Untitled Document

Folks,

Since the publication of PP5E, some of the work I've been doing has convinced me that its recommendations on hiraloam sharpening need to be revised. Special attention should be paid to the first two items below, which are the most important.

Hiraloam (high Radius, low Amount) refers to the practice of reversing the conventional Unsharp Mask filter settings of low Radius and high Amount. In this technique, Amount is rarely set higher than 50% and Radius is often 30.0 pixels or even higher.

I recommended--and still do, more or less--that you start with a 500% Amount and increase the Radius to see what the general effect is. The idea is to add shape to objects and not to grossly lighten or darken them as a unit. So, if working on a face, you choose a Radius big enough to add depth to the eyes, but not big enough to lighten the entire face. Then, I said, reduce the Amount until a natural look is achieved.

**************

1) This last sentence needs to be modified. It turns out that most of what we find objectionable in this type of sharpening is the lightening halos, not the darkening ones. The effect is even stronger than in conventional USM, which is not what I expected. So, instead,

A) Make a duplicate layer and proceed to choose the appropriate Radius as above. Then, instead of reducing Amount to what looks good, reduce it to something clearly, but not ridiculously, too high. 60% usually works for me.

B) Apply Image, target top layer, source bottom layer, mode Darken, opacity 50%. I have not found any images yet where a lower opacity seems to work. (If you are sharpening in LAB, you must target the L channel on the top layer and blend the L below into it in Darken mode; Darken mode doesn't work when all LAB channels are active).

C) If necessary, reduce the opacity of the top layer to taste.

***************

2) When sharpening in LAB, only a crazy person would sharpen the A and B channels, right? That's what I thought, because it's true in conventional high Amount low Radius USM. But it isn't true in hiraloam. In every image where I've run a comparison, adding sharpening of the A and B has helped. It shifts colors around the edges enough to create noticeably improved transitions, but with the Amount set so low, not enough for a viewer to detect what's going on.

While there are technical advantages to doing this type of USM in LAB as opposed to RGB, I haven't found images (yet) where it makes so much of a difference that I'd recommend going to LAB just to sharpen. OTOH, the images in which this technique seems to work best are outdoor shots with greens and blues. Those are the kinds of images we're likely to be working on in LAB anyway.

*****************

3) Unlike conventional USM, a zero Threshold is usually appropriate in hiraloam sharpening of all three LAB channels, as the Amount setting is usually too low to provoke visible noise. The principal exception appears to be in skies.

******************

Hiraloam sharpening remains an adjunct to, not a replacement for, conventional USM. WIth these tweaks, however, I'm tinding I'm starting to use it on most images.

Dan Margulis

___________________________________________________________________________

Dan,

Thanks for the added notes on Hiraloam. I generally do something which may or may not be as effective (or may be a useful adjunct) which is to use the Blend-If sliders to reduce the effect on the lightest tones. I generally reduce the darkest tones somewhat too, but much less than the light tones. I find this gives me greater snap between the midtones where it's most needed without blowing out the highlights or blocking up the shadows. Sending small areas to pure white or black for sparkle seems a better use for conventional USM.

It seems to be particularly effective (needed) with Digital Captures, which is perhaps one reason that lately, like you, I've also been using at least a small amount on most images.

I'll be trying your technique to compare!

Ric Cohn

___________________________________________________________________________

While I find the discussion of Hiraloam of some technical interest, instead of going to any of this trouble I use Photokit Sharpener Pro, developed out of basic principles for addressing all these issues about which Bruce Fraser has written extensively. This is one of the finest sharpening tools available - no halos, and easily applied user controls for dealing with any other minor artifacts that may appear on the odd image.

Mark Segal

___________________________________________________________________________

Sounds like a side-by-side challenge would be appropriate. I'll be glad to supply some iamges and set up a web page for this. My best recent images are all ballet shots with high ISO so there is a noise issue. But if I dig, I'm sure I can find some others that will be interesting. But it doesn't have to be all about me. I'll take the first 3 contributions. You have to be willing to release your raw to be considered.

