retouching process, and, when performed,
performed preferably in a specific channel only, the most common targets
being L or K. Now let’s proceed to the real meat of this chapter.
How
In many cases sharpening can be achieved in two totally different
ways:
have something in common: they both can be achieved by starting with … blur.
Yes, it was not a typo. In this part reader has to deal with a rather
controversial idea that to sharpen image we need to blur it, at least
initially. Then we can compare the blurred version with the original,
find the difference and finally enhance that difference in a way that
provides some sort of a sharpening effect, be it local a contrast
enhancement or adding strategically placed small halos of opposing
luminosity values. It goes without saying that no such travesty as a
manual edge selecting is attempted during the whole process.
For those of you who, like yours truly, find it hard at first to grasp
the concept of “sharpening via blurring” I created a little separate
tutorial conveniently named “Sharpening With Blur” that can be located
here.
Since I will be referring to it further down this review, here’s the
executive version of how you can sharpen the image using nothing more
than Gaussian Blur along with some blending modes and styles (which, by
the way, are covered separately here:
on, we’re ready for diving into our sharpening tool of choice, the
Unsharp Mask dialog.
USM by the Numbers
The USM dialog offers its user three control points: Amount,
Radius and Threshold. We’re going to see what each one does and learn
when to use which value.
Let’s start from the end and take a look at the Threshold parameter.
This parameter controls our Difference mask from the “Sharpening by
Blur” example. Zero means mask is used as is, including noise and all.
Greater values gradually exclude smaller differences, thus, in fact,
leading us to another rather controversial revelation: one can use a
sharpening tool as a noise removal one! As many of other things in this
chapter, it does not make much sense at first, but if you think of it a
little bit more, it actually does make a perfect one. Remember, we
started our sharpening process with the plain bold Gaussian blur. And
what the blur does, amongst other things? It kills the noise, and rather
effectively at that. So if during the rest of the process we somehow let
this noise-killing effect to survive, we might end up with two goals
reached instead of one: we would have our image both sharp and noise-free!
Let’s do a quick tour on the Threshold path.
Here’s a 100% on untouched rather noisy original I took at ISO 1600
during the Night Battle of Moorpark Civil War Reenactment in November 2006:
Next go four USM-versions of the same crop, all with 500% Amount,
1-pixel Radius and different amounts of Threshold, the latter ranging
from 0 to 50. At this point we’re not looking at the sharpening effect
at all, but rather at the blurring effect of the Threshold parameter.
For easier comparison, a little window in the middle of each image shows
the part of the original image.
As it is easy to see, despite of huge sharpening Amount value, the
appropriately large Threshold values can help us to keep the noise under
control.
Next parameter (we’re going upwards) is the all-important Radius. Its
value controls the width of the halos. In our “Sharpening via Blur”
example we had to work the Blur radius, the Levels amount and, finally,
the BlendIf marker position to deal with it. USM basically gives us a
single entry point to achieve whatever size halo we want to.
Here are several variations, all with Amount 500%, Threshold 25 and the
Radius accepting the values of 2.5, 5.0 and 10.0.
The last in our list of parameters, and the first in actual USM dialog,
is the Amount. It controls how radically colors and luminance are pushed
aside on the borders. It also serves as a separate topic for the “How
Much Is Too Much?” chapter-in-chapter in the book. While this topic
seemed to cause certain disagreement between Dan’s beta readers and in
photographic community in general, Dan sounds rather supportive for
“sharp as much as you want, and if you don’t hear any complaints, sharp
more” idea.
Interestingly and rather understandably, this parameter also plays an
important part in the alternative way of sharpening, which we we’ll
cover in the following section.
Halo World
The two largest issues one may encounter when dealing with
sharpening are the noise and the halos. Both, if not controlled, tend to
make our images downright ugly. So we need to learn how to get our image
to look sharp while keeping both of those undesirable guests, um, out of
the picture.
