"smart" sharpening software

[Glenn Kroeger]

I've read numerous comments about "smart" sharpening algorithms in use. Bill Atk inson mentions them. Charles Cramer refers to a script by Deke McClelland called "Sharpen Only Edges". I have also been refered to Nik SharpenerPro software.

Anyone tried comparing these. What works well. I am interested in output to a Li ghtJet, but want something that won't kick up grain as much as Adobe's build in unsharp mask algorithms.

Thanks for any input.

[Pete Andrews]

Someone just e-mailed me the Deke sharpeninmg method, and I haven't had time to try it yet. It sounds very fiddly.<be> I've never found the unsharp mask in PS to exaggerate grain, but I seem to use different settings from most people. My settings are usually around a 3 to 6 pixel radius, threshold level setting of 7, and 40 to 50 percent sharpen.I see people advocating sharpen settings of up to 500%, ridiculous!

[fred_1006]

go to: http://www.ickybits.com and try out Robert's Ultrasharpen 3 tool. I have been using his Ultrasharpen 2 action for quite some time and find that it works really well. I just downloaded 3 yesterday and haven't had a chance to try it out yet.

[tim atherton]

Try here: ftp://ftp.pinkheadedbug.com/pub/ the Deadman tool is by far the best, but I don't know if the instructions are with it.

Just load it into Photoshop in the Actions Palette. To get bck on topic, it does amazing things with many of my 4x5 scans - way better than Photoshop USM.

Tim A

[William Leviit]

Sharpening is best done while making the scan. The LinoColor Scansoftware has an excellent sharpening filter.

[John Hennessy]

Glenn

A lightjet cheatsheet I got from somewhere (maybe the Calypso site?) says not to sharpen too much (not over .9 pixels and 0 threshold) and to do it in photoshop NOT the scanner. The scanner sees only one line a time so how can it calculate lines it hasn't seen yet or lines it has passed over and "forgotten" about?

If you're getting too much "grain" you need to mask off some parts of your image before applying the USM; for instance, I never apply USM to the sky or water with reflections, etc.

As for you actual question about other sharpening tools: I haven't tried any but plan to do so now that so many have been mentioned.

[Joshua L. Slocum]

Question- In large format photography, why is electronic sharpening necessary? We all know that it's nothing but the augmentation of local contrast - hence, not "real" sharpness - and it doesn't add detail. So why is it so popular? Is it necessary to make up for a loss of sharpness in the scanning process? If so, why scan? For all the hoopla I hear about the "incredible sharpness" of Fuji Frontier digital prints, for example, every one I've seen looks totally fake- like video tape. Seriously. . .uniform, square-like, totally unnatural texture across the whole image, passing off completely binary tone differences that don't exist in the visual world (i.e. lines never look as sharp in a scene as they do in a digital/video rendering of a line which is basically "on" or "off") as "sharpness." Can anyone enlighten me?

[John Hennessy]

Josh:

Why do photographers go to so much trouble making unsharp masks in a wet darkroom? To enhance the appearance of sharpness. It is absolutely possible, and in the case of digital USM, not unlikely to over do it.

But, assuming sharpness is a legitimate goal for a given image, a wet USM or its digital namesake is a wonderful tool if correctly used.

[Glenn Kroeger]

Josh: Yes, it is easily possible to overdo sharpening. On the other hand, I disagree that all sharpening is not "real" sharpness. First, lenses and film loose contrast at edges by virtue of the falloff of their MTF. Restoring that is actually restoring "real" edge sharpness lost by the behavior of optics and emulsions. Second, to accomodate the "unreal" limited contrast range of photographic papers, overall contrast is lowered in the scanning process. Restoring that at edges is essential. A good raw scan is much softer in appearance than the original image.

[Joshua L. Slocum]

Gentlemen:

Thank you very much! That's exactly the kind of answer I was looking for, and it makes total sense:)

[William Leviit]

High-end drum scans made for the purpose of going to off-set press (atleast at the Lithographers here in Germany) ARE done during the scan. High-end scan software such as LinoColor establish the necessary criterian (contrast, brightness, sharpness and color balance) using a detailed "pre-scan". The changes are then caluculated and applied during the fine scan.

[Bill Brady]

The Epson TWAIN driver does unsharp masking at scan time. This *is* the best spot in the chain to do it. Because the scanner driver has all of the physical pixels available to it, it is possible to actually increase sharpness as long as you aren't scanning at max optical resolution. (like increasing DMax with a higher bit count ADC) The problem is finding out whether any particular driver actually does this. I'm still not sure about the Epson driver.

[chris jordan]

hey guys-- there have been a couple of misstatements here about sharpening that i wanted to help correct so you don't end up with bad results.

--first, NEVER use the sharpening algorithms that come with scanners. they don't have the flexibility that Photoshop's USM has, and once you've done it, you're stuck with it. make your scans with no sharpening, and save them with no sharpening. sharpening is the VERY LAST step to do before printing.

