Reduce (USM) sharpening halos

Moderator: jsachs

tomczak
Posts: 661
Joined: April 25th, 2009, 12:56 am

Reduce (USM) sharpening halos

Postby tomczak » October 25th, 2011, 6:51 pm

Composite, in Soft Light mode, the input image with a Negative of High Pass filtered version of it.
Maciej Tomczak
Phototramp.com

tomczak
Posts: 661
Joined: April 25th, 2009, 12:56 am

Re: Reduce (USM) sharpening halos

Postby tomczak » October 25th, 2011, 9:08 pm

An example on a synthetic test image: 1=original, 2=oversharpened, 3=recovered.
Attachments
HaloRecovery1.jpg
Original
HaloRecovery1.jpg (11.83 KiB) Viewed 23848 times
HaloRecovery2.jpg
Oversharpened
HaloRecovery2.jpg (13.27 KiB) Viewed 23843 times
HaloRecovery3.jpg
Recovered (from Oversharpened, without Original)
HaloRecovery3.jpg (12.67 KiB) Viewed 23846 times
Last edited by tomczak on October 25th, 2011, 9:43 pm, edited 1 time in total.
Maciej Tomczak
Phototramp.com

Bob Walker
Posts: 47
Joined: April 25th, 2009, 9:08 am
What is the make/model of your primary camera?: Canon 7DII 5DIII
Location: Los Alamos, New Mexico
Contact:

Re: Reduce (USM) sharpening halos

Postby Bob Walker » October 25th, 2011, 9:15 pm

Maciej,

Do me a favor and compare this method with one I have used (was this a suggestion of den's maybe?).

Oversharpen the image as in your first step (producing image 2 -- maybe use unsharp mask with radius=3), then composite the oversharpened image in "darken" mode with the original. It takes away the bright half of the halos and leaves the dark half, which is usually the much less objectionable side effect of oversharpening. For me, the final result is more subtle than your interesting tip, but may be another tool in the toolbox.

Thanks,
Bob W

tomczak
Posts: 661
Joined: April 25th, 2009, 12:56 am

Re: Reduce (USM) sharpening halos

Postby tomczak » October 25th, 2011, 9:24 pm

Bob,

These are two different techniques - both useful:

1) the one that you mention works if you are doing the USM, and want to reduce halos produced as a result of it, by reducing the white portion of halos and leaving the dark portion intact. You need the original, unsharpened, image to do that.

2) The one I propose is for when you have an image already (over)sharpened, with halos that you want to reduce, but don't have the original anymore.
Maciej Tomczak
Phototramp.com

tomczak
Posts: 661
Joined: April 25th, 2009, 12:56 am

Re: Reduce (USM) sharpening halos

Postby tomczak » November 8th, 2011, 10:49 pm

It looks like I'm at least a few years behind the curve...

http://www.dl-c.com/discus/messages/523 ... 1239927925
Maciej Tomczak
Phototramp.com

tomczak
Posts: 661
Joined: April 25th, 2009, 12:56 am

Re: Reduce (USM) sharpening halos

Postby tomczak » March 13th, 2014, 10:57 am

Just an illustration of reducing sharpening halos on major edges by removing (or reducing) either black or white overshoots (technique #1 above). Removing white overshoots is done by Compositing unsharpened original with sharpened version of it in Darken mode (thus leaving the dark overshoots only).

Here it's done globally (i.e. all white overshoots are removed: on major edges where there are objectionable, and elsewhere, where removing them probably decrease overall visual sharpening effectiveness somewhat), but this can be refined by limiting overshoot removal to major edges by edge masking as suggested here:

viewtopic.php?f=5&t=994

This technique is pretty universal as all PWP's sharpening methods (with the exception of Bilateral Sharpen whose hard threshold can protect roughest edges from sharpening) will produce some sort of overshoots that may become annoyingly visible on well defined edges contrasted with smooth background of a different tone (e.g. dark hair and skin against lighter asphalt).

Since reducing overshoots on one side of the edge (most often white) can't be done interactively during sharpening (you need to sharpen first, then reduce one of the overshoots), it's sometimes difficult to judge how much sharpening to apply in the first step to optimize the process and make the image look decent at the end.
Attachments
Sharpened Minus White Halos.jpg
Sharpened Minus White Halos.jpg (13.2 KiB) Viewed 21841 times
Sharpened.jpg
Sharpened.jpg (15.02 KiB) Viewed 21832 times
Original Unsharpened.jpg
Original Unsharpened.jpg (11.37 KiB) Viewed 21827 times
Maciej Tomczak
Phototramp.com

den
Posts: 612
Joined: April 25th, 2009, 6:33 pm
What is the make/model of your primary camera?: Canon EOS-350D/Fuji X100T
Location: Birch Bay near Blaine, WA USA

Re: Reduce (USM) sharpening halos

Postby den » March 13th, 2014, 5:15 pm

Maciej... ...Others

I tend to control sharpening edge undershoot/overshoot with an 'active' default Mask Tool - Brightness curve applied to either the 'starting image' or its Luminance channel... ...then use that mask interactively with whatever sharpening transform being employed... ...Mask black amount controls the edge undershoot and mask white amount controls the edge overshoot based upon the image's globally mapped HSV-Values or Luminance gray tones...

The attachment below illustrates USM where the undershoot edge sharpening is set to a 70% amount preference with the overshoot edge sharpening set to a 20% amount preference to eliminate objectionable white major edge halos... ...preferences are set while viewing Preview at 1:1 Zoom image areas to avoid anti-aliasing artifacts.

Note... I favor the 'active' Luminance channel mask as most sharpening methods occur in the RGB color space as is the case for USM... ...but this is a nuance that will generally go un-noticed for most photo-realistic images.