We already know that noise can be partially controlled by setting the
Threshold parameter higher. In general, noise should be removed or
diminished at much earlier phase of our whole retouching process
(remember, sharpening should happen as the very last step before
printing?), so we can temporarily forget about whole noise issue and
concentrate on the halos. How can we get rid of those tell-tell signs of
some heavy post processing?
Well, after reading (and, hopefully, trying for yourself) all those
contradicting techniques – blur to sharpen, sharpen to get rid of the
noise – you won’t be surprised to hear the next Dan’ suggestion: to
kill the halos we need to increase them. That is, we are going to
drastically increase the Radius, which, as you remember, controls the
width of the halo. And to camouflage this bold move all we have to do is
to simply decrease the Amount. Hence goes Dan’s trademark brand name for
this technique, “HiRaLoAm: High Radius, Low Amount”.
Let’s see how this technique works. I pickup up another image and have
chosen Amount 50%, Radius 50 pixels and Threshold 5 (to keep the sky
noise down a bit). This is a 100% crop view with the little windows
showing the part of the original (softer) picture.
As you can see, the sharpening effect, while not immediately obvious, is
definitely there. What, actually, was exactly our point: sharpen – yes,
(visible) halos – no.
While HiRaLoAm can be effectively used for the halo-less sharpening,
it’s not always wise to use this particular method (details to follow).
Few pages down the road Dan offers another method to get rid of them.
Using conventional and rather aggressive USM, make two copies of the
sharpened layer, set one blending mode to Darken, another to Lighten and
then play with opacity and/or BlendIf sliders to bring the halos down
while keeping the sharpening...
(Hey, wait a sec, didn’t we do it in our “Sharpening via Blur” example? ;-)
Which Is The One?
Having learned so much of some many ways of sharpening (and we
didn’t even touch sharpening in Camera RAW), it seems only natural to
ask ourselves, or, in this case, Mr. Margulis: which technique shall we
use?
The answer is, as you may have already guessed, “it depends”.
Dan suggests conventional USM (high amount, small radius) for the images
featuring a lot of clearly defined edges. He brings an example of a
yacht harbor, filled with masts, ropes, etc. I’d say the image of the
“firing squad” in our “USM by the numbers” section can be a good example
for this technique, too.
Since HiRaLoAm seems to work more like a shape-defining method and also
tends to lose fine details Dan suggests to use it where no such fine
details could be lost. He offers a marine shot featuring mostly sea and
sky. I think the plane shot above is a good target practice for this
method. You don't want to use it on the foliage or other texture rich
images,
He also suggests mix and match, depending on the image, desired result
and the amount of the copious spare time on your hands. For instance,
try conventional USM first followed by HiRaLoAm, both with the milder
settings than ones chosen if used separately. Or, in totally different
direction, ultra HiRaLoAm (R: 100, A: 20%), followed by regular HiRaLoAm
(R: 20, A: 50%), complete with conventional USM.
The Final Words
In this review we covered only the most important parts of the
chapter, anywhere from 50% to 75% being left for the diligent reader of
the original book to get through for him/herself.
Yet we have successfully learned that
final words from the chapter itself:
Happy Sharpening! :clap
Cheers! :1drink
Fifth Edition
By Dan Margulis
Chapter 6: Sharpening With a Stiletto
Review by Nikolai Sklobovsky,
www.photosocal.com
Foreword
First of all, we need to understand that perceived sharpness has often
nothing to do with our sharpening techniques.
Some researches shows that in many cases "the number one" winner can be
simply a higher resolution image, closely followed by increased
contrast, followed by adding a small amount of noise (but not too much).
The effect of the conventional spatial sharpening (as in PS Sharpen
command) only follows in the fourth place. Nevertheless, since this
chapter is about sharpening we are going to consider various aspects of
it, however subtle its results could be on the perceived image sharpness.
