The 1/100th Rule

It's funny how much you pay attention to sharpening once you actually pay attention to sharpening. I remember back when I had no idea what it actually did or how to use it, let alone a term like "capture sharpening." Heck, with my Kodak point and shoot 12 years ago, the only "adjustment" I ever made to an image was zooming it in and out on the screen!

So here I am today, a 33 year old chap with ADHD, OCD and enough Photoshop sliders and gizmos to keep my constantly revving mind cranking for eons, not to mention enough literature out there to keep me studying until I'm two feet in the grave and elbows on banana peels.

So, my latest little fixation is this whole concept of matching your sharpening radius to a measurement of the finest detectable detail the human eye can pick up on. I'm not sure if it was Bruce Fraser who first introduced the concept or not, but he certainly made it quite well known.

The theory makes complete sense to me: you have pixels that get crammed together at different "tightnesses" based on how much cramming your output device is doing. So, if you want a 1/100in. radius then that would be 3px for 300ppi, 1px for 100ppi, 2.7px for 270ppi and so on. And now, my curiosities:

1) 1/100th of an inch, which is the common number recommended, seems awfully large. For a print to a high-end Epson at 360ppi output, that would mean a radius close to 4 pixels! (3.6 according to the formula). I'm in no way a printing guru but isn't that an incredibly large radius?

2) Consider this: let's say you are viewing an image on a monitor that for the sake of simplicity, has a 100ppi display output (which is a reasonable average these days). If you sharpen for the web/that monitor, you may use a radius like myself and many others use, which is .3px. I find this to have quite a nice punchy effect (definitely noticeable). Now considering everything so far, that would mean that your and my eyes are making out a tremendous amount of detail at 1/333 of an inch; more than three times smaller than the aforementioned 1/100th! (100 pixels per inch/.3 pixels)

Now I know there's a difference between monitors and printers and different output devices, and that every image is different from the next, and that ink spreads on paper, but I'm talking quantifiable numbers and mathematics here; a pixel is a pixel. An inch is an inch. 360 multiplied by 1/100 is equal to 3.6. These are not really subjective measurements so I would reckon that there should be some truth to some areas of this. What do we think?

All digital editing is essentially one operation; changing the tonal level of one, two or all three RGB values in a pixel. We do many different variations on the theme in order to achieve different effects, but the truth is that raising or lowering those three numbers is always what we are doing. Sharpening is increasing local contrast, in places (edges) where there already exists a differential in luminosity between adjacent pixels, sharpening increases it - it makes the darker pixel a little more dark and the lighter pixel a little more light. So how do you darken 0.3 of a pixel? Obviously you can't. It's not like the physical world where once they thought the atom was the smallest particle of matter and for the last 150 years or so they have been finding smaller and smaller sub-atomic particles. In the digital imaging world the pixel is as small as it gets. Tonal alterations can only be made to the entire pixel, not just a third of it. So what does it mean to set 0.3 radius? Well, sharpening affects not just the two pixels immediately adjacent the edge but also their neighbors, with the effect diminishing as the distance from the edge increases. The pattern of this effect, how quickly it fades out, is determined by the radius setting. Think of four pixels in a line radiating from the edge, A, B, C and D. Radius = 1.0 might mean X change in A, X/4 change in B, X/8 change in C and no change in D. Radius = 1.5 could be A has X change, B has X/2, C has X/4 and D has X/8. Radius = 0.3 might be A has X, B has X/10, C and D have no change. The point is, the detail enhancement is always on the pixel level and resolution can't be any higher than that afforded by the number of pixels.

I generally just use smart sharpen with radius 1 pixel and I move the strength slider. Since I switched to Nikon I don't sharpen much, they're super sharp already and sharpening makes them look fake.

Elie is awesome with his information.
Thanks Elie

windpig wrote in post #16243676
Elie is awesome with his information.
Thanks Elie

Yes. Yes he is - always.

tim wrote in post #16243450
I generally just use smart sharpen with radius 1 pixel and I move the strength slider. Since I switched to Nikon I don't sharpen much, they're super sharp already and sharpening makes them look fake.

