USM in PS 7.0 - DPChallenge Forums

04/25/2006 04:13:06 AM · #1

Ok, I am working on getting a working knowledge on USM workings. Links would be fine to answer questions. I'm not afraid of reading. I just get a little overwhelmed sometimes as this is a pretty big topic and there's a lot of information out there, much of which seems to go all over the place with the settings used.

I was very glad to get a foundation by getting the suggestion to START at Canon's suggestion of 300%, Radius 0.3, Threshold 0. I took Kirbic's suggestion and now use a threshold of 3 most of the time.

Now First things are first.

I understand the Amount, and I half-understand the concept of Radius, but Threshold seems to boggle me.

Can anyone define Threshold?

Second things second.

I watched some FANTASTIC video tutorials from Radiant vista and I now feel reasonably comfortable with the idea of using an adjustment layer to sharpen.

The first question is obviously: Is this legal in Basic? If I use the layer mask, is that legal? If it's not, I would assume that using adjustment layers would be off-limits because they automatically generate layer masks.

The next questions that raise themselves largely because the tutorial I watched did not cover USM, only smart sharpen which is only in CS2.

I would like to become comfortable using USM because I feel that it is a fairly basic tool. I can bring those skills into CS2 if I choose to upgrade later.

I notice that when I have a picture at full size, it often needs a fair bit of sharpening to make it look good. On the other hand, when I resize down to DPC compatible 640x640 or thereabouts, I notice that my sharpening settings are VERY strong, often bordering on oversharpening.

Does using multiple passes help prevent oversharpening?

Would it be safe to say that the primary difference is the ratio of the radius in the two different sizes?

If I have a picture that has 4 times more pixels, does the radius of the sharpening need to be quadrupled? I can't see that this could possibly be right, but I don't know the math.

Can anyone help with these possibly basic or at least foundational questions?

My deepest thanks!

I am especially interested in this because I have been asked to process around 100-150 of my own pictures of kids at this school I have been subbing at and submit large sized pictures. I have not taken pictures of every child in the school, so I can't rightly call this a serious job, but it's a major step in the right direction. The school has been quite impressed with the pictures I have taken. Indeed they have held out the opportunity to actually do 2 more shoots, possibly for money later on.

One of these may be as early as THURSDAY!

04/25/2006 06:09:03 AM · #2

bump

shots

Fujifilm FinePix S7000

04/25/2006 06:20:13 AM · #3

Just a little tip...

Try converting your RGB image to Lab color via the image / mode menu..

hide the a and b chanels...

apply USM....ammount 200 %, radius 1, and threshold 0

unhide a and b

convert back to RGB...

this gives much less edge haloing..

Message edited by author 2006-04-25 06:21:57.

04/25/2006 06:48:57 AM · #4

I'll give it a whirl. I took abot 120 more shots today.

On one hand, while it's nice that my crap shots to good shots ratio is changing quite rapidly, it's pretty scary how many shots are actually worthy of editing.

I pulled 12 decent shots out of the first 36 taken...

Sitting at 254, That's a lot to edit!

shots

Fujifilm FinePix S7000

04/25/2006 07:01:50 AM · #5

I made a photoshop "action" for the technique...easy !

04/25/2006 07:41:38 AM · #6

Yup. I just got my new pictures sorted and the first thing on my list is ACTIONS... Didn't get a huge response on my thread on how to do this, but I'll muddle my way through it.

I still need to figure out what settings I actually want before I do the action. Some pics need more sharpening...

Sony 16-50mm f/3.5-5.6 E-mount

04/25/2006 08:16:40 AM · #7

Dude, that is AMAZING! Beautiful results on that action. It takes a bit longer, but it works GREAT!

I made an action for it too.

Anything to say on multiple passes? Just to experiment, I ran it again and it wasn't exactly the sort of thing you would want to do twice in a row :)

jmsetzler

Sony Alpha a6000

04/25/2006 08:22:19 AM · #8

There are lots of different methods of sharpening that work equally well. The one thing to watch out for is the simple fact that when it looks sharp enough on the screen, it is generally too sharp for printing...

Canon 12x IS USM (S3 IS built-in Lens) -- 36-432mm (35mm EQ)

04/25/2006 08:49:12 AM · #9

Interesting. Any particular reason why this might be?

The Radiant Vista Tutorials said to check them out at 50% to see a more accurate view of how they will look in print. This seems accurate, 50% on each side would roughly represent the area of view at 4x the resolution.

