![](3D"http://www.bythom.com/Images/Slides.gif"=20)
=20
![](3D"http://www.bythom.com/Images/sharpening.jpg"=20)
Sharpening=20 101Probably the=20 one area for which lots has been written but little understood is = the=20 topic of sharpening. This article attempts to clear up some of the = confusion, and offer a few useful options. |
Contents: |
Sharpening is easy to = understand at=20 the simplest level=97given two digital photos, one sharpened and one = not, you=92ll=20 always point to sharpened one as having more detail and being =93better = focused.=94=20 Yet, if you were confronted with two sharpened photos, one well done, = and the=20 other less so, you might not be able to differentiate between the=20 two.
Unfortunately, most = digital=20 photographers have learned sharpening by fixed rules that specify exact = values=20 to use in Photoshop's filters, and therefore neither understand why they = use the=20 tools they do, or how they might be using them rightly or wrongly. This = article=20 attempts to clarify the problems created by digital imaging, and how to = apply=20 the correct amount and type of sharpening to fix them.
Why Is Sharpening Necessary?
The=20 nasty truth underlying all digital recording techniques is that they = turn analog=20 signals into discrete samples of the original. CD players, for = example,=20 sample sounds at 44 kilohertz (i.e., 44 thousand times a second) and = record each=20 sample using 16 bits of data. The frequency of digital sampling and the = amount=20 of data sampled determine how well the analog original can be = reproduced. The=20 coarser the sampling, the less the digital recording is like the=20 original.
State of the art consumer = digital=20 cameras have a sampling frequency of 4500 by 3000 pixels (e.g., Kodak = Pro 14n),=20 with the amount of data recorded being 12 bits for each of the red, = green, and=20 blue colors. These numbers are actually relatively crude compared to the = analog=20 reality, where detail and color variations are nearly infinite. The real = world=20 sports an infinite number of shades of blue in the sky and an endless = amount of=20 detail, but your digital camera only captures between 1000 and 4500 = pixels of=20 horizontal detail in perhaps thousands of shades of possible blues. = While that=92s=20 pretty darn good, it does cause two resolution-oriented = problems:
- Detail smaller than the = pixel=20 size is usually lost.
- Where=20 transitions between details occur within the area of a single pixel, = the=20 transition usually results in a digital value that is neither of the = original=20 values.
This = second problem=20 is what makes details in your photographs look fuzzy. The classic = example is=20 that of a diagonal transition line that transects a pixel. The pixel can = either=20 be white, black, or some in between value. If the camera were to render = the=20 pixel as entirely white or black, then you=92d see an artifact known as = the=20 stairstep, so named because a diagonal line gets rendered as a series of = pixel=20 blocks that resemble a set of two-dimensional stairs. The alternative is = to=20 record the pixel as an "in-between" gray (which still produces a bit of = a=20 stairstep effect, but isn't quite as obvious). Neither case is correct, = and both=20 tend to reduce apparent sharpness.
All = digital cameras=20 use in-camera interpolation to detect edge transitions, and use some = form of=20 digital sampling to create =93in-between=94 values for those diagonal = lines. The=20 result? Instead of a precise transition from one pixel value to another, = diagonal details (and sometimes small horizontal and vertical details) = are=20 rendered as a more gradual transition from one color to another. Our = brains have=20 been programmed to see blurry or soft edges as being out-of-focus, thus=20 unmodified digital photographs always tend to look just a tad soft. = That=92s even=20 true of higher resolution cameras and scanners=97film images I=92ve had = scanned on=20 4000 dpi drum scanners still look a little soft in the detail = areas.
Worse = still, most=20 digital cameras employ what is known as an anti-aliasing = filter--essentially a=20 diffusion filter over the sensor. (The exception is the Kodak Pro 14n). = Why?=20 Because the Bayer pattern sampling used in digital cameras has a = tendency to=20 produce colored artifacts and moire patterns on small detail. By = blurring the=20 light slightly so that multiple photosites get some of the information = from a=20 particular detail, this lessens the chance that these hard-to-remove = artifacts=20 appear. Unfortunately, it also has a further tendency to make edges less = distinct.
