Reading mode: Light Dark

New Sharpening Concept

Go to last post
True but your statement could mislead people. Compared pixel for pixel the statement is true. If you assume about twice the number of Bayer pixels as Fovean pixels, then it is pretty much a wash (assuming some content in the green channel). The higher end Bayer cameras have factors of 2-almost 5 times the pixels as the current Fovean so they have equal or higher resolving power.

This assumes that there is some green channel as this channel tends to dominate the luminance and detail definition. If you had a pure red against black or red, then there is a bit of a penalty. Same for pure blue. These situations are very rare where there is no green at all.

I would say the sharpening of different channels different ways might have an impact but on the margin. Of somewhat more impact is different noise reduction in different channels. The blue often is pretty noisy with the red not all that far behind. Differential noise reduction sometimes can help.

On the same vein, if I was going to sharpen the blue channel, I would probably do a bit of noise reduction first to avoid amplifying the higher noise typically found in that channel. Same for the red. It is hard to pose general rules as in Bayer sensors, results are image content dependent because of the nature of the beast. When comparing with the Fovean, however, the larger number of pixels in the high end sensors makes this pretty much moot.

It is probably worth while experimenting if there is an issue with a particular image but I'm not sure it is all that worth while in the average workflow. Probably a bit of marketing going on here.
Stan said:
Bayer sensor cameras can't always provide the same degree of
apparent sharpness to things of different colors. The Sigma
doesn't really experience that issue. He made an attempt to give
those colors which were sampled less favorably more of the
sharpening. To try and even things out a bit is the idea.

Stan
Click to expand...
--
Leon
http://pws.prserv.net/lees_pics/landscapes.htm
 
..shouldn't improve so much is Bayer is working perfectly! If you examine the DIFFERENCE in sharpening between the back up lights and the red lamps you can see that this becomes more even after EQ. At the same time there is more overall sharpness, but that is not what he is trying to get across. You can sharpen globally with regular USM.

Here is some text which describes the strategy a bit. Basically it's USM that is applied differently to different RGB combos. Also, here are some samples, both are from Mike Chaney:

"I experimented with an adaptive unsharp mask filter this morning that is designed to equalize sharpness of colors that have been "smudged" by Bayer sensor interpolation. This filter is nothing more than an unsharp mask filter that doesn't sharpen B/W detail at all but applies sharpening to colored areas proportional to the number of sensors that contribute to color. For example, a black line on white paper will receive no sharpening while a black line on pure red will receive 100% of the indicated USM. Black on green gets 50% of the indicated USM. Desaturated green where the red and blue channels are not zero but are half the intensity of green receive 25% of the indicated USM. Etc.

Here are the results. I think this filter gives the Bayer images more of a 3D effect and it certainly balances sharpness between B/W detail and "primary" detail like black/red. Keep in mind that this filter took me about 10 minutes to write so it is in its early stages. I may add some code to "tone down" the intensity of the filter in shadows to reduce noise, etc."



Stan
 
After reading a number of posts in the past year or so, some are idiotically opinionated (with not much but gibberish, no facts, no insight) and some have a great clarity like the below post.

I just thought I'd write to thank you for the nicely written post that explains an aspect of digitial photography that is not as commonly understood as it should be. We should have "Post of the Day" awards for really nice gems like yours...

thanks again for the clarity you share (on sharpening at that, ha ha)
Sandman said:
Intrinsic,
Sharpness with film and digital are like night and day. Digital
images from DSLR cameras are intrinsically soft. They are softened
by an antialiasing filter that eliminates moire and other artifacts
in the digital capture process. In order to restore sharpness to
the image, you have three choices: adjust the in-camera parameter
settings to sharpen your images just after they are captured,
adjust the sharpness using software in post-processing, or have
soft images! Sometimes soft is OK and you will not have to do
anything (portraits, some macros, etc). For everything else, at
least some sharpening is required to get the same look and feel as
a film image.