Right after Mr. Margulis’ gem went out of PeachPress publishing and into
the world, Adobe released a new beta version CS3 of its flagship
product, Photoshop. Amongst numerous improvements and innovation there
was one that would definitely affect the whole chapter, if not the whole
book: I’m talking about the smart filters. Starting from this version,
an operator can apply any filter – including sharpening - as an
adjustment layer of some sort and later return to it to modify its
properties, visibility, opacity, mask and position in the layer stack.
Yes, they are not perfect and do have an effect on the PSD file size
(you can read a nice technical insight on them at John
Nack’s blog), yet I think it’s a huge step in a right direction.
However, while this opens a whole new dimension in a retoucher’s life,
most of Dan’s techniques remain valid and, naturally, very sharp.:-)
In the Beginning
As it’s easy to see from the title, the chapter 6 is mostly
devoted to the sharpening techniques.
Through the whole chapter Dan’s motto seem to be “sharpen as much as you
can get away with, and then some”. This logic based upon rather
down-to-earth fact that while many prints across all the US print houses
were rejected for being “too soft”, few, if any, were returned for being
“too sharp”.
This does not necessarily mean that over sharpening cannot hurt. However, ourgenerally photographically uneducated customers may not necessarily
recognize the issue and simply state that they don’t like the image
because “it makes them look bad” (or something else to that extent).
From the very start Dan suggests to use one and only one PS sharpening
tool, obscurely named USM (Unsharp Mask) and ignore all the rest, even
though “the rest” being nicely named Sharpen, Sharpen Edges, Sharpen
More, and Smart Sharpen. His reasoning, as in the rest of the book, is
simple: the level of control this tool gives you is unsurpassed by the
rest, which give you none or much smaller amount of it.
When
Before diving into the fine details of several sharpening
scenarios, Dan emphasizes a very important concept he suggests to follow
religiously: sharpening should be the very last step in the whole
retouching process, the only thing that can happen after it is printing.
And if somebody else is gonna work on this image after you done with it,
he suggests to leave the sharpening out of the picture altogether.
With the new PS CS3 it seems that you can put your sharpening into those smartfilter layers at any time and at least get an idea of what you gonna
have. Later you can move this sharpening group to the top of the layer
stack and even disable it if you need to work on something else. You
will always be able to return to it and adjust the radius, threshold,
amount and other things.
Where
Another preliminary part of the chapter deals with the question
of “where”, as in “where in the picture should we apply our sharpening”.
The answer is not always straightforward and usually image specific.
Unless you going to “take a trip to Selectionville” – the concept Mr.
Margulis quite obviously dislikes a lot – you should look very carefully
to what you’re gonna sharpen.
I personally don’t see anything criminal in “selective” sharpening, especially if onecan do that in the matter of seconds as opposed to spending half-an-hour
or so looking for particular color/channel combination. But I guess
that’s just me…:-)
However, Dan immediately provides some general ideas that can help you
to answer this important question.
First thing he suggests we need to do is to find out if there is a
particular dominant (not to be confused with primary) color we want to
sharpen – or to avoid sharpening at all costs. Green foliage of bushes
and tress, red walls of canyons, yellow sands of the desert, deep brown
of the tree bark are all typical examples of such dominant color we
would like to sharpen. Gentle pink of a female face or deep blue of the
skies are the examples where we would like to have as little sharpening
as possible, since otherwise the first would be covered with nasty pores
and the latter simply become very, very noisy.
Obviously, not all images would have such a “color boss”. Some would
have several (in different parts of it), some would have none (or simply
too many to pick one). However, in many cases we do have it, and this
information can be extremely helpful to our goal, which is to make the
sharpest image possible without the client noticing that we “cheated”… :-)
The next round of solution is based on the color space you chose to work
with. Dan’s mantra here: “avoid sharpening in RGB if you can”. Yes, you
can always Fade it to Luminosity only, or, with the new smart filter in
CS3 simply have it to affect luminosity only, but all that being said,
he suggests to perform sharpening in other color spaces, at least if you
plan to use them at all.