The=20
method by which most digital photographs are =93corrected=94 is to apply =
a=20
sharpening =93filter=94 using an image-editing program, such as =
Photoshop. What=20
these filters do is to detect transitions (edges) and make them more =
pronounced.=20
The easiest way to see this is to use two gray =
blocks:
- Open=20 your image editing program and create two adjacent gray blocks (it = doesn=92t=20 matter what size). The lefthand block should have the RGB value of 86, = 86, 86=20 (in Photoshop, doubleclick on the foreground color in the tool = palette, then=20 enter those values directly into the RGB section of the Color Picker = that=20 appears). The righthand block should have the RGB value of 43, 43, 43. = Be sure=20 that these blocks touch one another, and that you=92ve turned off any=20 anti-aliasing options in the tools you used to create them.
- Zoom=20 way in on your image (500-1000%). You want to see the transition line = between=20 the two blocks clearly, so place it in the middle of your = window.=20
- Use=20 the Sharpen filter (in Photoshop: Filter->Sharpen->Sharpen).=20
Note how the software = lightened a=20 single pixel column on the light gray side and darkened a single pixel = column on=20 the dark gray side (just each side of the cursor in the image, above).=20 Essentially, the transition is being exaggerated. At this close view, = your eye=20 can easily see the change, but now select Actual Pixels (View->Actual = Pixels). Since you know what to look for, you can probably still make = out the=20 light gray column (though I bet you can=92t see the dark gray one, = especially if=20 your monitor has been correctly color calibrated). Use the Undo command = (Ctrl-Z=20 in Windows, Apple-Z for Macs) to toggle back and forth between the = unsharpened=20 and sharpened versions. The sharpened version should indeed look = sharper. (For=20 extra credit, rotate the unsharpened version 45 degrees and repeat the=20 sharpening step; the unsharpened version should seem significantly = softer than=20 the sharpened version).
Okay, now you know what = to look for=20 in sharpening: sharpened digital images have a slight =93halo=94 at = edges, one that=20 can usually only be seen by enlarging the pixels or knowing exactly what = to look=20 for. If you performed the extra credit assignment in the last paragraph, = you=20 also are aware that diagonal edges can obtain interesting "artifacts" = when=20 sharpened (the vertical artifacts are rather mild-mannered in = comparison, aren't=20 they?).
The Unsharp Mask
For=20 our simple example, we used a straight Sharpen filter, which worked = fine.=20 Unfortunately, the real world is a little more complicated than two = shades of=20 gray with a straight-line edge between them. We need to be able to = control the=20 amount of sharpening that is done and how it is applied, and for that we = use the=20 Unsharp Mask. Contrary to its name, it does actually sharpen your image. = An=20 Unsharp Mask gets its name from the fact that back in film days, a small = piece=20 of glass was placed between the original negative and a positive mask to = exaggerate contrast along dark/light junctions; the glass put the = positive mask=20 slightly out of focus from the negative (in other words, an unsharp = positive=20 mask was used to "sharpen" an underlying image). The software variant = attempts=20 to reproduce that effect. In areas where there isn't a lot of detail, = there=20 wouldn't be much difference (the "Threshold" value controls how much it=20 considers to be a significant difference), so the function won't try to = change=20 values (changes are controlled by the "Amount" in the Unsharp Mask = settings). In=20 areas where there was detail, the original data and the blurred data = would be=20 significantly different, and thus the function senses there is detail or = an edge=20 there.
Probably the=20 biggest problem in using the Unsharp Mask is determining how to set the = three=20 controls (Amount, Radius, and Threshold for Photoshop users; other = programs=20 often use different terminology, apparently so that they don't receive = nasty=20 letters from Adobe lawyers). Most written advice I=92ve seen always = gives one set=20 of starting points, then leaves the rest up to you to determine = visually. And=20 every writer=92s starting point seems to be different. But if you read = the last=20 paragraph carefully, you'll already have some ideas about what these = values=20 might be.
