WOW!! - Unsharp Mask

[Spinners]

Wow, i just started to mess around with unsharp mask, and i must say WOW! it makes a huge difference in picture quality.. I do however wonder why all my pictures need to be touched up. I mean.. they apear to be in focus when i take them, yet, dont look as sharp as after i unsharp mask.

Do film photographers experience this too, or is it just a digital thing?


Also i would like to know what settings you have all found to be the best all around unsharp mask..

I use 250 for the percentage.. if i remember correctly, then 2 for radius, and 0 for threshold. I think i got those names right. Makes my pics very nice. And you can really see the difference in the prints.

Also, do you guys batch your unsharp mask, or one at a time.. cant imagine thats efficient?

[LeeO]

Spinners,
I use radius 1.0 and threshold 0. Amunt varies but is usually between 75 abd 100%

I don't batch sharpen anything as each image has it's own requirments.

[rdenney]

spinners wrote:
Wow, i just started to mess around with unsharp mask, and i must say WOW! it makes a huge difference in picture quality.. I do however wonder why all my pictures need to be touched up. I mean.. they apear to be in focus when i take them, yet, dont look as sharp as after i unsharp mask.

Do film photographers experience this too, or is it just a digital thing?


Also i would like to know what settings you have all found to be the best all around unsharp mask..

I use 250 for the percentage.. if i remember correctly, then 2 for radius, and 0 for threshold. I think i got those names right. Makes my pics very nice. And you can really see the difference in the prints.

Also, do you guys batch your unsharp mask, or one at a time.. cant imagine thats efficient?



Your settings might put a little white line around bright areas. You might try about .6 or .8 radius, 125%, and a threshold of about 10. The 10 setting keeps PS from finding edges in the middle of smooth gradients (as it will if threshold is 0). The small radius means that the lens is sharp, because the edge you are sharpening is less than a pixel wide. If the lens was fuzzy or the image out of focus, you'd need a bigger value. The amount controls how much the bright side of the line is brightened, and how much the dark side is darkened.

Film photographers deal with this, too, when they scan their work. I have a $2500 film scanner (a Minolta Multi), and I still have to use USM.

The edges aren't actually soft, but are rendered with a bit of smoothness caused by slight light spillage from one pixel cell to the next. It is really caused by aliasing, where the edge doesn't exactly line up with the boundary between pixel cells, and it's a digital fact of life. Most cameras apply USM or some other sharpening algorithm pretty aggressively in the camera, but Canon and Nikon both are pretty conservative in their DSLR's. This is good, because you can always add USM in PS, but if it's over done in the camera, you can't undo it.

But there is a sharpen setting in the camera that you can experiment with. I haven't, so I don't know how far it goes.

For making 4x6 proofs, I probably wouldn't bother with it. But for making display prints or for web presentation, I always work the image in photoshop anyway. But I only add USM to a copy when I know how it will be presented, and after I've scaled it to the output device. USM for low-res web images needs different settings than for high-res inkjet prints.

Rick "who wrote about this in another thread--do a search on 'pixel radius' to find it" Denney

[khenn]

Here's another way to do sharpening, and most of the time the results are as good if not better than Unsharp Mask. Plus you don't have to flatten your image and it has it's own layer so you can turn it off any time you want unlike Unsharp Mask.

1. On the Layer palette select your Background Layer and right click. Select Duplicate Layer.

2. With this new layer highlighted select Filter / Other / High Pass. Set the Radius to 10 and click OK. I've found that for macro's of flowers 10 works pretty well. If your working with portraits, try 5. Make sure you have the preview box checked and then you can see your results as you experiment.

3. Zoom into your image to Actual Pixels level so you can better see what you're going to do next.

4. Go back to the Layer Palette and select Hard Light from the left drop down. You may want to experiment between Hard Light and Soft Light. I prefer Soft Light.

5. Now go to the Opacity Slider and select a level of sharpening that seems best to you. Usually something between 20% and 70% will be best.

This was taken from the Luminous Landscape website at http://www.luminous-landscape.com/tutorials/high-pass-sharpening.shtml

[Spinners]

denny,

what do you mean by scaled to the output device.. are you talking like making an 8x10? im not sure i follow you there. I thought that you should scale down your images.. just send them to the printer full res, and let the printer make it 8x10? i think im missing something.. :(

[rdenney]

spinners wrote:
denny,

what do you mean by scaled to the output device.. are you talking like making an 8x10? im not sure i follow you there. I thought that you should scale down your images.. just send them to the printer full res, and let the printer make it 8x10? i think im missing something.. :(

I use different USM settings depending on the resolution of the image. For example, if I need an 8x12, I print at 240 and sharpen usually around .8, 10, and maybe 120%. But if I need a 500x800 pixel image for the web, I'll probably be using settings more like .6, 10, and 70% to keep it from looking oversharpened.