Speaking of the other spaces…
LAB provides a single logical winner, its L (Luminosity) channel. There
is no point, says Dan (and it’s hard to argue with him on this), to
sharpen either A or B channels since the sharpening is the game of light
and dark while both of those channels carry the color information only.
However, if your path leads you to CMYK and, especially, if the image
has only one important color, Dan suggests that one should forget about
both LAB and, of course, RGB and perform the sharpening in CMYK. And in
many cases, he says, you can sharpen K only and be done with it.
OK, so far we have covered the basics. Sharpening should be postponed
until the end of the whole retouching process, and, when performed,
performed preferably in a specific channel only, the most common targets
being L or K. Now let’s proceed to the real meat of this chapter.
How
In many cases sharpening can be achieved in two totally different
ways:
- increasing local gradients in the color/luminosity transient areas,
thus making for the narrower and better defined edges; - adding contrasting color halos, thus making edges to “stand out” more.
Dan reminds us that this second trick was known to humanities for ages:
eyeliners add a dark halo close to eyes’ whites and make eyes look
sharper.
have something in common: they both can be achieved by starting with … blur.
Yes, it was not a typo. In this part reader has to deal with a rather
controversial idea that to sharpen image we need to blur it, at least
initially. Then we can compare the blurred version with the original,
find the difference and finally enhance that difference in a way that
provides some sort of a sharpening effect, be it local a contrast
enhancement or adding strategically placed small halos of opposing
luminosity values. It goes without saying that no such travesty as a
manual edge selecting is attempted during the whole process.
For those of you who, like yours truly, find it hard at first to grasp
the concept of “sharpening via blurring” I created a little separate
tutorial conveniently named “Sharpening With Blur” that can be located
here.
Since I will be referring to it further down this review, here’s the
executive version of how you can sharpen the image using nothing more
than Gaussian Blur along with some blending modes and styles (which, by
the way, are covered separately here:
- Blur a copy of image – I used 10-pixel radius Gaussian Blur.
- Change the mode of the blurred image to Difference and merge it down
to have a difference layer in normal mode - Make a copy of that difference layer, name them “Dark” and “Light”
- With the “Light” copy apply generous portion of contrast boosting. I
used Levels tool and set its White Point marker down to 17 - Put this layer next to another copy of the original image, set its
blending mode to Screen and merge them together. - On this layer go to its blending options and change its Underlying
layer BlendIf slider Black point to 127. - Now repeat the last steps for the “Dark” copy of the difference
layer. The difference would be: invert it after contrast boosting, use
Darken instead of the Screen and use White Point BlendIf slider instead
of Black Point one.
on, we’re ready for diving into our sharpening tool of choice, the
Unsharp Mask dialog.
USM by the Numbers
The USM dialog offers its user three control points: Amount,
Radius and Threshold. We’re going to see what each one does and learn
when to use which value.
Let’s start from the end and take a look at the Threshold parameter.
This parameter controls our Difference mask from the “Sharpening by
Blur” example. Zero means mask is used as is, including noise and all.
Greater values gradually exclude smaller differences, thus, in fact,
leading us to another rather controversial revelation: one can use a
sharpening tool as a noise removal one! As many of other things in this
chapter, it does not make much sense at first, but if you think of it a
little bit more, it actually does make a perfect one. Remember, we
started our sharpening process with the plain bold Gaussian blur. And
what the blur does, amongst other things? It kills the noise, and rather
effectively at that. So if during the rest of the process we somehow let
this noise-killing effect to survive, we might end up with two goals
reached instead of one: we would have our image both sharp and noise-free!
Let’s do a quick tour on the Threshold path.
Here’s a 100% on untouched rather noisy original I took at ISO 1600
during the Night Battle of Moorpark Civil War Reenactment in November 2006:
Next go four USM-versions of the same crop, all with 500% Amount,
1-pixel Radius and different amounts of Threshold, the latter ranging
from 0 to 50. At this point we’re not looking at the sharpening effect
at all, but rather at the blurring effect of the Threshold parameter.
For easier comparison, a little window in the middle of each image shows
the part of the original image.