First, let=92s get rid of = one notion,=20 that there=92s a magic starting place that applies to all images. = Instead, let me=20 suggest that there are at least two starting places, and many images = need to use=20 selections to apply different sharpening techniques to different areas. = More on=20 that in a bit.
Next, let=92s make sure = you know what=20 each of the controls refers to:
Follow along at home: = Start up=20 Photoshop. Recreate your own simple gray block document to match the = example=20 (above). The lefthand block should have an RGB value of 86, 86, 86 and = the=20 righthand block a value of 43, 43, 43. If you use the Paint Bucket tool = to=20 create the blocks, make sure that anti-aliasing is off. You want solid = blocks of=20 a single color value.
Sharpening makes the edge = of the=20 lighter block a lighter value of gray, and the edge of the darker block = a darker=20 value of gray. Select Unsharp Mask from the=20 Filters menu and set starting values of = 100=20 for Amount, 2 for = Radius, and=20 0 for Threshold.
-
Amount: = determines the aggressiveness of the "sharpening" action. With your = simple=20 two-gray image, try amounts of 100, = 200, and=20 400 (make sure the Preview box is = checked in=20 the Unsharp Mask dialog so that you see the changes = as you=20 make them; you should also be viewing at Actual = Pixels size).=20 What you should see is that as the amount is increased, the colors of = the new=20 edges get more exaggerated. In other words, the light line that gets = added on=20 one side of the boundary gets lighter with each increase, the dark = line on the=20 other gets darker (though that=92s often more difficult to = see).
-
Radius: = determines how wide an area at the transition is affected. Try = increasing the=20 Radius to 4 and 8, and you=92ll see = that area=20 that is modified at the transition widens. Note, too, that the further = away=20 from the actual transition point you get, the less the Amount is=20 applied.
-
Threshold:=20 determines how much difference there must be between two adjacent = pixels=20 before any change is made. In our simple example, you=92ll have to = enter very=20 high numbers before you see how this works (try 25,=20 50, and 100). Note that threshold = and radius=20 interact a bit. With a Threshold of 100 and a Radius = of=20 1 or less, almost nothing changes, but if you = increase the=20 Radius, you=92ll start to see the effect again.
Okay, I wrote earlier = that there=20 isn=92t a magic starting place that applies to all images. While = that=92s true for=20 Amount, both the Radius and Threshold probably should be started at = specific=20 points:
-
Amount: = <determined picture by picture, and with some consideration to = enlargement=20 size, if applicable>
-
Radius: = start=20 with .5 and try to avoid going much higher, if = possible. I=20 believe it=92s better to apply Unsharp Mask twice = with=20 .5 and .3 than using an initial = radius of=20 .8. Why? Because any value larger than = .5=20 starts to affect more than one pixel beyond the transition point, = which starts=20 to produce more visible halos, especially if you need to use = aggressive amount=20 values. If you=92re printing with an inkjet printer, the dot gain you = get from=20 the ink spreading on the paper often masks these halos, so go ahead = and try=20 higher values if you=92d like, but only if you analyze the results = from the=20 final output (not the screen).
-
Threshold: start=20 with 0 and leave it there if your image is = relatively=20 noise-free. Using any other value for Threshold applies the filter = to only=20 parts of the image, and I believe there are better ways of handling = partial=20 sharpening than using Threshold (see
Sharpening Rules
Before continuing, = let=92s set some=20 specific rules down for dealing with standard = sharpening:
-
Perform all = other changes=20 (color correction, saturation changes, distortion corrections, etc.) = before=20 sharpening.
-
Save a copy of the = corrected, but=20 unsharpened version. (And you=92ve already saved a copy of the = original,=20 right?)
-
Use the Unsharp = Mask for basic sharpening, as it provides more control than = the other=20 related filters.
-
If submitting to a = professional=20 designer, send the unsharpened version, as you don=92t know how dot = gain may=20 change the amount of sharpening necessary.
-
If printing on an = inkjet printer,=20 error on the side of slight oversharpening.
- Try = to use a=20 Threshold of 0 and = Radius=20 of .5, if possible.