___________________________________________________________________________

John Ruttenberg wrote:

Unless I am mistaken, regular sharpening and the Photokit "plug" is not the same sharpening that Dan is talking of with HiRaLoAm (the closest Fraser method is midtone contrast sharpening, which uses much higher radius and smaller amount than what Dan is talking of).

I peronally use blend if sliders to limit the white halo (which is what Photokit would be doing, it is only executing Photoshop commands and does not actually add new USM code or anything), if not using separate lighten/darken layers or fades etc.

Stephen Marsh.

___________________________________________________________________________

Stephen,

Dan has alot of material on sharpening including the use of Hiraloam in PP5E where he discusses quite a few settings for different objectives ranging from special-purpose sharpening to other effects altogether, not all of which are necessarily targeted by PK Sharpener. My comment was a contribution aimed at the choice between pouring sweat over general sharpening issues in Photoshop or using PK. Where you say "Photokit is only executing Photoshop Commands", yes, but I would recommend deleting the word "only". While no one tool does everything, it is a very sophisticated set of highly versatile algorythms - of course all employing a range of Photoshop components - which took those guys a long period of time to develop and perfect. From what I've been reading tons of effort went into the concepts and numbers at the heart of that tool. I just haven't been tempted by much else since I started using it. But I do appreciate Hiraloam and in fact use it primarily for local contrast enhancement, combined with Blend If perhaps similarly to what you describe. By the way, Katrin Eismann has a nice piece on this combination in PhotoshopUser magazine December 2006 page 039. I recorded an Action for the whole sequence - it tames the tendancy of Hiroloam to clip highlights and shadows.

Mark Segal

___________________________________________________________________________

MARK SEGAL wrote:

Stephen,

Thanks for the reply Mark. I did not read the discussion that way, I was only familiar with v1.1 of PKS so if later versions offer such settings as HiRaLoAm then I knew I would be corrected. As a general side issue on the general topic of sharpening (not HiRaLoAm which was the specific topic at hand) then I would agree that PKS takes a lot of the monkey work out while still giving a flexible final result for further fine tuning if desired.

I think it is important to understand that this plug is perhaps similar to scripts/actions more so than what most people think of when they install a plug-in. Usually commercial plug-ins add extra capabilites or different/better ways of doing what Photoshop currently does, they do not leverage Photoshop code. I respect the team members behind the product and the gathered experience and knowledge that this product represents (which is what one is really getting for their money, not new features) - such things as an Automate plug-in or script are currently beyond my talents. This type of plug that calls Photoshop routines is a great concept and Adobe have many useful Automate plugs.

Thanks for the info, yes, an age old issue and Bruce Fraser did the same for midtone contrast 'sharpening' moves, where one does not wish to hit the shadows/highlights (those split blend if sliders truly are a great thing).

Sincerely,

Stephen Marsh.

___________________________________________________________________________

On 2/6/07 3:48 PM, "Stephen Marsh" wrote:

Regular sharpening? What©ˆs that in PhotoKit Sharpener? There are all kinds of radius and amounts in the product, all kind so masked generated with numerous parameters based on the capture or output devices specified by the user. Bruce©ˆs sharpening is not about a Œone size fits all©ˆ set of numbers for all capture and output devices because that©ˆs simply impossible!

I have no idea if it©ˆs the same (most likely not) but the big deal is the actual numbers necessary for a myriad of output devices having been defined by a heck of a lot of output testing. Is Dan©ˆs method output specific? Or does he subscribe that output sharpening for a halftone image is the same as a fine art ink jet or contone printer? IF so we need to lump that into the Raw challenge because, that©ˆs silly.

PKS as you wrote is 100% Photoshop. If we give you all the numbers plus the order of processes you could duplicate the effects 100%. But getting the numbers was the big work once the techniques (which are all over the web and in all kinds of books) is described.

Blend If is used in PKS for many routines. In fact, the layers are left intact so users can fine tune them if they feel the need (the vast majority don©ˆt).