As it is easy to see, despite of huge sharpening Amount value, the
appropriately large Threshold values can help us to keep the noise under
control.
Next parameter (we’re going upwards) is the all-important Radius. Its
value controls the width of the halos. In our “Sharpening via Blur”
example we had to work the Blur radius, the Levels amount and, finally,
the BlendIf marker position to deal with it. USM basically gives us a
single entry point to achieve whatever size halo we want to.
Here are several variations, all with Amount 500%, Threshold 25 and the
Radius accepting the values of 2.5, 5.0 and 10.0.
The last in our list of parameters, and the first in actual USM dialog,
is the Amount. It controls how radically colors and luminance are pushed
aside on the borders. It also serves as a separate topic for the “How
Much Is Too Much?” chapter-in-chapter in the book. While this topic
seemed to cause certain disagreement between Dan’s beta readers and in
photographic community in general, Dan sounds rather supportive for
“sharp as much as you want, and if you don’t hear any complaints, sharp
more” idea.
Interestingly and rather understandably, this parameter also plays an
important part in the alternative way of sharpening, which we we’ll
cover in the following section.
Halo World
The two largest issues one may encounter when dealing with
sharpening are the noise and the halos. Both, if not controlled, tend to
make our images downright ugly. So we need to learn how to get our image
to look sharp while keeping both of those undesirable guests, um, out of
the picture.
We already know that noise can be partially controlled by setting the
Threshold parameter higher. In general, noise should be removed or
diminished at much earlier phase of our whole retouching process
(remember, sharpening should happen as the very last step before
printing?), so we can temporarily forget about whole noise issue and
concentrate on the halos. How can we get rid of those tell-tell signs of
some heavy post processing?
Well, after reading (and, hopefully, trying for yourself) all those
contradicting techniques – blur to sharpen, sharpen to get rid of the
noise – you won’t be surprised to hear the next Dan’ suggestion: to
kill the halos we need to increase them. That is, we are going to
drastically increase the Radius, which, as you remember, controls the
width of the halo. And to camouflage this bold move all we have to do is
to simply decrease the Amount. Hence goes Dan’s trademark brand name for
this technique, “HiRaLoAm: High Radius, Low Amount”.
Let’s see how this technique works. I pickup up another image and have
chosen Amount 50%, Radius 50 pixels and Threshold 5 (to keep the sky
noise down a bit). This is a 100% crop view with the little windows
showing the part of the original (softer) picture.
As you can see, the sharpening effect, while not immediately obvious, is
definitely there. What, actually, was exactly our point: sharpen – yes,
(visible) halos – no.
While HiRaLoAm can be effectively used for the halo-less sharpening,
it’s not always wise to use this particular method (details to follow).
Few pages down the road Dan offers another method to get rid of them.
Using conventional and rather aggressive USM, make two copies of the
sharpened layer, set one blending mode to Darken, another to Lighten and
then play with opacity and/or BlendIf sliders to bring the halos down
while keeping the sharpening...
(Hey, wait a sec, didn’t we do it in our “Sharpening via Blur” example? ;-)
Which Is The One?
Having learned so much of some many ways of sharpening (and we
didn’t even touch sharpening in Camera RAW), it seems only natural to
ask ourselves, or, in this case, Mr. Margulis: which technique shall we
use?
The answer is, as you may have already guessed, “it depends”.
Dan suggests conventional USM (high amount, small radius) for the images
featuring a lot of clearly defined edges. He brings an example of a
yacht harbor, filled with masts, ropes, etc. I’d say the image of the
“firing squad” in our “USM by the numbers” section can be a good example
for this technique, too.
Since HiRaLoAm seems to work more like a shape-defining method and also
tends to lose fine details Dan suggests to use it where no such fine
details could be lost. He offers a marine shot featuring mostly sea and
sky. I think the plane shot above is a good target practice for this
method. You don't want to use it on the foliage or other texture rich
images,
He also suggests mix and match, depending on the image, desired result
and the amount of the copious spare time on your hands. For instance,
try conventional USM first followed by HiRaLoAm, both with the milder
settings than ones chosen if used separately. Or, in totally different
direction, ultra HiRaLoAm (R: 100, A: 20%), followed by regular HiRaLoAm
(R: 20, A: 50%), complete with conventional USM.