Most of us serious amateurs choose to do the sharpening in
post-processing, since we have much more control over the process
than choosing to do this in-camera. Once in-camera sharpening is
accomplished, there is no way to undo it or to modulate the effect.

For those of us who do our sharpening in post-processing, we are
always in search of the BEST technique to accomplish this goal,
hence the relevance of Paul's review of this article. We already
have dozens of techniques to choose from, each offering some
benefit and some disadvantage. What we mostly want to avoid is the
appearance of sharpening artifacts, like jagged diagonal lines,
halos, etc.

While there are many similarities between film and digital, this is
just one of the major differences between the two media. I can't
imagine why anyone would not like going digital. Ther are some
adjustments to make, but the leap is well worth it. Most of us are
on our third or fourth digital camera, so there is a wealth of
knowledge here on the topic and, as you might have gleaned by
jumping into this thread, we are all on the same quest of expanding
our knowledge of this topic every day!

Cheers,

jim

--
Shoot more, ***** less!
galleries at: http://www.pbase.com/sandman3
Click to expand...
--
Pixelitus Maximus
 
VideoProduction said:
Once converted out of RAW, the image is already in the RGB domain
and gray IS determined by RGB channels. Only LAB and CMYK have a
separate gray (Lightness/K) channel. I fail to see how the gray
pixels avoid being effected by the sharpening process in RGB mode.
Of course, you cannot sharpen R/G/B components in the other 2 modes.

My recent technique of choice for sharpening is as follow:
  • Complete editing the image.
  • Stamp visible to a new layer (create a new layer and press
Sh+Ctl+Alt+E)
  • Desaturate the Layer.
  • Apply Hi-Pass filter (1 to 3 works for me, experiment with your
setting)
  • Change layer blend mode to Overlay
  • Create an action for the above steps (for a very mild sharpening)
  • Run this action on any image requiring sharpening.
I like this process for the following reasons:
  • The process is non-destructive to the image.
  • You can stack the action for more sharpening.
  • You can reduce the fill % for less sharpening.
  • A combination of the above should give you the precise amount of
sharpening that you want.
  • You can select a copy of any of the R/G/B channel for the Overlay
layer.
  • You can use mask for fine tuning on the Overlay layer.
I usually copy & paste the red channel into the Overlay layer. It
has the least detail on skin complexion but plenty of other hard
edges for the Hi-Pass filter to pick up.

Regards,
Alan
Click to expand...
--
Tim
 
pbleic said:
This article is well worth reading. It is by Michael Chaney, the
author of Qimage. The method is incorporated in Qimage, but is a
bit hard to get at. In addition, you can buy it from Outback Photo
as a plug in.

Trust me, this is well worth reading. Different than anything you
have read on sharpening in the past.

http://www.outbackphoto.com/dp_essentials/dp_essentials_05/essay.html

--
Paul

------------------------------------------------
Pbase supporter
Photographs at: http://www.pbase.com/pbleic/photos
--------------------------------------------------
Copyright 2003 All rights reserved.
Click to expand...
Thank you Paul, very helpful. I had Qimage but did not know what that feature did and had not tried it. It seems to be image dependent as to how much effect it has, a welcome enhancement to some photos and no harm to the others.

Can't comment on the stand alone plugin as I haven't tried it. But I only use Photoshop for editing now and always print and size my photos in Qimage.

Thanks Again

--
Regards Rusty1

 
VideoProduction said:
David Jeffery said:
So how about using NIK sharpener in LAB mode.....I find that NIK
sharpener is often less brutal than USM.....and have read that
sharpening in LAB is best.....comments?
Click to expand...
Each time you convert from one mode to another, the image degrades
to some degree. It may not be a concern to many, but can be
objectionable to the purists.
Click to expand...
Not really an issue with 16 bit L*a*b mode, and even less so if sharpening is done however you like and then added as a layer on top of the original image. Doing so twice with one in Lighten and one in Darken is even better as the light/dark halos can be adjusted independently.

And, there's always Luminosity blend mode to simulate L*a*b sharpening, which is also available in PS Elements (no L*a*b).

