Sharpening - Part 2 - Separate control over light and dark halos

ruttrutt Registered Users Posts: 6,511 Major grins
edited March 20, 2006 in Finishing School
Introduction

In Part 1, I explained the basics of sharpening and led you through the process of finding good values for USM. Now I'm going to teach you a new technique that will allow you to sharpen more in many cases without over sharpening. This could be described as an advanced technique, but it is the most common enhancement to basic USM there is. Virtually every prepress professional knows this trick, and soon you will, too.

Here is an before/after of an image sharpened using this technique:

19668203-L.jpg

19668405-L.jpg

Notice the how well the illusion of sharpness works even at this size.Look especially at the stonework on the bridge. Yet the branches of the trees don't look over sharpened, which is the most common bugaboo of sharpening.

The idea is to gain independent control over the light and dark halos.Usually the light halos are more easily detected. In order to prevent visible light halos, we are restrained to more conservative USM values. But once we can independently specify the values for the light and dark halos, we can make the dark halos do their job without making the light halos give away the trick.


Quick start

The basic idea is really simple. Make 2 layers, one for the dark and and one for the light halos. We can do this by setting the blending options of the layers, one to Lighten and one to Darken. Then we can control the opacity of the light and dark halos independently by changing the opacity of each layer separately. I'll walk through the process that produced the image above in some detail, but in essence here are the steps.
  1. Make a duplicate layer.
  2. Sharpen the L channel of the duplicate layer in LAB as in Part 1, but leave the Amount set to 500. In other words, don't complete the final step.
  3. Convert the image to RGB.
  4. Make an additional duplicate layers. For now make the top layer invisible and the middle layer visible.
  5. Set the blending mode of the middle layer to Darken.
  6. Tune the opacity of the dark halos with the opacity slider of the middle layer.
  7. Set the blending mode of the top layer to Lighten and make it visible
  8. Tune the opacity of the light halos with the opacity slider of the top layer.

Walk through

I started out by following my basic recipe for tuning USM Threshold and Radius as described in Part 1, with two exceptions:
  1. I made a duplicate layer and sharpened it, and
  2. I left Amount at 500.
When I was done, my USM dialog looked like this:

19668942-S.jpg

And my image looked like this:

19668489-L.jpg

This looks pretty bad, but don't worry. The sharpening opacity is just too great an about to fix that.

Next I moved the image into RGB, made an additional duplicate layer, made the top layer invisible, and set the blending mode of the middle layer to Darken. Like so:

19668927-S.jpg

At this point the image looked like this:

19668501-L.jpg

Now there are no light halos at all and the dark halos are as dark as they can be. It's easy to see here how much the light halos contribute to the over sharpening of this image, especially in the tree branches. Look closely at this image, though, The dark halos aren't as noticible as the light ones, but at 100% opacity, they are very evident.

19676170-L.jpg

I turned do the opacity slider of the Darkenlayer until the halos nearly disappeared and that backed off so that I could see their effect. I ended up at 35%. This will definitely vary depending on image.

19668936-S.jpg

At this point a 100% zoom looked like this:

19676162-L.jpg

At this point the full image looked like this:

19668526-L.jpg

A strange effect without the light halos, eh? But interesting to see just the dark halos. Usually the light halos so overwhelm the dark ones that they are all we see.

I set the blending option for the top layer to Lighten and made it visible:

19668931-M.jpg


It looked bad. The 100% light halos overwhelmed the image and it looked virtually identical to how it looked before I tuned the dark halos.But now I could complete the tuning by turning down the opacity of the light layer until the image looked good. I used the layer eyes to compare the image with and without the light halos. I arrived at a value of 12% for the opacity of the light layer. I often find that the right opacity for the light halos is about 1/2 that used for the dark ones, but again this is image dependent.

19668934-M.jpg


At this setting the image was clearly sharper, but the halos were not visible. The tree branches which would have made such a mess if I'd used the same opacity for light and dark halos now look OK.The result was the after image from the top of the tutorial.

One nice thing about this approach is we aren't really done until we flatten the image. The layers are still there and we can make them visible and invisible and tune their opacity.


