Sharpening

On 3/23/2013 2:33 PM, Alfred Molon wrote:
In article , Eric Stevens
says...
Have you tried High Pass sharpening?
http://www.luminous-landscape.com/tu...arpening.shtml

Just gave it a quick try and it doesn't seem to be better than USM.


If you switch to LAB mode, and sharpen only the lightness channel. This
will minimize halos.
First create a new layer so if you oversharpen you can back it down.
Also, if there are areas you don't want sharpened, you can mask them off.

--
PeterN

In article , Tony Cooper
wrote:

First create a new layer so if you oversharpen you can back it down.

Always. For most adjustments.

rarely.

learn about smart objects and non-destructive workflows.

On 3/23/2013 6:22 PM, Tony Cooper wrote:
On Sat, 23 Mar 2013 16:17:16 -0400, PeterN
wrote:

On 3/23/2013 2:33 PM, Alfred Molon wrote:
In article , Eric Stevens
says...
Have you tried High Pass sharpening?
http://www.luminous-landscape.com/tu...arpening.shtml

Just gave it a quick try and it doesn't seem to be better than USM.


If you switch to LAB mode, and sharpen only the lightness channel. This
will minimize halos.

In LAB, you can pull the A and B channels in a bit on both ends of the
curve and pop the colors.

And do all sorts of things that are much harder to do in RGB.
My bible, and IMHO one of the most authoritative books on LAB color.
http://www.amazon.com/Photoshop-LAB-Color-Adventures-Colorspace/dp/0321356780



First create a new layer so if you oversharpen you can back it down.

Always. For most adjustments.

Also, if there are areas you don't want sharpened, you can mask them off.



--
PeterN

Alfred Molon wrote:
I was asking a generic question, and was curious to know to what extent
technology has progressed today.

Sharpening of digital images has not actually changed
any at all in literally decades (since well before
DSLR's were on the market).

There are exactly two basic forms that sharpening takes,
and everything else is just a mixture of the two or one
way or another to apply sharpen to only specific parts
of an image. That is, techniques such as "smart
sharpen", which usually means masking off only edges for
sharpening, or appling sharpening only to the luminance
channel in LAB mode, etc etc.

But there are significant distinctions between the two
basic kinds of digital sharpening that can be very
helpful to know about and understand. Unsharp Mask is
one type. What is best called High Pass Filter Sharpen
is the other (it should not be confused with the High
Pass Sharpen tool so titled by Adobe, which is
different.) Some people call it Convolutional Sharpen,
but USM can also use convolution so that is not a valid
distinction.

High Pass Sharpening, which I'll refer to as simply
"Sharpen" from this point on, uses a digital high pass
filter to detect high frequency spatial detail that
consists of a sequence of multiple tonal changes. Think
of a white picket fence outlined against a dark
background. Sharpen will increase the brightness of the
white parts and decrease the brightness of the dark
areas in between. Generally, the most change will be
made closer to the center of the tonal change. The
"amount" of change is one parameter and the "radius" (or
distance between edges, which is essentially the cutoff
frequency of the filter) is another, plus a third
parameter called "sigma" that is the distribution of the
change over the radius that is affected.

Visually, I=I=I=I=I=I is changed to this: | | | | | |

The way that Unsharp Mask works is different. It also
looks at tonal transition edges, but only single edges
and not multiple edges. The picket fence would be seen
as an average of the white pickets and the dark spaces
between, and considered just a single "grey" continuum.
But if there is an break in the fence (and open gateway,
for example), the transition from the average gray of
the fence on the left to the dark only background
through the gate would be seen as a transition, and then
on the other side the entirely separated transition from
the dark background to the average gray of the fence would
be seen also as a transition.

I=I=I=I=====I=I=I=I is changed to: I=I=I_H_===_H_I=I=I

Only the single transition is made more distinct

But it also has to be understood that the above represents the
"normal" adjustments for parameters. In fact USM parameters can
be set to see each picket in the fence, and the effect might be
somewhat similar for the two types.

But because USM will see a sequence as an average of the
values (that is where the "unsharp" in the name comes
from), the effects that it applies are irreversible. A
Sharpen tool is exactly the same algorithm that is used
for Blur, but with different parameters and either can
reverse the other.

The distinction between the two types comes into play
when an image from a camera that uses a Bayer Color
Filter is used. If no resampling is done, both types of
sharpening will probably have about the same effect and
almost any image will benefit from application of one,
the other, or both. After and image is down sampled
however, the effects of USM will usually be greater than
those of Sharpen. If the image is up sampled, the
effects of Sharpen will be greater than USM on most
images.

Note also that the higher the pixel resolution the
greater the amount and radius that is required to get an
effect.

The bottom line is that if you enlarge images for
printing it is very likely that Sharpen will be more
important than USM, while if you reduce images for web
display it is most likely that USM will be more
important than Sharpen.

