Untitled Document

While I happily blunder back and forth across color spaces looking for something that works, some of the forum members prefer to work exclusively in RGB. For those of you who do, you might find the following approach a little intriguing. With a little effort, who knows how much this can be expanded and improved?

On several occasions Dan has talked about the benefits of sharpening the Black channel in CMYK. And of course he's right. It does do a great job, especially with portraits and the like. Yes, I know, so please don't remind me that Dan has also proposed working with RGB channels in a similar manner. But I couldn't help but wonder if the key to the success of the Black channel method was not simply the fact that the Black channel is the lightest of the four. In that case, what will happen if we sharpen the lightest RGB channel? Works sometimes, doesn't work all that well other times. But why? Maybe because the contrast is not always quite right. Unlike the Black channel, the Red and Green channels may or may not be light enough to get the job done. OK, so why not make them into what we want instead of what we have? So I duplicated the background layer, sharpened the lightest layer (Red or Green, never Blue) then created a Curves Adjustment Layer and grouped it with the duplicate. Both the duplicate and the adjustment layer were set to Luminosity mode. Next I used the unconventional settings of 50 for the Amount and 50 for the Radius-unconventional because I did it for every image I tested, from grassy fields to portraits. Now I had a tool to adjust the mediocre results. The Hiraloam settings gave a dramatic sharpening effect. By editing the Luminosity curve I was able to get a considerable amount of variation in that effect. Not only that but I also had the utility of two different opacity settings (the adjustment layer, and the duplicate layer).

Think about it. No dramatic color shifts, no dark or light halos. Just adjustable sharpening. To my surprise it even worked with John Ruttenberg's image of the brick factory with the dried brush in the foreground.

But most satisfying of all was a field of wild ryegrass punctuated with stands of lavender flowers. That grass just about popped off of the screen, yet it didn't look at all artificial. The once soft grass looked about as real as grass in a photo can look.

So what's my point? I can't concentrate on one thing all that long when there's so much to explore. Some of you will no doubt find far better ways to create natural sharpening in your images, possibly based loosely on these observations. I'm eagerly looking forward to some new ideas. What I did was not based on rules or theories, so please feel free to twist and bend anything as much as you like. The image is the thing, not the way we get there.

Howard Smith

___________________________________________________________________________

Nice one!

Werner Tschan

--

STUDIO LTD

Atelier für Fotografie

Altenbergstrasse 8

CH-3013 Bern

T: ++41(0)31 332 88 33

F: ++41 (0)31 331 62 42

M: ++41(0)79 227 02 19

U: www.studio-ltd.com

___________________________________________________________________________

Howard,

What you are describing below sounds very logically thought-through and effective, and this isn't meant to sound like a soap commercial - it is a pure technical comment that I really believe. Before I bought PK Sharpener I tried all kinds of sharpening approaches using Lab luminosity, the weakest channel in CMYK, High Pass Sharpening, and you name it. After testing PK Sharpener over a wide range of images and image correction issues needing either various kinds of sharpening or smoothing or a combination thereof, I haven't felt it worthwhile doing anything else, because the ratio of quality to ease with that plug-in is so high. Of course, HIRALOAM should be distinguihed from sharpening per se - while the latter is meant to simulate an overall resolution improvement by emphasizing edge contrast throughout the image (usually), HIRALOAM is often more intended to enhance mid-tone contrast. When done on a duplicate image layer and combined with a blending option that actually confines its impact to mid-tones (as demonstrated by Katrin Eismann in a recent issue of PhotoshopUser magazine), it corrects a tendancy of HIRALOAM to cause low and high end clipping of an otherwise tightly positioned histogram.

Mark Segal

___________________________________________________________________________

"Mark Segal" wrote:

I have not seen the article: can you please describe the blending option that would constrict HIRALOAM to midtones? I would have thought of using a mask instead

Bill Theis

___________________________________________________________________________

Bill, sure. This appeared in PhotoshopUser magazine December 2006, page 039 and written by Katrin Eismann. She duplicates the background layer and implements Amount/Radius/Threshold at 50/20/0. Then she opens Blending Options and in the "This Layer" section drags the shadow slider to 70 and the Highlight slider to 185. Then she separates the triangular sliders and drags the shadow point to 0 and the highlight point to 255.

I find it so generally useful that I created an action of the whole process; where I see I need some livening-up of the mid-tons I launch the Action, then examine the outcome and see whether any further adjustments are necessary.