The Final Words
In this review we covered only the most important parts of the
chapter, anywhere from 50% to 75% being left for the diligent reader of
the original book to get through for him/herself.
Yet we have successfully learned that
- sharpening is, in fact, based on blurring;
- one can hide such nasty things as halos by increasing their size;
- the whole sharpening process can be used as a noise removal one;
- it’s preferable to sharpen per-channel (L and K being Dan’s
favorites) and on a separate layer. - two different methods of sharpening can be used for different target
areas and sometimes both can be mixed, but in any case Dan recommends to
postpone sharpening to the last possible moment. - those comfortable with blending modes can use bold conventional USM
values and then get rid of halos by using Darken/Lighten modes and
BlendIF sliders
final words from the chapter itself:
Dan Margulis wrote:Above all, in USM, be greedy at all times. Remember
the strategy of asking for the raise. There’s no fixed limit. The best
amount of sharpening is the largest one you can get away with.
Happy Sharpening! :clap
Cheers! :1drink
"May the f/stop be with you!"
0
Comments
There is one issue in your overview that I question and am interested in hearing others opinion on this as well as yours Nikolai. Making sharpening your last move a steadfast rule is not the way I see it. Dan has stated that it is a rule that can be broken if minor adjustments need to be made after your sharpening has been done. The best example I can use is when RGB is your output space. Very often you go to other spaces to make corrections and if you do then the sharpening should be done there. If you have gone to LAB to make your corrections then you should obviously do your sharpening in the L channel. However, there are times that after you are done with LAB and it's more blunt moves you may want to fine tune your image in the output space of RGB. You could reasonably perform your sharpening in LAB and then fine tune with minor corrections in RGB once you got there.
Good point. However, in Dan's LAB book he discusses a test he performed on a file, converting it from RGB to LAB and back again numerous times, with no discernable or measurable loss or degradation in color. That being the case, a final jump into LAB to sharpen could follow a series of color moves in various spaces, as the last stop on the tour.
—Korzybski
Great examples, explanations and illumination of an arcane science.
—Korzybski
A few points I think are worth adding:
neat summary-
now did I understand all of it?-
'it depends'-
seriously, good job-
OK, bad example, but that wasn't my point. The point is that corrections can be made after sharpening as long as they are not significant or major. I personally have not needed it myself but for others I was just saying that it is not a rule that needs to be followed religiously and never broken. It may be an irrelevant issue but I'd like to know what people think.
I agree. If you're nearly done with an image (including sharpening) and upon last look, you decide you need one more small correction, I have no problem doing it on top of the sharpening and calling the image done. It's not how you plan your workflow, but sometimes your time is better spent on other images than undoing your sharpening (which sometimes means going back to LAB mode) just to make a minor correction.
Homepage • Popular
JFriend's javascript customizations • Secrets for getting fast answers on Dgrin
Always include a link to your site when posting a question
Thanks for reading and commenting!
Re: when...
I'm standing by Dan's point ("sharpening should be the last step before print") for the following reason: unlike any other color correction effect, sharpening is not scalable. This means, sharpening scenario for 4x6 *must* be different from the one for 20x30. Everything else can be done at earlier stages (some steps are still recommended to be performed earlier than the others:-), but sharpening definitely is the one that should be performed *exactly at* the target resolution.
Given this rationale, everything else is "one size fits all", and as such can be done once and for all. Sharpening needs to know the exact target, hence it's the last step.
Yes, you are correct in your line of thought that *some* things can be done even later. Like a minor cropping that does not change a resolution. But that is pretty much it.
Besides, as I specifically mentioned, with the CS3 you can perform smart sharpen to get a preview at any point in your workflow, then disable it and move it to the top of the stack for the later use.