Advanced variation

Not that advanced, really once you've digested the basic technique of having separate layers for light and dark halos. Why limit ourselves to the same radius for the light and dark halos? Duplicate before USM and use different Radius and even Threshold values for each of the two layers. This takes a little practice, since you won't actually be able to tell how it's all going to look until you put it together. But it can often be invaluable. A smaller Radius for the light halos in the above image might have been a good thing because the light halos live in the fine tree branches and the dark halos live in the far coarser stonework on the bridge.


Moral

The idea of sharpening is to get away with exactly as much as possible but no more. Often the light halos are the limiting factor. By separating them, we can use much higher values for the dark halos, creating a far greater illusion of sharpness without being detected. (And you thought it was just that great new lens, eh, mate?)
If not now, when?

Comments

  • ruttrutt Registered Users Posts: 6,511 Major grins
    edited April 14, 2005
    I fixed a simple but very basic problem in the description of the procedure. You have to make a duplicate layer before sharpening and sharpen it. It's very important to have the original unsharpened image as a background!
    If not now, when?
  • cletuscletus Registered Users Posts: 1,930 Major grins
    edited April 15, 2005
    Great work, rutt thumb.gifclap.gif
  • ginger_55ginger_55 Registered Users Posts: 8,416 Major grins
    edited April 15, 2005
    You all, I plan to do this, just not today that I know of. I am very excited about the second part of this, but my body kind of shut down after stressful doctor visits yesterday, etc.

    And I am so excited that in one yr from June my daughter and family are moving to south carolina, 2 hrs from me. It is about all I can think about. I keep thinking I am going to announce it here, but haven't decided on people or wide angle. Sara, daughter, her husband and two grade school grandkids. I just can't concentrate on sharpening. I am going to do it, though. I am not uninterested. I am very interested.

    ginger
    After all is said and done, it is the sweet tea.
  • DJ-S1DJ-S1 Registered Users Posts: 2,303 Major grins
    edited April 15, 2005
    rutt wrote:
    I fixed a simple but very basic problem in the description of the procedure. You have to make a duplicate layer before sharpening and sharpen it. It's very important to have the original unsharpened image as a background!
    Oh good, I thought I was just being extra dumb! I couldn't see any difference the other way and I thought I was doing something wrong. Thanks for the tutorials, I know they take a lot of time to put together. thumb.gif
  • thomas1973thomas1973 Registered Users Posts: 1 Beginner grinner
    edited May 16, 2005
    Excellent tutorial! thumb.gif

    Just to add an extra tip:

    For getting the maximum sharpening for a particular image, I start out by using the technique described here. But sometimes parts of the image needs more sharpening than others. I adjust the USM values and the layer opacity for the "Darken" and "Lighten" layer to match the optimum sharpness for the parts of the image needing the most sharpening. So far it follows the original tutorial, but...

    Then you can add a layer mask to each of the sharpening layers (by being on the layer in question, and clicking on the square with the circle inside, at the bottom line of the layer's palette). Pick a black brush with a soft edge. Adjust the brush opacity (around 30% is often good), then paint over the areas that need less sharpening, do another pass to further reduce sharpening if needed.

    Great for images that need different sharpening in different areas :-)


    Thomas.
  • wanderingeyewanderingeye Registered Users Posts: 43 Big grins
    edited June 15, 2005
    Rutt (and Thomas),

    Thanks so much for your help with this, I really appreciate it!
    David Cothran (Wandering Eye)
    http://www.wandering-eye.com

    "the days run away like wild horses over the hills"
    Bukowski
  • colourboxcolourbox Registered Users Posts: 2,095 Major grins
    edited June 15, 2005
    It looks as if this tone-zoned sharpening is what's going on behind the new Smart Sharpen in Photoshop CS2, when you put it in Advanced mode. You get separate control over highlight and shadow sharpening, including Amount, Radius, and Fade. Would be interested in your thoughts on Smart Sharpen.
  • ruttrutt Registered Users Posts: 6,511 Major grins
    edited June 15, 2005
    colourbox wrote:
    It looks as if this tone-zoned sharpening is what's going on behind the new Smart Sharpen in Photoshop CS2, when you put it in Advanced mode. You get separate control over highlight and shadow sharpening, including Amount, Radius, and Fade. Would be interested in your thoughts on Smart Sharpen.

    Nothing you can't do with the techniques I outlined. Perhaps a little more convinient. The important thing is knowing what's really going on.
    If not now, when?
  • PlasmodiumPlasmodium Registered Users Posts: 65 Big grins
    edited September 21, 2005
    Thanks, Rutt.