Bruce pretty much provided a number of sharpening recipes in his fine book. What he didn©ˆt provide and what PKS does provide is specific numbers for sharpening parameters based on literally hundreds of hours of testing a heck of a lot of output devices. Not just a press and not just to Œlook good©ˆ on screen. In fact, most of the routines look pretty darn awful on screen (and the zoom ratio plays a huge role). Sharpening by eye is simply an exercise in futility. That is until we have displays that can output at least 300 dpi!

Andrew Rodney

___________________________________________________________________________

Hello Everyone,

Sharpening is still one of those frustrating tasks for me. And although I have looked at PKS and similair products, as a beginner I am purposely staying away from them until I think I have a thorough understanding of what steps are being performed by them on my behalf.

I have also read (once) Bruce Frasier's book on sharpening and look forward to studing it further. The results I got from using the recipes in his book weren't satisfactory compared to simple "one pass" sharpening techniques I had been using prior. I hope upon further studing his book I can improve on my intial results, as I did find the idea of a "multi-pass" sharpening workflow very sensible.

On the subject of Multi-pass sharpening, it seems like it should be easy enough to work out Input Sharpening numbers in USM for someone like me because I use one camera model for the majority of my work. And I use one scanner to scan film. Also, Output Sharpening numbers should be easy for me to figure out because I send all my photos to one printer, an Epson 2400. Both these tasks can be scripted. Thus the need to purchase a plug-in that takes into account a multitude of input and output devices seems unnecessary to me.

Besides, it's the "creative" sharpening that needs to be applied to each individual scene that's the hardest part to figure out! And so far, at least for me, individual inspection of each photo on screen and printed is the only way I've found to get satifactory sharpening results.

Again, ironically, it may be only after I become proficient at sharpening images manually that I can really trust to hand off some images to an semi-automatic plug-in.

Murray DeJager

___________________________________________________________________________

Maybe Andrew can speak up, but I believe the hiraloam equivalent in PhotoKit Sharpener is Haze Cutter. I think it's equivalent to Contrast Mask in the original PhotoKit product. I don't use either of these. There's also the Haze Reduction routines in the TLR Sharpening scripts/actions available for free download from here:

http://www.thelightsrightstudio.com/photoshop-tools.htm

All these appear to be variants of the same theme, with Blend Ifs to limit the effect in the highlights/shadows.

Stephen Best

Macquarie Editions

___________________________________________________________________________

--- Andrew Rodney wrote:

So how come the numbers for Contone 300 and Inkjet 360 Glossy are identical? Just teasing, but really you should let other users speak for your product and not get so defensive. It's a great product for those that are happy to just push the buttons, but it's not the last and only word in sharpening.

Stephen Best

Macquarie Editions

___________________________________________________________________________

I have NO commercial or other relationship with PixelGenius therefore I can talk about this product with complete objectivity. I have experimented with USM and three other sharpening programs before using PK Sharpener, and I agree fully with Michael Reichmann's evaluation of the product in his review on Luminous-Landscape - and he has no commercial relationship with ANY company. That evaluation says that it is the best tool on the market for sharpening photographs period.

Your statement that it's a great product for those that are happy to just push the buttons is frankly insulting to the developers of the product and to its experienced users - amongst which I count myself as one, thank-you very much.

If you really knew anything in depth about PK Sharpener you would never have made that statement because you would have known that it is distinctly NOT for people who are just happy to push buttons. You need to apply yourself to using this tool with care, forethought and experimentation. It's beauty is the amount of control over the results it allows its users, and with that control comes a responsibility on the part of the users to know how to implement those controls intelligently. That takes judgment and experience. PK works fast - after you decide how to work it!

While Andrew says that you can't predict results of sharpening from a monitor for the reasons he states, he is only partially right about that. Where he is right is that unlike colour management you do not get an obvious and reliable match between display and print for the appearance of sharpness. What he didn't say is that there is something we can do about this. Firstly, as he alluded to but did not explain, when inspecting the results of a sharpening move, we set the image magnification ratio to 25% or 50%. At these values there is no anti-aliasing of the monitor image and in that magnification range depending on the image size the effect most closely resembles (but doesn't mimic) what will appear on paper. Secondly, one gains experience comparing prints versus the display images for various sharpness settings till one reaches a point when the outcome becomes more predictable based on the display appearance. I have been using this product successfully (to judge from peer

review of my output) since it first appeared on the market and that includes the transition from CRT to LCD display, which required some "re-education".