So, are there exceptions?
Sure, there always are, it's an art, not a math.
HTH
Thanks, appreciate it!
Thank you for the comments!
I agree, your #1 point is a great solid approach, I should've mention that in the review.. :bash
I'm glad you've added it:-)
Re #2: While I agree that an "even" (power of two, i.e. 100%, 50%, 25%, etc.) scale factor makes for a generally better looking screen representation and shows less artefacts, an alternative - and quite popular - suggestion is to keep the whole image visible (Ctrl/Cmd+0) and concentrate on the preview section of the USM dialog. This way you can get both general view and the specific details.
Cheers!
Thank you!
Glad you liked it! :ivar
I've been known to do this, but it doesn't show color when sharpening only a single channel. So I really do like looking at an even power magnification. I have found that otherwise I tend to sharpen too little, being afraid of the artifacts that won't, in fact, impact the final result.
Valuable point (as always:-).
I guess, again, "it depends.."
Well I just observed myself work on a couple of images and I realized that there is one step I almost always do after sharpening (if I do it at all). And what's more, I've watched Dan in his classes, and he does it the same way. I think that's where I got it.
My final step (when I do it) is a trip to CMYK to establish good shadows and deep blacks. (See this recent example.) Almost always I have sharpened in LAB first. Under special circumstances I'll hold back and sharpen that K channel in CMYK, especially portraits of older people. But I've gotten good at doing that in LAB and keeping it away from skin (blend-if sliders, inverted K channel mask, there's plenty of ways).
So, anyway, when there's a trip to CMYK for the shadows, both Dan and I usually do it after sharpening. At least that's what I've observed.
Interesting commment.
Let me ask you this, however: would the result be much/any different if the shadow treatment was done prior to sharpening?
The trip to CMYK is not benign like the trip from RGB to LAB and back. You don't want to do it more than you have to. If the destination color space is CMYK, then there should be just one move to CMYK at the very end. If CMYK isn't the destination, then I suppose you can go to LAB for color and contrast enhancement, go to LAB via RGB to set the shadows, then back to LAB for sharpening.
I don't think you'll actually get any gain from doing it that way, but it's more steps. Dan thinks we are weenies because we don't target CMYK, in case you haven't figured that out. Pros print their stuff on high volume photo offset presses. And sleep in the press room for the duration.
I understand Dan's attitude toward CMYK as well as the perils of going back and forth, in and out of this space.
I think that the two primary reasons for postponing the sharpening are:
- you need to know the final size/resolution, and
- any bold contrast/color boosting moves may ruin your carefully selected threshold/amount values and give away your sharpening treatment
As Isaac mentioned in his first reply, any minor actions are most likely to be OK.And yes, if you're sharpening in LAB and finally going to CMYK to make the final adjustments with the intention to stay there for good - I agree, it make total sense to avoid extra steps.
This is basically what I was talking about. I, personally, have not performed any tasks after USM myself but I thought it should be made known that it is not a rule that can never ever be broken when someone needs to make a move after words. Rutt's example is a good one.
I have been hitting View>Print Size and then CTRL/CMD+= to move up to the next viable round number. With images of 300ppi that puts you at 25% usually. I like it here as opposed to 100% because you see a bunch of stuff that never prints at 100% as was mentioned. It was also preferred by a beta reader in the book. I don't know if this is the best way or not but that's the way I do it.
I had a conversation with Dan recently and he took the opportunity to tell me about things he has decided since PP5E went to press.
Rutt: Right.
Dan: Wrong! Sharpen all three channels in LAB.
Rutt: Why? Oh, you get some color action that way?
Dan: Right!
Here is an example. I wanted to make a print of this old shot of mine and thought I'd reprocess to see if I could do better now that I know what Dan would do with it. I followed the his move from Chapter 4 and then tried to decide how to sharpen. Maybe too much noise for conventional RGB and not many edges. So I chose to sharpen in just HiRaLoAm. Radius: 35, Threshold: 4, Amount: 45.