Lab Color Sharpening =96 Using the Luminosity =
Data
Some books and seminars = recommend=20 converting your document to the Lab Color space=20 (Image->Mode->Lab=20 Color), and then applying an Unsharp Mask to the Lightness = channel=20 (Click on the Channels tab in the = Layers=20 window and make sure that Lightness is the only channel = selected before applying the filter). After you=92ve sharpened your = image this=20 way, you convert it back to RGB Color=20 (Image->Mode->RGB=20 Color).
The=20 reason for using a technique like this is that sharpening colored edges = means=20 that the Unsharp Mask uses the color data to determine how to modify the = edge.=20 This sometimes results in minor shifts in color in the sharpened areas. = But the=20 changes are small. For example, the small blue area between the pole and = the=20 building front in the image below shifts ever so slightly bluer when = using the=20 Unsharp Mask on the original RGB (a value of 90, 132, 187 shifts to 107, = 153,=20 210 using Lab Color sharpening, but to 109, 155, 213 using RGB Color=20 sharpening). If you aren=92t using color profiles for all of your = equipment and=20 attempting to maintain color matching between acquisition and print, = sharpening=20 using only the luminosity data is probably overkill, though, as any=20 miscalibration between your monitor and printer likely produces a bigger = shift.
If=20 you are minding every last little bit of color change, be aware = that=20 changing color space introduces rounding errors in color values (and the = error=20 is more severe in 8-bit images than in 16-bit images). That same pixel = value I=20 noted before changes from 90, 132, 187 to 90, 133, 188 after converting = to Lab=20 Color and back to RGB Color, without making any other changes to the = document.=20
While I=92ve never seen a = color=20 changed in ways that I found destructive or problematic by rounding or = shifting,=20 I have noted very subtle changes in areas of gradated color patterns. At = 1000%=20 view I sometimes see a few random pixels shift in sky tones when making = these=20 changes. (Best way to see for yourself: make the adjustment from RGB = Color to=20 Lab Color and back to RGB Color. Then zoom way in to an area that has a = subtle=20 gradation, then use the History palette to click back and forth between = the=20 original RGB and the reconverted RGB. If you=92re satisfied you=92re not = seeing any=20 color shifts worth worrying about, then go ahead and use the Lab Color=20 sharpening method.) If you do decide to use the Lab Color method for = sharpening,=20 convert to Lab Color and back only once. Rounding is cumulative, = so=20 bouncing back and forth between RGB and Lab Color is a no-no!
A=20 better way to sharpen just the luminosity channel without changing to = Lab Color=20 is to perform sharpening as you usually would (e.g., using = Unsharp=20 Mask), then choose Fade Unsharp Mask from the=20 Edit menu. In the dialog that appears, change the Mode = from=20 Normal to Luminosity and click on the=20 Okay button.
Edge Sharpening
Portrait and nature=20 photographers have one thing in common: we hate seeing sharpening = artifacts on=20 large blocks of color. Any underlying noise in the image tends to bubble = up to=20 visible, and skies or skin tones with too much detail in them don't look = as good=20 as gently gradated ones. And if you know what bokeh is (the character of = out-of-focus areas) and care about it, you'll be particularly bothered = by what=20 happens to noise in out-of-focus areas. Indeed, the first thing I look = at in an=20 image to see if sharpening is detectable is the non-detailed, = out-of-focus=20 areas. I've seen all kinds of techniques that attempt to deal with the = problem=20 of pulling up sharpening artifacts (deselecting the continuous tone = areas before=20 sharpening, using history brushes to paint out sharpening in certain = areas,=20 etc.). But the following technique is my favorite. Essentially, you = build a mask=20 of just the edges in your image, then apply sharpening to the areas = under the=20 mask. Here's one set of steps you can use in Photoshop:
- Open = your image,=20 as usual. Perform all your other corrections on it before sharpening, = as=20 usual.=20
- When = you're=20 ready to sharpen, click on the Channels palette and = create a=20 new channel. This new channel will eventually be your mask. Change = it's name=20 to "Sharpening Mask."