Is it the last and only word on sharpening? Of course not - on that we agree.

Mark Segal

___________________________________________________________________________

on 2/7/07 10:27 AM, Andrew Rodney wrote:

Andrew,

I'm used to looking at sharpening effect at 100% magnification. Until your post I wouldn't have thought it to be an exercise in futility. What am I missing ?

Lee Clawson

2/\V/\7 Studio

P.S.-- And a 300dpi display !!!

___________________________________________________________________________

On 2/8/07 8:06 AM, "Lee Clawson" wrote:

The display is a very low rez device (72-96 dpi). The zoom ratio is important but it should be based on the output device. The old 'look at 100%' was recommended for two reasons. One, if you are not using an even divisible (say a zoom of 33%), Photoshop sub samples the preview to give you this ratio and its not at all accurate (for sharpness and some don't realize in some cases, even color!). The other reason was folks were thinking about a halftone repro. So say you are sending a 300ppi image to a halftone dot using 150 linescreen. You're going to be using four image pixels to make one halftone dot and therefore, you would be far better off soft proofing the sharpening at 50% which equates to using one pixel for the display.

The recommendations over the years has been, make the sharpening look slightly crunchy but what does that mean? And slightly crunchy will appear differently at different zoom ratio's. Bottom line is, you can't sharpen visually (well unless you've output so much content to a single device you train yourself what it should look like).

What's quite interesting is to use a sharpening routine that's based on actual output testing, output the image and then compare it to the display. The image on screen looks awful (depending on the zoom ratio it looks even worse) but on output, it looks great. Or view the same image on an LCD and CRT; which is right?

Maybe in addition to a raw versus JPEG and high ISO digital versus film, we need a sharpening iron chef challenge too. Now if only the parties to be would agree to proving their points of view in front of an audience.

Andrew Rodney

___________________________________________________________________________

I've voiced this issues before, but they're probably worthwhile iterating again. I have however reworked some of the output routines for my own use (they're just repackaged actions anyway) as these work well in push-button mode.

1. For film use at least (which is most of my work) the input categories are fairly broad, and too broad for my use. Even Bruce admitted this and implied that it was going to be refined in PKS 2.0. For example, I'd use different parameters for Ektachrome (which is relatively grainy) and Fujichrome (which is less so).

2. For some of the negative input routines, the PKS routines took forever with large files and the results that weren't appreciably better than I can achieve with other and faster means.

3. For grainy/noisy originals, I'd prefer to use a more specialized routine like Noiseware Professional and get this to sharpen as well in the same pass.

4. There's no visual means (preview) in PKS to determine the settings prior to input sharpening.

5. If you find you regularly use say a 40% opacity, there's no way to store this as a preset.

6. The output routines don't take account of the working space gamma for the Blend If parameters. To be fixed in 2.0 I believe. My actions do.

7. PKS doesn't work on grayscale or Lab files.

8. I now do combined capture/creative sharpening as a Smart Filter on a luminosity layer with the CS3 beta (see an earlier thread of mine). PKS can't handle anything other than a simple layer structure with a mandatory Background layer.

9. For web output, it depends more on the downsampling method. I generally find that Bicubic Sharper doesn't need further sharpening.

10. A few other limitations which I can't think of off the top of my head.

So I'd say I was fairly familiar with PKS, but judged it unsuitable for my work. If however you fit into the PKS mould, then it's not a bad tool. It's just a product which replicates what you can do yourself with Photoshop tools/actions. Most here I would have thought are more interested in the how and why of sharpening rather than the expediency of using a single tool.

Stephen Best

Macquarie Editions

___________________________________________________________________________

All -

I've read the discussion on sharpening with great interest. Here's a few ideas to consider:

-Sharpening may be applied on a layer - so it is non-destructive to your image

-You can use a layer mask to control where the sharpening is applied - this lets you "paint" sharpening on and off

- Try this: add a layer, go Filter>Other>High Pass - when the dialog box pops up, set it to (2) two pixels, click OK. Your screen will go completely gray -

don't worry - then change the layer blending mode to Hard Light or Soft Light - voila! It's a different look than USM.....