--
Floyd L. Davidson http://www.apaflo.com/
Ukpeagvik (Barrow, Alaska)

Floyd L. Davidson wrote:
Alfred Molon wrote:

I was asking a generic question, and was curious to know to what extent
technology has progressed today.

Sharpening of digital images has not actually changed
any at all in literally decades (since well before
DSLR's were on the market).

It's just like transportation technology: Nothing new since
the wheel has been invented. Horse drawn carts - wheel.
Cars - wheel. Trains - wheel. Planes - wheel.

There really has be no change at all since the invention of
the wheel, see?


There are exactly two basic forms that sharpening takes,
[...]
High Pass Sharpening, which I'll refer to as simply
"Sharpen"
[...]
Unsharp Mask
[...]

- warp sharpening

That makes 3, doesn't it?
Typical well informed Floyd.


But because USM will see a sequence as an average of the
values (that is where the "unsharp" in the name comes
from),

Nope. It's an analogue film technique, using a somewhat
blurred low contrast positive on top of the negative to
partially cancel low frequencies, but not high frequencies.
According to some sources it's a technique from the 1930's.

The distinction between the two types comes into play
when an image from a camera that uses a Bayer Color
Filter is used. If no resampling is done, both types of
sharpening will probably have about the same effect

.... if you're blind drunk, that is, with the emphasis on
*blind*. Simply trying it yourself will show that.


Note also that the higher the pixel resolution the
greater the amount and radius that is required to get an
effect.

You're assuming identical output sizes, which is not given.

-Wolfgang

Wolfgang Weisselberg wrote:
Floyd L. Davidson wrote:

There are exactly two basic forms that sharpening takes,
[...]
High Pass Sharpening, which I'll refer to as simply
"Sharpen"
[...]
Unsharp Mask
[...]

- warp sharpening

That makes 3, doesn't it?
Typical well informed Floyd.

Sorry, that is another thing like "smart sharpen" and
half a dozen other names that people give to specific
implementations of USM or Sharpen.

But because USM will see a sequence as an average of the
values (that is where the "unsharp" in the name comes
from),

Nope. It's an analogue film technique, using a somewhat
blurred low contrast positive on top of the negative to
partially cancel low frequencies, but not high frequencies.
According to some sources it's a technique from the 1930's.

You don't have a clue how it works do you! Hilarous.

The blur is used to average a sequence of variations.

Yes it was done with film using analog techniques long
before digital photography existed. That doesn't change
the theory of how and why it works though.

That is not different than Sharpen, which was also
developed and understood with film before digital
existed. Look up Eberhard Effect, Adjacency Effect,
and/or Mackie Lines.

The distinction between the two types comes into play
when an image from a camera that uses a Bayer Color
Filter is used. If no resampling is done, both types of
sharpening will probably have about the same effect

... if you're blind drunk, that is, with the emphasis on
*blind*. Simply trying it yourself will show that.

You seem to be all of that. What's the point of such
drivel.

Note also that the higher the pixel resolution the
greater the amount and radius that is required to get an
effect.

You're assuming identical output sizes, which is not given.

What I specified was "higher the pixel resolution", and
that was specified because that is exactly what makes a
difference.

You'll have to define "output sizes" in terms of pixel
resolution to make an un-ambiguous statement out of what
you've said.

--
Floyd L. Davidson http://www.apaflo.com/
Ukpeagvik (Barrow, Alaska)

Eric Stevens wrote:

On Fri, 22 Mar 2013 23:38:33 +0100, Alfred Molon
wrote:

Is there anything better than unsharp mask to sharpen an image? More
specifically, to compensate for a not so sharp lens or a (small)
autofocus error.

Have you tried High Pass sharpening?
http://www.luminous-landscape.com/tu...arpening.shtml


Thanks for this pointer. It is my new favorite method of sharpening. Because
it is a layer, it is entirely reversible at any time, unlike USM.

This article
http://photo.tutsplus.com/tutorials/...h-pass-filter/

suggests a couple more interesting uses of high pass.

Thanks,
Doug

On 2013-03-31 10:41:18 -0700, Douglas Johnson said:

Eric Stevens wrote:

On Fri, 22 Mar 2013 23:38:33 +0100, Alfred Molon
wrote:

Is there anything better than unsharp mask to sharpen an image? More
specifically, to compensate for a not so sharp lens or a (small)
autofocus error.

Have you tried High Pass sharpening?
http://www.luminous-landscape.com/tu...arpening.shtml


Thanks for this pointer. It is my new favorite method of sharpening. Because
it is a layer, it is entirely reversible at any time, unlike USM.

This article
http://photo.tutsplus.com/tutorials/...h-pass-filter/

suggests

a couple more interesting uses of high pass.