Mark Segal

___________________________________________________________________________

first, thanks Mark for the Karen Eismann midtone blending. It's really obvious now that I see what it is...

so Howard, could you please add just a bit more detail? I am very interested in your method for sharpening with curves but have just a bit of confusion:

--- Howard Smith" wrote:

OK here is where my CMYK confusion sets in: by lightest Black channel you mean that the ink is low (like below 50%). The lightest RGB channel means to mean that it is mostly above 128 (which means it would look like an extremely heavy black channel)... or do you mean that you want the channel to be below 128 and "look" more like the black channel?

next question: what do you mean by "group"? I could see selecting the now sharpened Red or Green channel and Image>Adjustments>Curves to it but I take from the rest of the description that this is not it. Do you make a "Set" (the little folder where you can drag layers) and put the layer with the sharpened channel in it and on top of this a curves layer?

presumably on the "lightest" RGB channel, yes?

edit how? which curve were you changing? I am guessing you were adjusting only the curve for the channel that was sharpened, yes?

it is this last statement that makes me believe that you had a curves layer and not apply curves to the sharpened channel.

Thanks in advance! I see great possibilities and pardon me for not understanding the Photoshop language enough to follow your description

Bill Theis

___________________________________________________________________________

Several people were confused by my explanation. A post like my original one is a compromise between giving enough information and possibly insulting more experienced Forum members by putting in too much fine detail that is second nature for many of them. Hopefully the following expanded information will answer all the questions that have been presented, both on and off the Forum.

First of all, the basic procedure is just that. Yet I fear that many folks were trying to read too much into it and thereby got confused, being unable to believe that a procedure that can give such good results does not involve a lot of complex work.

In channel blending we need not be concerned with color. Granted that the final result depends on color, the channels themselves are nothing but grayscale images. You have to keep that in mind when blending them. A LAB channel can be blended with a CMYK channel just as easily as it can be blended with a channel mask in RGB. Mask or channel, they're just grayscale images.

Why use the lightest grayscale image (the lighter, Green channel in my grass image) instead of the darker Red or Blue channels? Because the goal is to darken only the areas that will give more contrast to individual grassblades, not to the entire image. What you're seeking is to darken the darker, background grassblades in the Green channel to make the lighter foreground grassblades stand out more. Keep in mind that darkening an RGB channel is the same as substituting darkness for the color it represents. The Green channel is made active in the duplicate image so that it will be the only channel sharpened. The use of Unsharp Mask causes the dark areas in the Green channel to darken still more. It also darkens the lighter parts of the channel, but not as dramatically. The result is that the composite image now shows darker accents among the green grassblades, with just a little darkening of the greens overall. This gives more contrast between the individual blades of grass. Had we sharpened the much darker Red or Blue channels, the enhancement of the greens would have been mediocre compared to the result of sharpening the Green only.

Sure, we can use the Unsharp Mask on the background layer, but the result would be relatively fixed. We could edit it some by changing the values for Amount and Radius, but not enough for the best result. By sharpening only the Green channel on the duplicate layer, we give ourselves some room for editing by adjusting the duplicate layer's opacity. It's one of Dan's favorite techniques for fine-tuning. But we can do still more if we use a Luminosity Curve to further adjust the enhanced contrast in the sharpened Green channel. To avoid editing contrast in the background layer as well, we "group" a Luminosity Curves Adjustment Layer with the duplicate layer, confining the editing to the duplicate layer's still active Green channel. Now, if the highlights are blown out we can restore them by adjusting the highlight end of the curve. If the shadows among the grassblades are not dark enough, we can darken them by adjusting the midtone region of the Luminosity curve. If the really dark shadows are too dark, we can lighten them by adjusting the shadow region of the curve. You can't do that with the Unsharp Mask alone.

If that's still confusing, work your way through the steps a few times, then consider what you're doing to the image. Like just about everything in Photoshop, it becomes more and more simple the more you work with it.

Duplicate the background layer with Ctrl-J (Command-J). Change its blending mode to Luminosity.

Create a Curves Adjustment Layer above the duplicate layer. Press Ctrl-G (Command-G) to link the adjustment layer to the duplicate layer. Change the adjustment layer's blending mode to Luminosity.

Click on the duplicate layer to make it active. Open the Channels palette and click on the Green (or Red) channel, whichever is lightest overall. To keep only the chosen channel editable and yet be able to see the results as you edit, click the eye icon for the composite channel.

Choose Filter/Sharpen/Unsharp Mask and use the same values for the Amount and Radius (I used settings of 50 for each). You will be sharpening only the chosen channel, but you'll be seeing the result in your image. Threshold, in case you wonder, was left at 0.