But you have to resize the image first, because USM acts on pixels. If you use USM, and then downsample it for the web, your sharpening ends up on the bicubic interpolation floor.

Here's my workflow: I scan or capture at the highest resolution. I correct the color, contrast, and curves for the Adobe color space on my calibrated monitor, and then apply just a small amount of USM to make the dust stand out so I can clone out the dust (not needed for the 10D--but mega-needed for my Minolta Dust-Magnet Scanner). Then I save it as a full-resolution Adobe PSD file. It has been corrected.

Now, let's say I want to make an 8x10 print. I open the file and immediately resave it under the same name with "8x10" appended to it. Then I crop it to 8x10 proportions, and then determine the number of pixels I need so that it will print at 240 pixels/inch on my printer (its optimum). I then resize it to that number of pixels. Then I turn on the output preview for my printer, and then target the color, contrast, and curves for the printer. But since I've recaled it, I need to apply USM again at the new resolution. I do that and then I print the file.

Then I might want to put it on my web page. I'll again crop it to the proportions I want, resize it to the smaller numer of pixels, target the color, contrast, and curves for the Windows/RGB preview, once again apply USM, and save it as a PSD file and then as a JPG file.

Each time you resample the image, you have to apply sharpening again, because the edges have been redefined by the pixel interpolation during resizing. That's why I don't apply it until I know how I'm going to display it, except for just a bit to make the dust visible.

Does that make any sense at all?

Rick "noting the important difference between correction (to make it right for Adobe) and targeting (to make it right for the output device)" Denney

[Spinners]

Whats the difference in the output if:

i send one to the printer at full res, wont the printer size it correctly?

or

i send one to the printer that i cropped to 8x10?

still abit confused.

[Jud Dagnall]

Another USM technique is as follows, although it needs to be done before adjustment layers are added. I have found that this method is better for preserving colors while still enhancing detail.

1. Convert the image to Lab Color mode

2. Select the Lightness channel ONLY

3. Sharpen

4. Convert back to your original mode

[Leighow]

Jud Dagnall wrote:
Another USM technique is as follows, although it needs to be done before adjustment layers are added. I have found that this method is better for preserving colors while still enhancing detail.

1. Convert the image to Lab Color mode

2. Select the Lightness channel ONLY

3. Sharpen

4. Convert back to your original mode



Luminous Landscape seconds Jud's method which I started using a few months ago when I began shooting in RAW. Luminous Landscape notes (as I recall) that you do not want to sharpen the color!

Collectively, you have all taught me a great deal about UMS. I wil be more systamatic in my approach and avoid "messing up" my master files by saving AFTER UMS !


HOWIE

[Spinners]

as neato as this USM stuff is, i am having a heck of a time figuring out what really looks better, and how sharp is to sharp.. Now granted i can tell when i have way oversharpened.. But i suspect no matter how i adjust these 3 settings i can accomplish the same image using all 3 settings in different combinations..??

I cant believe you all work every picture by hand! the last wedding i took pictures at, i had 200. Now i want to process them and develope my workflow.

I find USM to be awesome, yet doesnt seem efficient to adjust all pics individually, and i know they all would need USM. I printed some 4x6's on matte to see if i could tell the difference, and boy i did. sigh...

[Kevin Connery]

I'm surprised this hasn't been stated, yet, but it'll help explain WHY sharpening of some sort is needed for virtually all digital images.

The sampling that it does, either via a scanner or the camera's sensor, is on a grid. Since edges on the subject have no obligation to fall exactly ON that grid, the image has to make a determination. And, since few things are 100% in focus, those edges also tend to be somewhat soft. Thus, a pure black to pure white edge will probably be captured as intermediate shades of gray--looking 'soft'.

Sharpening adds the APPEARANCE of sharpness by increasing contrast at those edges.

The downside is when the increase is clearly visible (halos) instead of just being subliminal.

For most people's eyesight, a radius of under 1/100th of an OUTPUT inch is usually not going to be visible unless the amount is very high. So, a typical 300 DPI output medium can use up to about 3 pixels of radius for moderate amounts (up to about 100-150%) and 2-2.5 pixels for higher amounts. Web images, typically viewed at 1/72 to 1/100th of an inch on a monitor, can't survive that large a radius: 0.5 to .75 or so is a better choice.