Thanks,
Doug

One word of warning when using high pass filter sharpening, it is not
always the best choice for all sharpening tasks, and should be applied
with a subtle hand.
In images with strong textures, if one is not careful, it is easy to
over cook high pass sharpening, which results in artifacts noise, and
obvious over sharpening. At first look it might seem the image has
gained some "Pop", but on closer examination you will discover it is
over cooked.

So, unless you want that over cooked look try and keep the pixel radius
in the 2-6 range. If you start moving above a radius of 10 you will
start seeing artifacts in most images. Then you will get slightly
different results by your choice of layer blending mode. Your best
choices will usually be, "Overlay", "Soft Light", "Hard Light", or
"Vivid Light".

--
Regards,

Savageduck

Floyd L. Davidson wrote:
Wolfgang Weisselberg wrote:
Floyd L. Davidson wrote:

There are exactly two basic forms that sharpening takes,
[...]
High Pass Sharpening, which I'll refer to as simply
"Sharpen"
[...]
Unsharp Mask
[...]

- warp sharpening

That makes 3, doesn't it?
Typical well informed Floyd.

Sorry, that is another thing like "smart sharpen" and
half a dozen other names that people give to specific
implementations of USM or Sharpen.

Interesting: You're either too stupid to use Google to do some
research, too arrogant to check your own assumptions or too
stupid to understand what you read.

Or are you saying that USM or Sharpen squashes and stretches
pixels in areas --- i.e. moves them around --- but do not at
all change the pixel values? Really?

Here's the article from 1999:
http://www.cs.technion.ac.il/~gotsma...ementByI-D.pdf
aka http://preview.tinyurl.com/cp3r749
Read it and try to understand it.

I expect your apology and you to concede you were wrong.

But because USM will see a sequence as an average of the
values (that is where the "unsharp" in the name comes
from),

Nope. It's an analogue film technique, using a somewhat
blurred low contrast positive on top of the negative to
partially cancel low frequencies, but not high frequencies.
According to some sources it's a technique from the 1930's.

You don't have a clue how it works do you! Hilarous.

Indeed --- your claims are hilarious.

http://www.iovs.org/content/12/6/461.full.pdf
A 1973 article on unsharp masking in fundus photography.
Also describes how one can create the blurred unsharp mask.
Of course completely in analog film.

The blur is used to average a sequence of variations.

i.e. partially cancelling out the low frequencies.

Yes it was done with film using analog techniques long
before digital photography existed. That doesn't change
the theory of how and why it works though.

Indeed. It partially cancels out the low frequencies (as they
are not smoothed away by blurring the mask) but not the high
frequencies (which *are* smoothed away by blurring the mask).

Which is what I said.

That is not different than Sharpen, which was also
developed and understood with film before digital
existed. Look up Eberhard Effect, Adjacency Effect,
and/or Mackie Lines.

USM is the very same algorithm in analogue and digital. Make
unsharp positive mask, apply mask with appropriate density
(either by exposing it for a shorter time than the negative
to the paper or by using a low density film for the mask --
or in digital use the opacity slider).

High Pass Sharpen in film works during development and is a
chemical process which is a one-time chance to get it right.
In digital, it's applied as a mask well after the equivalent
of development.


Now --- how do you do warp sharpening in film? How do you
move film grains?

The distinction between the two types comes into play
when an image from a camera that uses a Bayer Color
Filter is used. If no resampling is done, both types of
sharpening will probably have about the same effect

... if you're blind drunk, that is, with the emphasis on
*blind*. Simply trying it yourself will show that.

You seem to be all of that. What's the point of such
drivel.

The point is that I happen to have eyes.


Note also that the higher the pixel resolution the
greater the amount and radius that is required to get an
effect.

You're assuming identical output sizes, which is not given.

What I specified was "higher the pixel resolution", and
that was specified because that is exactly what makes a
difference.

Yep: a compact camera sensor has a higher pixel resolution
(per area) than a FF sensor, even when the FF has more pixels.


You'll have to define "output sizes" in terms of pixel
resolution to make an un-ambiguous statement out of what
you've said.

So you scale down one copy and scale up the other copy of the
identical photo ... and poof.

Oh, and there's a difference if you have the identical pixel
resolution and a 4x6 inch print or a 40x60 inch print ...

-Wolfgang

Douglas Johnson wrote:
Eric Stevens wrote:
On Fri, 22 Mar 2013 23:38:33 +0100, Alfred Molon

Is there anything better than unsharp mask to sharpen an image? More
specifically, to compensate for a not so sharp lens or a (small)
autofocus error.

Have you tried High Pass sharpening?
http://www.luminous-landscape.com/tu...arpening.shtml

Thanks for this pointer. It is my new favorite method of sharpening. Because
it is a layer, it is entirely reversible at any time, unlike USM.

You can put USM (set on the strong side) on a separate layer and
change it's opacity. And of course you can use a layer mask
to only apply at the edges.

-Wolfgang