The result is likely to be something less than what you wanted. Don't fret. Click on the Curves Adjustment Layer's curves icon to open the Curves dialog box. For my own purposes, I make the gradient bars light at the top left and light at the bottom right. Move the bottom left point on the Composite curve a little to the right to increase contrast. No magic about this. You have to start somewhere. If you don't like the result, move it back and make whatever blends or moves in the curve that get you close to the result you're hoping to get.

When you think you've gone as far as you can, try adjusting the opacity of both the Curves Adjustment Layer and of the duplicate (sharpened) layer.

If the Red (or Green) channel doesn't work on a particular image, go back and try it again with the other channel.

Basically, what you're doing is sharpening a single channel in the duplicate layer, adjusting the contrast of that channel with a Luminosity curve, and blending the resulting changes in the duplicate layer with the original image, using Luminosity mode so you get only contrast changes, not color shifts.

I have no supporting theories for all this, just results that get the job done for me. Hopefully it will work as well for others. But my real goal is to stimulate some creative thinking and see how much better this can be done. Sharpening is like using Curves. There are an infinite number of adjustments and settings that are possible. The ideal would be to find a basic method that can be modified easily to give superior results with a great variety of images. As an example, using Curves in luminosity mode is one of those simple things that can be modified easily to improve a great number of totally unrelated images. It will be great if we can find a way to use a basic Unsharp Mask technique that can have universal application with only a few adjustments.

Some Forum members have expressed their concern over asking questions that might make them look inexperienced. To which I can only reply that stupid answers are common, but there is no such thing as a stupid question. For myself, I have little patience with those who adopt a condescending attitude when answering a legitimate question. Or, look at it this way. Most of us are not here to teach, but to learn. We do that best when we question what we don't know and share what we do.

If the procedure is still not clear to some of you, feel free to contact me either in the Forum or off-line. As for posting my test image of the field of grass (I've tried it on several others of course), most of my experimental work is done on commercial images from Webshots. Since they are copyrighted they can't be posted. If it's really necessary, though, I can take one of my own less professional images and post it.

Howard Smith

___________________________________________________________________________

"Howard Smith" wrote:

Hi,

Should this method of sharpening be applied after scanning or only after the file has been re-sized for final printing?

___________________________________________________________________________

Sharpening can be done any time to see what the effect will be, but the preliminary result should be deleted before making any further edits. As Dan has emphasized repeatedly, it is also useful to make a duplicate of the image and sharpen that. The result can be blended into the Background layer by changing opacity. You might also consider sharpening the duplicate layer, editing the Background layer alone, then blending the duplicate, sharpened layer into the final result in Luminosity mode. If it doesn't work out, you can always delete the duplicate layer. To avoid problems with enhancement of sharpening artifacts, sharpening is usually the last editing step in image preparation.

Howard Smith

___________________________________________________________________________

There are also arguments to be made for sharpening as a first step in a digital workflow (after spotting). This is explained in great detail in Bruce Fraser's book, as well as in the manual for PhotoKit Sharpener. This may have been what the OP was referring to. The 36- page Sharpener manual is a great tutorial on sharpening and it can be downloaded without cost.

http://www.pixelgenius.com/sharpener/manual.html

--Rich Wagner

___________________________________________________________________________

Howard, I do recommend that you read "Real World Image Sharpening with Adobe Photosdhop CS2" by Bruce Fraser, to gain a much better understanding of the role of sharpening as a process through the image editing workflow, and in particular why leaving sharpening to the end is not particularly robust advice, although I grant you it is very common. One of the signal contributions of this book is to carefully explain the preference for a process one implements throughout the workflow. By the way, the only mention in the book about PK Sharpener is a disclaimer on page xv to underscore that the book is about philosophy and technique, not about the product which embodies them. I do use that product routinely, however, and mention this to make the point that it provides a technically substantiated, non-destructive sharpening workflow solution whose implementation is far more efficient than what you describe below.

Mark Segal

___________________________________________________________________________

"Howard Smith" wrote:

no one, not even me, could be confused after reading your excellent step by step instructions in your last post! Thanks.

I have been playing with the technique and find it quite useful so far. I did try a new wrinkle: make the two layers part of a layer set. I can then change the blending of the set from "Pass Thru" to "Luminosity" which allows me to change the curves and sharpened layer to "Darken", one of Dan's tricks. Besides the blending % for the two layers, you have the (redundant but linked) blending for the layer set.

I saw no need to duplicate the layer set and change to lighten so that I could control the sharpening's lightening seperately from the darkening. The darkening part works well enough.

I'm sure you can think of a much more elegant means of getting darkening only with a luminosity blend...

So I'm still playing with it although I'm not so sure I know WHY it works like it does

Bill Theis