L only sharpened:
All threee channels sharpened:
The difference is subtle at first, perhaps you'll want to download the originals and look at them in Photoshop instead of your browser.
Original with all 3 channels sharpened
Look at the shadows in the sky. See how the yellows are less desaturated?
Interesting. Great shot, by the way, whichever method you use.
No question that there's enhanced color with all three channels, but is it to be trusted? (The following samples are full-sized crops.)
In this area of dark clouds we seem to have just increased color noise. I could easily see the order of these images reversed to demonstrate LABs unique ability to reduce such noise, with the original being the desired result.
_____
Here, the brightening of yellows comes at a cost to detail in the pastel hues.
_____
Here, the clouds are now blue. But, are they really blue? Feels a bit like the old "Is it a bug, or a feature" problem with software glitches.
If pushing for more color in areas that seem too desaturated is the goal, Dan's Man-from-Mars technique would seem to be more reliable, as well as more customizable.
—Korzybski
Hey, no fair, showing such enlarged detail. I sharpened for the size shown and probably overdid it in all dimensions to make Dan's point.
Also this is an older shot and I only have a jpeg. I think Dan might have a raw version.
There is a lot of noise here that I need to address somehow in order to get a good large print (printed these this AM and they sucked.) Dan's verion in PP5E looks darned good, eh? He says that ALL he did was the LAB adjustments he shows in Ch 4, but we know that isn't quite true since he converted to CMYK for the book.
It's not enlarged, it's the actual full size version that you linked us too, which is the version you sharpened, yes? My crops are small regions but actual size. The difference becomes less noticable in the reduced version that you printed in the thread, but that's not the point.
—Korzybski
Since we know the HiRaLoAm actually does sharpening via blur (see the original review) and can be considered as a shape-defining-noise-removing mechanism I can see that his approach may just work, especially for the landscapes.
I will give it a try over the weekend. The difference may be subtle, but it may be worth it:-)
I've thought this through again and realized that I don't quite agree with the accepted wisdom of sharpening after resizing, especially resizing for a the web as in a dgrin contest 800 on the long side.
Resize first then sharpen and you lose a lot of control. Radius of .1 not enough, .2 too much. Sharpen first and you might have a full pixel or more of reasonable range to choose from. Similarly for HIRALOAM; resize first and you lose fine control.
This old saw is the second cousin of the right advice: sharpen at some size close to the expected viewing size (and an even multiple of the actual size.)
Then resize. I suppose I'd reverse this advice if the resizing were some sort of up-resing.
At least that's my current opinion.
So, I guess your p.o.v. now is that sharpening should go before downsizing, but not necessarily before upsizing.
But at any rate do you still agree that it should be one of the last steps in the workflow, i.e. applied after all the other manipulations (with the notable exception of the size change)?
And before a final trip to CMYK if there is going to be one.
Usually, it a little loser. If I was in LAB, then it's the last thing I do before leaving LAB. There are some final things I might do in RGB (selective color, change duplicate sharpening layers and change blending modes to lighten/darken with individual opacity, trip to CMYK for shadows, last minute color tuning with curves) and these will come after LAB sharpening.
Dan's point "if the image is destined for CMYK - good chances that you need to sharpen in CMYK".
I don't work in CMYK myself, so 1) I don't have my own experience, and 2) the whole CMYK point is rather theoretical to me, nless we simply talking about using its channels where appropriate.
But your suggestion seems to be in a major contradiction with the book...
Dan likes K channel sharpening in CMYK a lot. But in practice, I've seen him sharpen in LAB and then go to CMYK and then sometimes sharpen again. Or sharpen a duplicate early in CMYK and then use the K channel as a stamp in RGB, sometimes before a trip to LAB to enhance colors. (Sort of like Machiavelli's tomb.)
There are guidelines and then there are the reasons behind the guidelines. Sharpen late all other things being equal. But sometimes they aren't.