--don't sharpen the image until you make a final print. do all of your photoshop tweaks and save the file. then resize the image to the print size, do your sharpening, and print it out. if you want to save that file, save it as a different file, so the original non-sharpened image will be available in the future so that if you ever want to make a different sized print, you can resize it again and sharpen it for that size.

next, someone's suggestion here of using a high radius (i.e., 6 pixels) and low amount (i.e., 50) is a terrible way to sharpen. there are lots of reasons why it degrades the quality of your image, but for now i'll just say that top-notch printers who really know what they're doing, never sharpen that way. among the fine-art photographic printers there is some consensus on how to sharpen, which is to follow these concepts:

--do your sharpening ONLY while looking at the image in 100% mode (actual pixels). if you look at it in "View print size" or anything less than 100%, you will end up with an oversharpened image.

--use a threshold of zero.

--use a very low radius-- never more than 1 pixel, and usually in the 0.3 to 0.4 range.

--because you're using a low radius, you will need a large "amount". try starting with the amount at 300, and go from there based on how it makes your image look. stay in 100% view mode!

--after applying the unsharp mask, go to FADE, and fade it at 100% (i.e., it doesn't fade at all), but put the fade in LUMINOSITY mode instead of normal mode. what that does is turns the sharpening effect into a black-and-white-only effect, so it doesn't increase the saturation of the color of your image. this is a super important step!! it accomplishes the same thing as converting to LAB and sharpening only the L channel (but without doing the conversion, which can degrade your image depending on what RGB colorspace you're in).

--sharpen selectively in different areas of the image by making a copy of the background layer and sharpening that, then painting the sharpened areas selectively. that way you can sharpen some areas more, other areas less, and for areas that are supposed to be blurry, you can leave then unsharpened. you can even make several copies of the background layer and sharpen each one differently, and paint different areas from each one. this technique makes for a more 3-dimensional look and eliminates the creation of grain in smooth areas where you don't want it.

--another way to sharpen is to use the "sharpen" tool, which actually is a fairly sophisticated tool. but it almost always oversharpens the image, so after applying it, go to "FADE" and fade it back. stay in 100% view mode when doing this, and also put it in "LUMINOSITY" mode instead of normal.

okay, those are my nuggets for the day. happy printing.

~chris jordan, Seattle

[chris jordan]

p.s.: there are other sharpening engines out there other than Photoshop's USM, but having done lots of testing with them the consensus among the fine-art studios around the country is that Photoshop's USM is still the best if properly applied using the principles set out in my previous mesage. ~cj

[Mani Sitaraman]

Josh:

Agree with you 100% about Fuji Frontier prints and their unnatural appearance. What keeps me going back for more is the superior masking and exposure/color control, though.

The sharpening can be really awful-to the point of obliterating differences in sharpness between very different classes of lens.

[Ellis Vener]

Great discussion! Thanks for all the intelligent ( meaning: relevant) advice and input and a great question too!

[Paul Schilliger]

I use the same methods and levels as exposed by Chris Jordan and find it working well. I tried however the UltraSharpen 2 script and find it really worthy of interest. Could not find any information on Sharpen Only Edges, but I think UltraSharpen must be pretty similar in the way it works, with the ex ception that SharpenOnlyEdges works on the red channel only, have I read. As Chris reminded me, I have made good experiences resizing images with bilinear mode (instead of bicubic). Fine detail s are preserved and the images especially if they are small (web images) are much more detailed.

[Carles Mitja]

Some points about USM,

a) Any digital image, both from scanners and digital cameras, will need some amount of USM in order to be sharply viewed at final step (screen, inkjet, photogrphic paper, etc.). There are a lot of particular characteristics of digital imaging process that induces lost of sharpness.

b) The amount of USM necessary for a given picture depends on, essentially, three questions: The first is the scanner or digital camera own performance. The second arises from the amount of present detail (image frequency) in a given picture. The third, finally, is related with the final step of visualization (screen, inkjet, photogrphic paper, etc.).

c) All the stated above concludes to the impossibility of the existence of any rule of thumb about wich is "the best" amount of USM or "the best" smart software to apply it.

d) There are at least two questions to have into account: It seems obvious that we need some test of our equipement (scanner and/or camera) to try the best amount of USM to apply to our pictures. In second place we need to analyze EACH picture in particular and the amount of texture and detail to decide the appropriate amount of USM to be applied. It is too important to realize if it is necessary or advisable to apply the USM or the same USM to the whole picture area.

e) Using the USM tool a gaussian blur is applied to a copy of the original image to produce a new image with no fine detail; the blurred image is then substracted (mathematicaly) from the original image to extract the fine detail. This fine detail is added to the original to highlight the detailed areas.

f) The RADIUS determines the amount of gaussian blur first applied (Probably large files need more radius than small files to be the same effect). The AMOUNT defines de quantity of detailed version added to the original. The THRESHOLD indicates how different in brightness (from 0 to 255) must be two adjacent pixels to enables the USM be applied.

g) The threshold permits to avoid some areas to be contrast enhanced by the USM, like the face skin in a portrait or the sky in a landscape. May be interesting to analyze previously the neighbor pixel differences in those areas with the INFO Palette in order to decide the thershold to be applied.

Disagreement or discussion will be welcome,

Carles Mitja