- Click on the RGB=20 channel for your image.=20
- Select the=20 entire image (Select->Select = All).=20
- Copy = the entire=20 image (Edit->Copy).=20
- Click on the=20 Sharpening Mask channel you created in Step #2. Paste the image in=20 (Edit->Paste). It'll appear in = black and=20 white, but that's what we want, so don't worry.=20
- Use = the Find=20 Edges filter=20 (Filter->Stylize->Find = Edges) on the Sharpening Mask channel. It'll turn into = something that=20 looks a bit like a line drawing. Our next steps will be to tweak this=20 mask.=20
- We = want the=20 black to be really black and the white to be really white in our mask, = so use=20 Levels to make the adjustment=20 = (Image->Adjust->Levels).=20 Move the black point in from the left, and the white point in from the = right=20 (the triangles under the histogram). How much you do this is one of = the=20 critical choices you'll be making, so take your time, and remember = that the=20 black areas are what are going to be sharpened (the white areas won't = be=20 sharpened).
- We = want to hide=20 the actual sharpening of the edges, so we need to make sure there's a = smooth=20 transition from white to black in our mask. To accomplish this, use a = small=20 Gaussian blur on the image (say 2-4 pixels)=20 = (Filter->Blur->Gaussian=20 Blur).=20
- Now = we need to=20 tell Photoshop that our Sharpening Mask is just that, a mask. To do = that, we=20 need to select the black bits only. Use=20 Select->Load Selection. In the = dialog=20 that comes up, make sure that Sharpening Mask appears as the Channel = and that=20 the Invert box is checked.=20
- Click on the=20 Layers tab and make the image visible again. You should see the = selection=20 created in Step #10 superimposed over the image.=20
- Get = the image=20 ready for sharpening. Choose=20 View->Show->Hide = Edges=20 (Cmd/Ctrl-H) to make the selection invisible, then choose=20 View->Actual Pixels and scroll = the image=20 to an area with which you want to evaluate the sharpening.=20
- Finally, we're=20 ready to sharpen. Select=20 = Filter->Sharpen->Unsharp = Mask and play with the values. Note that you can use much = higher=20 amounts than usual, as the halos tend to fall outside the selected = areas.=20 Radius should be a low value (.5 to = 1.0),=20 and Threshold should be 0. If you find that the = sharpened=20 area is too restrictive, go back to Step 8 and change your = mask.=20
Yes, = that's a lot=20 of steps, but you can easily make them into a Photoshop Action (make = sure to=20 click on the icon to make Steps 8 and 13 editable during playback). = This, by the=20 way, isn't the only way to perform an edge sharpening technique. I've = seen at=20 least a half dozen variants on the idea of making a mask for the = details, then=20 sharpen only at the mask.
Using Sharpening to Change Contrast
Here's = something I=20 hadn't noticed before. Then one day I made accidentally entered some = "stupid"=20 values into the Unsharp Mask filter and noticed a change to the image = that was=20 interesting. So I did some research. It seems that a number of digital=20 cogniscenti have been using the Unsharp Mask to make slight bumps in = image=20 contrast. It's really quite simple to do: just select a low value for = Amount=20 (10-25%), a very high value for Radius (200-300 pixels), and a 0 = Threshold. This=20 technique actually produces a more likeable contrast boost, in my = opinion, than=20 does the Contrast command. Just be sure to save a = version of=20 your image before using this technique (it's not undoable at a later = date). You=20 also perform this action prior to doing a "real" sharpening.
Selective=20 Sharpening
If you = want to=20 apply sharpening only to some areas (say the eyes of a portrait but not = the=20 facial skin, do this:
- Create a=20 duplicate layer of what you want to sharpen.=20
- Apply your=20 sharpening to the duplicate layer (you might also want to slightly=20 oversharpen).=20
- Create a=20 Layer Mask for the duplicate layer.
- Paint black into=20 the areas of the Layer Mask that you don't want=20 sharpened.=20
- Adjust the=20 Opacity of the layer downwards if you = oversharpened.=20
|
|