Fading to luminosity is similar to converting to LAB and sharpening just the L channel, and doesn't involve a mode switch. Run the USM as desired, and Edit > Fade Unsharp Mask and specify luminosity from the pulldown blend menu. That avoids possible color fringing effects as different colored edges get sharpened. (USM operates on each color independently of the others; luminosity tells it to only affect the overall brightness, and to 'leave the colors alone'.

That said, there are many exceptions to the rule. The classic example is a close-up shot of a woman's face for a cosmetic ad and an orange. For the same resolution output, you would NOT want to use the same sharpening for both. Oranges look "better" if all the texture shows, but most people wouldn't want that same level of texture emphasis in a closeup of a person's face.

My personal batch workflow does a middle of the road sharpening. Resize for the web; USM at 100%, radius 0.7, threshold 5 for digital capture and 10 for scans, faded to darken only, followed by a USM at 100%, radius 0.4, same threshold, faded to luminosity. That gives a larger amount and radius for the darkening 'halo' than the brightening ones, and, for my work, that's appropriate. For individual images that are being reworked, I use amounts anywhere from 50% to 500%, a radius ranging from .3 to 5.0, and a threshold between 0 and 32, depending.

YOUR settings will depend on your output medium (LightJet and Frontier work well with lower amounts than inkjets do), resolution, and your content.

[Spinners]

i have found.. at least so far that 4x6's seem to come out well at aroun150% and a radius of 2. 0 for threshold. images are sharp and crisp. Color does not seem to be affected, or at least i havent run into it. i havent printed a close up shot yet at those settings, but im going to do that next.

Now if i am going to print an 8x10 then i would need to definetly increase the threshold correct?
Also what good is using a low radius, and a low threshold do, if you can even see the chanes on your screen. at least i cant really tell the difference. i really dont see a difference till i radius 2 and threshold 0.

[rdenney]

spinners wrote:
i have found.. at least so far that 4x6's seem to come out well at aroun150% and a radius of 2. 0 for threshold. images are sharp and crisp. Color does not seem to be affected, or at least i havent run into it. i havent printed a close up shot yet at those settings, but im going to do that next.

Now if i am going to print an 8x10 then i would need to definetly increase the threshold correct?
Also what good is using a low radius, and a low threshold do, if you can even see the chanes on your screen. at least i cant really tell the difference. i really dont see a difference till i radius 2 and threshold 0.

I don't find that larger prints need a high threshold. In fact, they generally need less.

Yours eyes have to be the guide. If you want the camera to do this for you, then set the sharpen function in the menu to maximum.

But there are some software packages that purport to sharpen automatically, though I haven't used them and therefore will defer to others. Judging from the work of those who use them, they are quite good, and I expect I'll be getting around to it eventually. Most of my digital work has been with scanners, and I spend so much time dealing with the scanner than another few seconds to apply USM hasn't been worth automating. That will probably all change now that I'm wearing the bits of this CD card.

Rick "for whom efficieciency hasn't been an issue--until now" Denney

[kellylipp]

I fooled with this a bunch and then got some advice from another contributor here and fooled a bunch more and then got lazy and downloaded Fred Miranda's Sharpening tools. Now talk about taking the chicken's way out: improved my photos a bunch and I didn't have to figure out the 50 PS steps that Fred uses!

I would love to see a professional seminar offered someplace in the world where us hacks could watch some of you studs tweak your photos! I'm a learn by watching kind of guy and this would be very helpful.

For now, I'll use the tools that others develop, work on my own stuff and wait until a pro in my area is offering a seminar.

Kelly

[Spinners]

where did you download or buy these tools?

[rdenney]

spinners wrote:
where did you download or buy these tools?

www.fredmiranda.com

Rick "looking at them myself" Denney

[Pekka]

kellylipp wrote:
I fooled with this a bunch and then got some advice from another contributor here and fooled a bunch more and then got lazy and downloaded Fred Miranda's Sharpening tools. Now talk about taking the chicken's way out: improved my photos a bunch and I didn't have to figure out the 50 PS steps that Fred uses!


Learning is always better in long run, here's some:
http://www.outbackphoto.com/workshop/photoshop_corner/essay_05/essay.html
this is method which was used first by Ultrasharpen ( http://www.ultrasharpen.com ) and was described in detail already 2001 in http://www.creativepro.com/story/feature/11242.html and http://www.creativepro.com/story/feature/12333.html

Other sharpening pages (there are tons of them in net, do a search, I found these with "Copernic Agent Basic"):
http://www.peachpit.com/photoshop/pdf/sharpening.pdf
http://www.guyd2.com/dpe/ss.html
http://www.astropix.com/HTML/J_DIGIT/USM.HTM
http://www.lonestardigital.com/photoshop_quicktips.htm
http://www.creativepro.com/story/feature/9343.html