Topics

question about sharpening


Gerald Bakker
 

Through my website, someone asked me the following question about sharpening. 
I could not find a satisfying answer myself, so I hope that someone here is able to help him (and me) further.

=====================================================================================
Below is my current sharpening practice (edge mask sharpening in 2 phases):
 
* Phase 1: Unsharp mask sharpening with strong settings to enhance Major and Large-scale Details around edges on Darken blend mode layer and Lighten blend mode layer respectively.
* Phase 2: A weak USM sharpening to better reveal Fine and Small-scale Details residing on non-edge areas besides sky-like smooth surfaces on top of the Phase 1 output layer.

Honestly, I don't know how to target "Fine and Small-scale Details residing on non-edge areas".
=====================================================================================

Some further background info: this sharpening method is elaborated in a few articles written by a Ron Bigelow. In these, he wrote about sharpening Phase 2 as follows:
     - Look for areas where Brightness (HSB readouts on Info panel) differences between bright and dark spots close to each other are small (where fine details are);
     - On these areas, apply a weak Unsharp Mask sharpening to let fine details reveal and not give negative impact on details residing on edge areas.

It seems that Ron Bigelow's articles are no longer available online, at least the one where this 2-phase sharpening is covered.

So, two questions really:
1. How to construct a mask revealing "fine details" but excluding real/major edges?
2. Is anyone familiar with this two-phase approach or can provide a copy of Ron Bigelow's sharpening articles?

Thanks,
Gerald Bakker
http://geraldbakker.nl


kirkhere@...
 

I cannot answer your sharpening question BUT I can point you to the internet archive site, which has most of Ron's web site and articles here:
http://web.archive.org/web/20180413211305/http://ronbigelow.com/articles/articles.htm


Gerald Bakker
 

Hi, thanks. Someone else mailed me with the same tip.
That answered my second question. The first of a series of 6 articles can be found here:


At the bottom of that page, there is a link to the second one, etc.
The crucial instruction about the second sharpening phase seems to be in part five, and reads as follows:

"Two Pass Sharpening

Edge masks also open up the opportunity to utilize a two pass sharpening approach. In the first pass, edge masks are used to sharpen the more significant edges in the image. Since only the edges are being sharpened, larger sharpening amounts can be used. This will make the major detail really come to life. A second sharpening can then be carried out on a separate layer. This sharpening does not use an edge mask. Rather it is aimed at sharpening the fine detail in the image. For this purpose, a fairly low level of sharpening is applied."


Maybe that answers the first question too: just apply a bit of sharpening on the full image, to target small detail that was masked out by the edge mask of the first pass.


Gerald Bakker

http://geraldbakker.nl



George Machen
 

(I am quoting below almost the entire earlier post, usually against good netiquette, because it represents conventional wisdom that Dan bucks.)

Dan has explained from time to time that USM suffers an often-undesirable inherent flaw of sharpening most strongly precisely the detail in an image that already is quite sharp. This can be seen theoretically when applying USM to a gradient wedge. (Take a gradient from 0% to 100%, cut it in half horizontally, then flop the top half to produce the center representing an edge running from completely distinct to none at all.) Here is an example of such a USM-sharpened gradient wedge:

<https://i.imgur.com/EJJ5sfV.jpg>;

Notice that USM increasingly applies the greatest Amounts and Radii to the already-sharpest edges. Years ago on this list, Dan put out a call for methods to accomplish a "reverse-Threshold" (a control that would do the opposite of USM's Threshold, that would instead minimize sharpening for strong edge contrast). One little-noted of the many advantages of the halo maps of Dan's PPW panel sharpening function is the ability to do just that. In Dan's MPCW, p.. 219, 3rd full ¶, he describes how to curve a halo map to effect a reverse threshold.

This technique could avoid that crunchy look one often gets with many images when strong sharpening through an edge mask is used with that two-pass method.

-- George Machen

On Aug 25, 2019, at 3:07 PM, gc.bakker@hccnet.nl [COLORTHEORY] wrote:

...The first of a series of 6 articles can be found here:

<https://web.archive.org/web/20180420095408/http://ronbigelow.com/articles/sharpen1/sharpen1.htm>;

At the bottom of that page, there is a link to the second one, etc.
The crucial instruction about the second sharpening phase seems to be in part five, and reads as follows:

"Two Pass SharpeningEdge masks also open up the opportunity to utilize a two pass sharpening approach. In the first pass, edge masks are used to sharpen the more significant edges in the image. Since only the edges are being sharpened, larger sharpening amounts can be used. This will make the major detail really come to life. A second sharpening can then be carried out on a separate layer. This sharpening does not use an edge mask.. Rather it is aimed at sharpening the fine detail in the image. For this purpose, a fairly low level of sharpening is applied."

Maybe that answers the first question too: just apply a bit of sharpening on the full image, to target small detail that was masked out by the edge mask of the first pass.

Gerald Bakker

http://geraldbakker.nl


Beat C
 

> Gerald wrote :
>> 1. How to construct a mask revealing "fine details" but excluding real/major edges? 

A bit late, I was too busy with the other thread that I started. 

How about the following : 

original layer
duplicate org layer on top with heavy USM and a threshold that is not too low
set this USM layer into Difference mode
make a stamp layer (“Edges") from these two layers (keeping the two). It will reveal only where the two layers differ, being the major edges.
put the USM layer back to Luminosity (or Normal)
copy original layer again on top and add a layer mask
Apply image : load the rgb, or any of the 3 channels of the stamp (“Edges") into the layer mask. Choose the channel wisely :-) 
Invert the layer mask (or do the inversion with Apply image). The edges get dark, the rest light.
Don’t forget to switch off the stamp layer (“Edges")
Apply sharpening to this layer (with the mask) with mild USM settings (for the details enhancing).
Now apply a strong contrast enhancing curve to the layer mask. Where the separation between black and white in the layer mask will be set, determines the character and range of the details sharpening
The black in the mask will make that the first sharpening (the real edges) are left alone.

Beat


Beat C
 

An additional step might be to use the original mask “Edges" (before inverting) as a layer mask on the layer that has the heavy USM for the edges. The mask (uninverted) lets only the sharpened edges through. Although I have already anticipated this by setting the Threshold fairly high in the first USM, this might add something.

Beat