This may be a more direct equivalent approach then a Composite-Darken, followed by a Composite-Lighten transformation sequence where an over-sharpened image version of the 'starting image' is used as the Overlay.

...den...
Attachments
capture_13032014_115531.jpg
capture_13032014_115531.jpg (47.96 KiB) Viewed 21839 times

tomczak
Posts: 661
Joined: April 25th, 2009, 12:56 am

Re: Reduce (USM) sharpening halos

Postby tomczak » March 16th, 2014, 5:48 am

I've experimented some with Den's sharpening through an interactive brightness mask. It often works in reducing white halos on rough edges - and, importantly, it's easy to adjust and preview the results interactively.

The reason I think it works is that the white overshoots are on the brighter side of an edge and the black halos on the darker side, so the y=-x brightness Amount mask tends to emphasize black halos and de-emphasize white halos regardless of what brightness zone the edge (detail) is in. There is always a brightness gradient across the edge, and the brightness amount mask takes advantage of it.

There is a word of caution though: such inverse brightness mask will also regulate sharpening amount globally and will reduce overall sharpening (both black and white overshoots) of details in brighter zones of the image and maintain overall sharpening in the darker ones. That may be desirable sometimes, but makes it difficult to fine-tune white/black overshoots only and decouple them from the image tone: it may be difficult to reduce white overshoots on major edges in darker zones and properly sharpen fine details in brighter zones.
Maciej Tomczak
Phototramp.com

tomczak
Posts: 661
Joined: April 25th, 2009, 12:56 am

Simulating AS upper roughness threshold - reducing major hal

Postby tomczak » March 24th, 2014, 2:28 am

Here is an example of trying to reduce major edge halos in AS. In effect, this technique tries to sharpen only lower (or middle) roughness ranges and protect the major edges from sharpening (and thus prevent or reduce the associated dreadful major halos).

The method is a little laborious and not quite interactive - which makes adjusting sharpening parameters and thresholds a bit of a guesswork, but I think it is quite an improvement over either masking the major edges (which may not get the right, halo-producing edges or mask too much and degrade sharpening effect) or Den's brightness gradient masking (which may affect global sharpening by brightness, regardless of edge roughness). The reason that the roughness - major or minor - has the same AS-definition and allows for smooth transitions between sharpened and unsharpened elements of the picture.

It works by:
1) AS sharpening the image with no threshold (or with some lower threshold to exclude noise from sharpening)
2) AS sharpening only the major edges by moving the lower threshold sliders to the far right of roughness histogram
3) Subtracting 2) from 1), thus leaving only lower roughness over/under shoots
4) Compositing the original, unsharpened image with the 3) in Soft Light mode.

p.s. AS, unlike Bilateral Sharpen, can use soft roughness thresholds (by separating white and black threshold sliders) to further improve the appearance of the sharpened image. Hard thresholds were use in this example. Sharpening and threshold parameters used are probably not ideal, but sufficient to illustrate the technique. Both white and black halos on major edges are reduced - targeting only the white overshoots can further optimize the sharpening, but, if done by hand, further complicates the technique.
Attachments
AS Major Halos Reduction.jpg
Original --> AS sharpened all --> AS sharpened major edges only --> Offset difference between AS1 and AS2 --> Composite of Original with the AS difference in Soft Light
AS Major Halos Reduction.jpg (43.41 KiB) Viewed 21610 times
Maciej Tomczak
Phototramp.com

den
Posts: 612
Joined: April 25th, 2009, 6:33 pm
What is the make/model of your primary camera?: Canon EOS-350D/Fuji X100T
Location: Birch Bay near Blaine, WA USA

Re: Simulating AS upper roughness threshold - reducing major hal

Postby den » March 26th, 2014, 1:34 pm

Maciej... ...Others

A big thank you for taking the time in off message board discussions regarding your technique to limit major edge halos with the Advanced Sharpen - Sharpen tab. Off/on for the past two years since you first suggested adding upper limiting threshold/blending in addition to the existing threshold/blending black/white sliders... ...I have been struggling to understand.

Now I do and I find the technique extremely useful for developing the micro-contrasts I prefer for 'gallery quality' images...

The following was shared with me and I pass on to interested Others along with my own added comments...
(1) Maciej's AS1 settings...
AS1.jpg
AS1.jpg (48.48 KiB) Viewed 21521 times

(2) Maciej's AS2 settings...
AS2.jpg
AS2.jpg (47.98 KiB) Viewed 21527 times

(3) the Offset Difference may be found for PWP7 under Transformation/Special Effects/Difference transform where Input = AS1; Subtract image = AS2; Operation = Offset Difference; and Amount = 100... ...press F1 Help when the transform is 'active' for an explanation...
(4) My preference is not so conservative as Maciej's so I prefer to use a Composite - Hard Light blend with a reduced amount that would normally be applied if Soft Light were used.

Does this technique work? Indeed it does... ...in the following illustration, a 3456x2304 pixel image of the Lower Yellowstone River Falls & Canyon was Resized using the Bicubic method to a width of 400 pixels [top image] which of course was softened... ...Maciej's Advanced Sharpen-Sharpen technique was employed with settings suitable for the 400w pixel image resulting in the bottom image with no objectionable dark/white edge halos and fine detail restoration...
IMG_0991-3_AS1-AS2OffDiff_80HardLite.jpg
IMG_0991-3_AS1-AS2OffDiff_80HardLite.jpg (47.95 KiB) Viewed 21532 times
...It is really difficult to fully appreciate the improvement at message board posting resolutions for full sized image... ...Be sure to try this yourself and pixel peek an image on your own monitor...

Thank you again Maciej...

...den...


Return to “Picture Window Tips”

Who is online

Users browsing this forum: No registered users and 1 guest