I am assuming this is a rough estimation at the difference between 72 PPI ish on a computer screen (which would require more sharpness to actually look sharp) and the standard of 300DPI of printers (about 4x) which would require less sharpness to look sharp.

This could also be because of sharpening done by the processers at the print shop...

I thought it was possible to not have the sharpening done there if you just ask them.

I am aware that there are many methods, but I don't have any currently working for me. I don't feel that the Canon 300 0.3 0 works equally as well as the method provided by Shots. Not entirely understanding the way it works, I am ill equipped to play around with it in ways that are meaningful and helpful, and especially consistent.

I am quite beginner, so with a HUGE field of variation, I would really appreciate some starting points...

GeneralE

Canon PowerShot S3 IS

04/25/2006 10:07:58 AM · #10

The basic idea of threshold is to set a limit as to "how different" in color two pixels have to be before the filter is applied.

If you have smooth gradient areas you don't want to sharpen between every little change, so you set a higher number. If you have lots of little details, you might set a lower number. I usually use a TH of either 5 or 7.

Yes, two applications of USM at "lighter" settings will sometimes look better than one application with more aggressive settings.

Yes, larger images (more pixels) require a larger diameter setting, but not four times as big.

I have some examples of sharpening and over-sharpening here.

kirbic

Canon EOS R5

Canon RF 24-70mm f/2.8 L IS

04/25/2006 10:19:54 AM · #11

Let me try to address a couple of your questions from above...

Threshold sets the minimum contrast difference for which sharpening takes place. If an edge has a contrast difference of less than the threshold value, it is not sharpened. Noise in an image usually consists of a LOT of low-contrast edges. Assuming that noise levels are not extremely high, it's usually easy to avoid "sharpening noise" using threshold.

I have not seen the Radiant Vista tutorial on sharpening using an adjustment layer. Use of the adjustment layer in Basic is legal if:
1.) The adjustment layer is applied in "Normal" blending mode
2.) The (automatically generated) layer mask is not used (remains blank)

With regard to radius and the size of the image, there is really no set ralationship. All depends on how sharp the original is. For instance, a 5D image using a very sharp prime under optimal conditions yields an original that requires very little sharpening, even though the image is quite large. The optimal sharpening radius for 5D images seems to be 0.2, basically as low as you can go. An image from a 6Mpx DSLR using a "consumer zoom," OTOH, might need a radius of 0.6 or above. the bottom line is that the required radius depends on how closely the image approaches maximum possible sharpness.

Sharpening on the luminosity channel (by converting to LAB) can give excellent results. Converting to LAB and back can have undesirable effects on color, however, since the gamuts of the spaces are different. To avoid the conversion, sharpen and the Edit>fade and select "Luminosity." this has the same effect, without the conversion.

Multiple passes of USM at lower amount can produce superior results... try sharpening twice, once fading to "darken" and the second time fading to "lighten." This reduces halos somewhat.

Nikon AF-S DX NIKKOR 18-200mm f/3.5-5.6G ED VR II

04/25/2006 10:24:55 AM · #12

Thanks so much General!

I find so far that Shots' suggestion works famously for pictures that were in focus, but I did a pretty large number of very low light shots on some pretty active kids where the focus isn't always perfect, so I've got a handful of pics that need rescuing.

It seems that I am able to get some pretty good results when working at 640x480 without severe oversharpening, but when I am working with larger pictures, I can't get anything sharp in pictures that need rescuing.

EDIT: Working on this as we speak.

One important thing to note, Adjust your levels, curves and things before sharpening (or in my case testing sharpening) what a difference it makes.

The Radiant Vista tutorial lists 5 basic adjustment layers. The 6th is not legal in basic.

They are:
Levels for Darkness and Brightness
Curves for Midtone Contrast
Curves for Color Balance (or color casts)
Selective Color for Color Balance
Hue/Saturation

There's a 6th they recommend which is a Curves for Contrast based on luminosity.

I have just complete building around 20 actions while editing some of my pictures and checking these things and double checking that video Tutorial.

In 2 hours, my output quality has improved visibly.

Lots O Fun.

Thanks for the tips, keep em coming!

Still having trouble balancing USM. Too many variables.

What setting do you use for the amount of fading for the luminosity on the USM?

Interestingly, I have noticed that some of the pics I have are indeed VERY sharp. At 100% magnification, the tiniest hairs in their eyebrows are sharp lines.

At this point, if add ANY sharpening, it is too much. I've left the last 3 pics unsharpened.

Message edited by author 2006-04-25 11:31:47.

swankFoto

Fujifilm X-E1

04/25/2006 12:37:25 PM · #13

[post for profile bookmark]