    What I've been doing, which may or may not make sense, is to use a similar technique with SmartSharpen instead of USM. I sharpen the 'lighten' and 'darken' duplicate layers twice each -- once with smartsharpen set to remove lens blur and once with it set to remove gaussian blur. Then as you suggest I vary the opacity and use layer masks to fine tune.

    Now I do have one question. When I sharpen in RGB I do a 100% fade to luminosity. How does this differ functionally from making the change to the L channel of Lab? I'd presume they're the same, because the changes are restricted to luminance either way, but I'm not so sure about the mathematics of it.
    Paul

    My Gallery

    "Hammer my bones in the anvil of daylight..." -Beck
  • XO-StudiosXO-Studios Registered Users Posts: 457 Major grins
    edited September 21, 2005
    Plasmodium wrote:
    Thanks, Rutt.

    What I've been doing, which may or may not make sense, is to use a similar technique with SmartSharpen instead of USM. I sharpen the 'lighten' and 'darken' duplicate layers twice each -- once with smartsharpen set to remove lens blur and once with it set to remove gaussian blur. Then as you suggest I vary the opacity and use layer masks to fine tune.

    Now I do have one question. When I sharpen in RGB I do a 100% fade to luminosity. How does this differ functionally from making the change to the L channel of Lab? I'd presume they're the same, because the changes are restricted to luminance either way, but I'm not so sure about the mathematics of it.
    The main difference is in the banding that leads up to the sharpening. Lab can do the calculations by going to an imaginary color, while RGB will immediately have to default to white, which rather than result into sharpening, it leads to a white halo in lighter colors. Look for more on this in the upcoming chapters of the review of D Margulis' book.

    XO,
    You can't depend on your eyes when your imagination is out of focus.
    Mark Twain


    Some times I get lucky and when that happens I show the results here: http://www.xo-studios.com
  • Klaatu Baradda NiktoKlaatu Baradda Nikto Registered Users Posts: 4 Beginner grinner
    edited March 19, 2006
    Look Ma'! No Mode Changes!
    Greetings All.
    This is a very informative forum with very helpful individuals.

    Concerning this very interesting sharpening approach, here's a bit of a shortcut to achieve similar results without having to change color modes.

    Run the USM twice and immediately after each run go to Edit>Fade.... Set the Fade of the first run to Darken and the second run to Lighten. Adjust the Opacity of each Fade command to taste.

    Variations of this technique could include separate layers for each USM run (or at least a single layer with both runs) but that might be overkill since Opacity can be controlled through the Fade command.

    Very interesting and clever approach Rutt. Thanks for sharing it.
  • SystemSystem Registered Users Posts: 8,186 moderator
    edited March 19, 2006
    not that I'm the official committee, but welcome!-

    george

    ps you got carpenter with you?-
  • PlasmodiumPlasmodium Registered Users Posts: 65 Big grins
    edited March 19, 2006
    Run the USM twice and immediately after each run go to Edit>Fade.... Set the Fade of the first run to Darken and the second run to Lighten. Adjust the Opacity of each Fade command to taste.

    This won't work in LAB mode, and the main point of this thread was to illustrate the advantages of using LAB for sharpening. In LAB mode there is no 'Lighten' or 'Darken' option either in the fade or layer blending options -- they're grayed out. So you have to convert to RGB preserving the layers in order to do it.

    The other disadvantage doing it with fade instead of on a layer is that it prevents you from having a layer in which you can vary your opacity or selectively mask out certain areas. For instance high contrast edges get sharpened more heavily than low contrast edges; so a standard sharpening run is likely to oversharpen your really high contrast edges, so it's nice to be able to selectively mask these oversharpened edges. A fade by itself doesn't give you that option.

    Finally, the other reason I would avoid using fade is that it doesn't allow you to differentially compare the lighten and darken layers. In other words you can't turn one on and off and then the other on and off to compare them with one another. With fading the process is just plain done when its done and you have to just accept the results.