What Elements won't allow for is sharpening the black channel of CMYK, which can be useful.

A good book to learn about all of this is "Professional Photoshop" by Dan Margulis. It's also fun to read.

http://www.amazon.com/exec/obidos/ASIN/0764536958/radellaf-20
 
Hi, Daniella.

I think the original poster's (Stan) intent was to try to illustrate what Sigma's "3D effect" is supposed to look like. :) But yes, the flamingo and pelican shot looks a little overprocessed.

I see what he's talking about. To me, the portrait shows the 3D effect more effectively than the flamingos.

But...

I'd be interested to see how normal USM on the luminance channel compares to this color EQ USM method on the original images. So far, I've only seen comparison between the original and the color EQ USM'd images.

-j.
Daniella said:
Tim O'Connor said:
Hmm. I guess its personal preference, (perhaps I like softish
looking images?) but the right hand image sucks to my eyes. (like
its been USM'ed way too much)
Click to expand...
I also don't like a lot of sharpening for portraits..they look
better when a tad soft. That was not a good exemple for this
sharpening as it is usualy prefered to have portraits not that
sharp.
Click to expand...
 
OOOH, good tip on using two layers to be able to adjust light and dark haloes independently. I've only used one layer and had to artistically compromise how I wanted the image to look.

Thanks! :)

-j.
Renard DellaFave said:
VideoProduction said:
David Jeffery said:
So how about using NIK sharpener in LAB mode.....I find that NIK
sharpener is often less brutal than USM.....and have read that
sharpening in LAB is best.....comments?
Click to expand...
Each time you convert from one mode to another, the image degrades
to some degree. It may not be a concern to many, but can be
objectionable to the purists.
Click to expand...
Not really an issue with 16 bit L*a*b mode, and even less so if
sharpening is done however you like and then added as a layer on
top of the original image. Doing so twice with one in Lighten and
one in Darken is even better as the light/dark halos can be
adjusted independently.

And, there's always Luminosity blend mode to simulate L*a*b
sharpening, which is also available in PS Elements (no L*a*b).

What Elements won't allow for is sharpening the black channel of
CMYK, which can be useful.

A good book to learn about all of this is "Professional Photoshop"
by Dan Margulis. It's also fun to read.

http://www.amazon.com/exec/obidos/ASIN/0764536958/radellaf-20
Click to expand...
 
...the tailamps...it's easy to get the red lens "right", but then the monochrome back up lights get too much sharpening. Since they are nearly B&W, they are already pretty sharp, but the red lenses are pretty sharp yet both are in the same focal plane. That's part of what hurts that 3D effect..

Stan
 
jjc said:
OOOH, good tip on using two layers to be able to adjust light and
dark haloes independently. I've only used one layer and had to
artistically compromise how I wanted the image to look.

Thanks! :)
Click to expand...
Welcome. I think I got it from Stephen Marsh's website. If not maybe Deke McClelland's Photoshop 7 Bible or Dan Margulis's Professional Photoshop.

Try here:

http://www.amazon.com/exec/obidos/external-search?tag=radellaf-20&keyword=photoshop%20professional&mode=books

Stephen Marsh's website
http://members.ozemail.com.au/~binaryfx/ST_USM1.html
he also has a good action:

"EPLBVHI USM - Edge Protected Luminosity Blend Variable Halo Intensity Unsharp Mask.
USM is applied to high contrast edges and flat areas are ignored.

The read me stop message is disabled by default. A flat file will result. Undo the last step or disable the flatten command in the action to have access to the split light/dark USM halo layers."
http://www.retouchpro.com/forums/showthread.php?t=2767

File posted on my own site (no registration required) if you prefer:
http://homepage.mac.com/renard/ls/auto_eplbvhi_usm.zip
 
It was quite popular when I originally posted it.

http://forums.dpreview.com/forums/read.asp?forum=1031&message=7038550
Renard DellaFave said:
jjc said:
OOOH, good tip on using two layers to be able to adjust light and
dark haloes independently. I've only used one layer and had to
artistically compromise how I wanted the image to look.