David Saffir

___________________________________________________________________________

Hello,

Andrew Rodney wrote:

I have one final question on this subject because I can't seem to get this comment by Andrew out of my mind!

Is there a reason why the resolution (ppi) of monitors has not changed since the beginning of computer time? Or is someone working on the problem right now? Assuming it is a problem.

Thinking about Andrews comment has got me thinking that it shouldn't be technologically impossible to produce such a monitor. Hell, even one with 150 ppi would seem like a huge improvement!

And it seems, at least to me, that it would probably be an easier monitor to manufacture than one with a larger color gamut then present monitors.

Murray DeJager

___________________________________________________________________________

"Murray DeJager" wrote:

They have. But there's always going to be a disconnect between what you see on screen and what you get on output ... that is until the screen can render the actual dot pattern/gain of output. Which means we're talking about resolutions like 1200/1440/1800dpi. The best we can realistically hope for in the future is some form of adjustable emulation. In the meantime this argues for evaluating overall image quality and sharpness on screen to some personal standard and having a procedure which can translate this optimally (and blindly) for whatever the output process is going to be. Which is PKS's approach for output and it works well enough.

I've standardized on 100ppi for my monitors but there would seem to be a trend for newer monitors to go for increased area over resolution. There's probably technical reasons why it's easier to provide increased luminance and colour depth with larger pixels. For B&W only there are technologies with really high resolutions.

Stephen Best

Macquarie Editions

___________________________________________________________________________

There are physical limitations on this. In the CRT days, the resolution was limited by (a) how finely and sharply the electron beam could be focused, and (b) the phosphor dot pitch. Dot pitch is the spacing between dot centers, it effectively this determines the size of each individual dot. One phosphor dot is not a pixel; theoretically you could write an entire image onto a single phosphor dot. But the physical limitation is on the dot spacing: how close are two adjacent red dots, for example - as you can't use the space between dots. The smaller the dots, the closer they are together, for a given screen size. This is why monochrome CRT monitors, with their continuous phosphor coating, have much better resolution capability than do color ones.

In the LCD world, each liquid crystal cell is a subpixel (a pixel is composed of at least three sub-pixels, RGB), and there is a physical limitation as to how small a cell can be. Might be this size of a cell could be reduced down to some point, but we also need to etch a lot of wiring and thin film transistors onto the glass substrate in order to control the liquid crystal. Might never reach 300 (x3) pixels per inch.

Similar limitations apply to other advanced display technologies, but they are not much used for digital photography. For example, a plasma display has to be pretty large before it can have HDTV resolution (in the 50" diagonal range), because each plasma cell (a fluorescent light)takes up a finite amount of space.

Randy Hoffner

___________________________________________________________________________

Actually, ppi has changed over time. The newest Apple laptops have a ppi of 150. There aren't too many monitors that have a ppi of exactly 72 or 96 ppi.

This gets weird, though, because people often confuse dpi and ppi. They aren't interchangeable. It's the source of lots of confusion and frustration. The other issue is that most GUIs are resolution dependent. When you look at the icons and text on a high-ppi monitor, everything is tiny. That's why a lot of people buy these monitors then run them at a lower resolution, which screws up what they're seeing any more because everything has to be interpolated (LCD panels have a native resolution they're designed for).

One of the big additions to Leopard will be a resolution-independent interface, so when you change your resolution, it won't change the size of everything on screen. I'm not sure if Vista is resolution independent or can be retrofitted to be.

Matthew Rigdon

___________________________________________________________________________

Dan,

Thanks, that helps quite a bit. I had already been doing the HIRALOAM on A&B in LAB; on some images the results are really quite amazing.

Cheers,

_andrew webb

___________________________________________________________________________

I concur with andrew webb... Thanks, Dan, for the suggestion as I would never have thought to HiRaLoAm the A,B channels

Bill Theis