    Here's the easiest way to do it -- just create an action triggered by a function key. I press F2 and the whole thing is done instantly, conversions to LAB and back to RGB and all -- and the layers are named, set to their blending modes, and they have layer masks. That's how to get the best of all worlds.
    Paul

    My Gallery

    "Hammer my bones in the anvil of daylight..." -Beck
  • Klaatu Baradda NiktoKlaatu Baradda Nikto Registered Users Posts: 4 Beginner grinner
    edited March 19, 2006
    Greetings Plasmodium and thanks for the well thoughtout reply.
    Plasmodium wrote:
    This won't work in LAB mode, and the main point of this thread was to illustrate the advantages of using LAB for sharpening. In LAB mode there is no 'Lighten' or 'Darken' option either in the fade or layer blending options -- they're grayed out. So you have to convert to RGB preserving the layers in order to do it.

    Understood (didn't miss that point). However, you actually have to "convert" to Lab FIRST.

    The method outlined was intended as an alternative to converting to LAB then back to RGB.
    The other disadvantage doing it with fade instead of on a layer is that it prevents you from having a layer in which you can vary your opacity or selectively mask out certain areas. For instance high contrast edges get sharpened more heavily than low contrast edges; so a standard sharpening run is likely to oversharpen your really high contrast edges, so it's nice to be able to selectively mask these oversharpened edges. A fade by itself doesn't give you that option.

    If you read the post again you will see the issue of using separate layers for each run (to allow control of Opacity after-the-fact) is addressed (although, admittedly without much detail).
    Finally, the other reason I would avoid using fade is that it doesn't allow you to differentially compare the lighten and darken layers. In other words you can't turn one on and off and then the other on and off to compare them with one another. With fading the process is just plain done when its done and you have to just accept the results.

    This is addressed similarly above. If you use separate layers for each run, you will have accomplished the same thing without leaving RGB (including the ability to compare effects). See below for a more robust outline of this RGB layering method.
    Here's the easiest way to do it -- just create an action triggered by a function key. I press F2 and the whole thing is done instantly, conversions to LAB and back to RGB and all -- and the layers are named, set to their blending modes, and they have layer masks. That's how to get the best of all worlds.

    The steps for such an action would be even fewer (faster) without the changing of modes. No?

    To be clear (assuming that one would want to pursue this "layered technique" in RGB) they would need to do the following:

    1) Create 2 duplicate layers of the Background and turn the top layer off, then select the 2nd layer and set it's Blend Mode to Darken.
    2) Run the USM, set the Fade command to Luminosity. Don't touch the Fade's Opacity.
    3) Select the top layer and set it's Blend Mode to Lighten.
    4) Run the USM, again set the Fade command to Luminosity. Again, don't touch it's opacity.

    From here you have 2 layers that have separated the dark and light halos which can be adjusted with their respective Opacitiy sliders. Or, if you so desire, a Layer Mask can be added for further control.
  • dandilldandill Registered Users Posts: 102 Major grins
    edited March 20, 2006
    ...

    1) Create 2 duplicate layers of the Background and turn the top layer off, then select the 2nd layer and set it's Blend Mode to Darken.
    2) Run the USM, set the Fade command to Luminosity. Don't touch the Fade's Opacity.
    3) Select the top layer and set it's Blend Mode to Lighten.
    4) Run the USM, again set the Fade command to Luminosity. Again, don't touch it's opacity.
    Thanks for this, which seems neat.

    Dan
    Dan Dill

    "It is a magical time. I am reluctant to leave. Yet the shooting becomes more difficult, the path back grows black as it is without this last light. I don't do it anymore unless my husband is with me, as I am still afraid of the dark, smile.

    This was truly last light, my legs were tired, my husband could no longer read and was anxious to leave, but the magic and I, we lingered........"
    Ginger Jones
  • Klaatu Baradda NiktoKlaatu Baradda Nikto Registered Users Posts: 4 Beginner grinner
    edited March 20, 2006
    Mmm... Yummy!
    Okay, while eating giant helpings of humble-pie, we retract our input.

    The RGB color space DOES in fact ADD & SUBTRACT from the luminosity of an image. Therefore, running the USM first in RGB and then Fading it to Luminosity will add or subtract the subtle color luminance from an image. This, unfortunately, leaves you with halos (though still much more subtle than purely running USM in RGB mode alone). This is something Lab mode completely avoids.

    Faster though our steps may be, Rutt's method is still KING!

    Thanks again for sharing this, Rutt.
Sign In or Register to comment.