Thanks! :)
Click to expand...
Welcome. I think I got it from Stephen Marsh's website. If not
maybe Deke McClelland's Photoshop 7 Bible or Dan Margulis's
Professional Photoshop.

Try here:

http://www.amazon.com/exec/obidos/external-search?tag=radellaf-20&keyword=photoshop%20professional&mode=books

Stephen Marsh's website
http://members.ozemail.com.au/~binaryfx/ST_USM1.html
he also has a good action:

"EPLBVHI USM - Edge Protected Luminosity Blend Variable Halo
Intensity Unsharp Mask.
USM is applied to high contrast edges and flat areas are ignored.
The read me stop message is disabled by default. A flat file will
result. Undo the last step or disable the flatten command in the
action to have access to the split light/dark USM halo layers."
http://www.retouchpro.com/forums/showthread.php?t=2767

File posted on my own site (no registration required) if you prefer:
http://homepage.mac.com/renard/ls/auto_eplbvhi_usm.zip
Click to expand...
--
Paul

------------------------------------------------
Pbase supporter
Photographs at: http://www.pbase.com/pbleic/photos
--------------------------------------------------
Copyright 2003 All rights reserved.
 
Renard DellaFave said:
Renard DellaFave said:
Renard DellaFave said:
Indeed. I wish it was presented as a method with instructions rather than a "black box" Qimage function.
Click to expand...
Click to expand...
Click to expand...
The detailed strategy is discussed here:

http://forums.dpreview.com/forums/read.asp?forum=1027&message=7094401
Renard DellaFave said:
Renard DellaFave said:
I imagine you could achieve something similar by using two levels of USM on two different layers and creating a layer mask for "greens" for one and one for the other hues for the other. Again, it would be nice if new methods were presented in such terms so they'd be generally useful without having to run yet another proprietary plug-in, or worse, yet another proprietary image processing application.
Click to expand...
Click to expand...
He varies the strength of the sharpening according to the contributions from the R, G, B sensors. So areas with values that are equal (monochrome) such as a resolution chart receive no sharpening. And more sharpening depending upon colors and whether 1,2,3 of the three colors contributed.

What he is trying to do is make the focus consistent across the focal plane. So things next to each other of different colors are more nearly seen at the same sharpness. Flower petals versus green stems.

Here's an image where he points out obvious variations in sharpness according to color within the DOF using 10D and Sigma images in the subsequent URL.



http://forums.dpreview.com/forums/read.asp?forum=1027&message=7067679

HTH.

Stan
 
pbleic said:
It was quite popular when I originally posted it.

http://forums.dpreview.com/forums/read.asp?forum=1031&message=7038550
Click to expand...
Well, yes, I do like it...it's basically is what I was suggesting at first, with the addition of the specific USM numbers to use.

Main idea: use lighten & darken on 2 USM'd layers,

possibly with:
Edit-> Fade-> Luminosity
after USM and before duplicate.

To do that, you do need to determine USM settings. They will vary depending on display size, image type, and noise levels. Nothing wrong with .5 500% 0 as a starting point, though.

The Auto EPLBVHI action makes the USM more selectively applied, allowing higher settings without over-emphasizing noise. It includes the Main idea above. It is FAR superior to using a non-zero Threshold setting in Photoshop's USM filer.

--
--- RAD
 
I know what he's trying to do, and there are non-quantitative explanations of the theory, but no explanations of how to play around with this sort of thing within Photoshop.
 
.. he's a nice guy and may be able to give you some suggestions for a PS approach. If something comes up I'd certainly hope to know about it! In the mean time I have the EQ in my qimage.

Stan
 
You must log in or register to reply here.

Keyboard shortcuts

f
Forum
m
My threads
Mobile site
About
Editorial content
Cameras & Lenses
Community
All content, design, and layout are Copyright © 1998–2025 Digital Photography Review All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
When you use DPReview links to buy products, the site may earn a commission.
©GPS Media - Guides, Products, Services.
Back
Top