I'm a bit confused - you generally use a 1 pixel smart sharpen but you also claim that you don't sharpen much. By "much" do you mean "often" or that you just use a very subtle amount? I would assume that regardless of the camera/lens quality, a bit of input and output sharpening would still be desired.

tzalman wrote in post #16243430
All digital editing is essentially one operation; changing the tonal level of one, two or all three RGB values in a pixel. We do many different variations on the theme in order to achieve different effects, but the truth is that raising or lowering those three numbers is always what we are doing. Sharpening is increasing local contrast, in places (edges) where there already exists a differential in luminosity between adjacent pixels, sharpening increases it - it makes the darker pixel a little more dark and the lighter pixel a little more light. So how do you darken 0.3 of a pixel? Obviously you can't. It's not like the physical world where once they thought the atom was the smallest particle of matter and for the last 150 years or so they have been finding smaller and smaller sub-atomic particles. In the digital imaging world the pixel is as small as it gets. Tonal alterations can only be made to the entire pixel, not just a third of it. So what does it mean to set 0.3 radius? Well, sharpening affects not just the two pixels immediately adjacent the edge but also their neighbors, with the effect diminishing as the distance from the edge increases. The pattern of this effect, how quickly it fades out, is determined by the radius setting. Think of four pixels in a line radiating from the edge, A, B, C and D. Radius = 1.0 might mean X change in A, X/4 change in B, X/8 change in C and no change in D. Radius = 1.5 could be A has X change, B has X/2, C has X/4 and D has X/8. Radius = 0.3 might be A has X, B has X/10, C and D have no change. The point is, the detail enhancement is always on the pixel level and resolution can't be any higher than that afforded by the number of pixels.

Elie - this answer you gave is technical, thorough, simple, straightforward, logical, thought-provoking, interesting and clearly explained all at the same time. I thank you and at the same time resent you, for now you have planted a new seed of distraction in my twisted brain!

RandMan wrote in post #16244477
I'm a bit confused - you generally use a 1 pixel smart sharpen but you also claim that you don't sharpen much. By "much" do you mean "often" or that you just use a very subtle amount? I would assume that regardless of the camera/lens quality, a bit of input and output sharpening would still be desired.

I would disagree. It depends on what you want from your images. If you want them to look as true-to-life as possible (I.E. what they looked like to your naked eye), you may well not want to add any sharpening. Bird photographers tend to do a lot of (meaning strong) sharpening to emphasize the feather detail. Personally I find that distracting and not realistic-looking, but many people like it.

archer1960 wrote in post #16244534
I would disagree. It depends on what you want from your images. If you want them to look as true-to-life as possible (I.E. what they looked like to your naked eye), you may well not want to add any sharpening. Bird photographers tend to do a lot of (meaning strong) sharpening to emphasize the feather detail. Personally I find that distracting and not realistic-looking, but many people like it.

I will see your disagree and raise you one disagree. An image out of camera will naturally be softer than what you are seeing in real life due to various digital filters and processing that occurs inside of the camera as you snap the shutter. Sharpening does not always have to add an "effect" per se, but rather can serve the sole purpose of bringing something that's fallen below a normal level back up to "equilibrium." So for your bird example, you could apply just a little bit of sharpening to bring it back to what you may say from the straight perspective of the human eye.

The amount of sharpening added outside of the camera will depend on if you are shooting RAW or in camera JPEG's. For RAW images you will have to add some sharpening, at least some Input and Output sharpening will need to be applied, as well as possibly some creative sharpening. If you shoot SOOC JPEG then the camera will apply sharpening (eqv to the Input/Output sharpening applied to the RAW file) to the image as it processes it in the camera. You may want to add additional creative sharpening to the file afterwards. Either way the sharpening is applied to the image at some point in the process.

Alan

RandMan wrote in post #16244477
I'm a bit confused - you generally use a 1 pixel smart sharpen but you also claim that you don't sharpen much. By "much" do you mean "often" or that you just use a very subtle amount? I would assume that regardless of the camera/lens quality, a bit of input and output sharpening would still be desired.

I don't sharpen often, sorry. Whatever ACR does by default is generally enough, I find applying sharpening to Nikon files makes them look fake. I played with sharpening a lot more when I shot Canon.

Not trying to put Canon down, it's just my experience. I understand the focus systems are better these days, 5D3 is meant to be miles ahead of the 7D.