Just when I’ve gotten used to sharpening by using a tool called unsharp mask, I now read from Peter Bauer (Photoshop User issue July/August 2004) that sharpening can also be accomplished by Gaussian blur. He says to:
1) duplicate the layer, 2) Change blend mode from normal to overlay, 3) apply heavy Gaussian blur to the top layer, 4) add the high pass filter and adjust slider until edges are sharp as possible.
Could someone please explain how sharpening can be done by adding blur? I understand that unsharp mask is a darkroom term/technique but this high pass filter just does not make sense.
Well, other than the fact that I don’t see the image getting sharper no matter how much high pass you apply I do notice some major color shifting which is not acceptable. Definitely not one of the better sharpening methods.
I saw some tonal changes but no actual sharpening, at least on one image. I also tried Hard or Soft light blending as well. Variations on the theme, but not much sharpening.
The high-pass filter makes sense, but the blurring does not.
The high-pass filter works hand-in-glove with the Overlay blend mode in that the mid-gray it produces in low color-gradient areas is the neutral color for that mode. That is, Overlay does nothing to the substrate when the top layer is mid-gray—it’s as if the layer did not exist in those areas. The deviation from mid-gray is governed by the steepness of the color spatial gradient and controlled by the high-pass filter slider.
Tones produced in the filtered top layer that are above mid-gray are lightened and tones below are darkened by amounts that increase with the deviation, and this emphasizes the contrast across edges.
So far, so good. We get the sharpening we seek without blurring the top layer. I can’t see what the advocated blurring brings to the party.
In the past, I have used this method (sans blur and certainly sans invert) with the soft light, overlay and hard light blending modes to achieve differing levels of sharpening, with hard light producing the strongest effect. Now that vivid light has been added to the mix, it holds the distinction of producing the most sharpening. FYI.
I think you might appreciate knowing that for a SELF-BLEND (top layer a duplicate of bottom layer and unaltered by blurring, inverting, etc), there is NO difference whatsoever between Overlay and Hard Light blends, and Soft Light is a muted version of them reproducible by adjusting opacity.
Gayle, in fact, sharpening does not make an image less blurry, it creates the illusion that an image is less blurry by exxagerating local contrast, with a strongly contrasted halo. Blur can be used to create the halo…
Usually, as Ho, I never use blur for the high pass method
"…Ho means using the High Pass, hard light method …"
Pierre, I see what you mean. It’s not a self-blend after applying the high-pass filter because the top layer is no longer identical to the bottom layer. My oversight. Thanks
No biggie… anyways, Ho didn’t state "high pass" in his post, it was implied from the OP…
BOB means that if the blurred copy is blended (using substract blending mode, from apply image) with the original one to create a contrasted halo, you are doing some sharpening. (that’s how it was done in the ancient times <g, d&r> of wet processing) here is an example of a manual USM: < http://www.fortunecity.com/victorian/canterbury/222/unsharp. htm>
I have suspected all along that the sharpening techniques used are the same as the manipulation of rise time in electronics. In an ideal system, one starts at some voltage level, say zero, and in zero time the value changes to another level. That’s ideal; in real life, you have a certain time lapse that occurs between level shifts and that time is the controlling factor as to how fast your system can respond. The shorter the time, the faster the system.
But wait! There’s more! Again, ideally, as you transition between levels, you would like to see a response that is essentially gaussian. With that being so, as the event takes place, viewing the edge on an oscilloscope, you would see the corner smoothly starting up, a straight line between the 10% and 90% values of the level change, and the corner being smoothly but crisply turned to the flat portion of the waveform. Now, in an effort to decrease the transition time, we can introduce effects that "Bump up" the corners so that now, instead of starting and finishing with a smooth transition a negative going bump (undershoot, sometimes called the "anticipation bump!) at the beginning, and a positive bump at the top, called overshoot. These bumps in the waveform can be of almost any size, but the larger they are, the more unstable the system, until you get ringing on the waveform, not very desirable!
All this is exactly like the sharpening effects in the Unsharp mask, as the math details. You can substitute electrical terms for the visual terms and get the same answers. Therefore they are rough equivalents of each other. I have not seen the equivalent to ringing in cases of oversharpening, but that may be what the artifacts of oversharpening consist of.
Anyway, just a bit of babble from an old time analog engineer!
I understand that unsharp mask is a darkroom term/technique<<
Just to help your understanding of what is happening:
In Darkroom terms, unsharp masking is achieved as follows:
A soft (out of focus) positive is combined with the negative. This reduces the overall contrast, but doesn’t reduce edges as much as ‘solid’ areas. The overall loss of contrast is compensated for by printing on a higher contrast paper. This brings back the overall contrast of the image to ‘normal’ but boosts the contrast at the edges.
The unsharp mask filter ACHIEVES exactly the same, but the method you are describing actually duplicates the steps of conventional film unsharp masking. (Although you appear to have omitted the step of inverting the mask layer).
Len, when I tried the technique last night, I wondered about the inversion, as Chris Cox claimed it was a subtractve process, but I saw no application of subtraction being applied. Thanks!
Lawrence, Because you "add" a negative to a positive you actually "subtract" the lightness values. I didn’t write "subtract blending mode", but overlay.
Also, when comparing local contrast enhancement with analogue electronics, rise time is blur radius, and the resultant contrast acts as a rise function with overshoots on both sides of the rise.
There is also an effect in film, called the adjacency effect, sometimes dubbed the Eberhard Effect. It is controlled by agitation procedure, and is dependent on the material as well.
I retried the above procedure again, and sure enough the invert is important. But as soon as I saw it, I recognized it as a variation on a High Pass sharpening procedure that Mathias V (sorry! can’t remember how to spell Mathias’ last name)put me on to. It works much better, and between his HP sharp and edge sharpening, I have ooodles of control.
There is another masking procedure, also called Unsharp Masking ( and a Sharp Masking also) the is used to control the shadow values in negs and highlights in transparencies. The Shadow/Highlight in PSCS is a perfect digital version, and with far greater range.
used to control the shadow values in negs and highlights in
transparencies. <<
That’s normally referred to as ‘contrast masking’ and generally uses a sharp low-contrast duplicate to increase contrast and a sharp low-contrast reversal to reduce contrast.
The Portland labs did call it unsharp masking back then. I don’t have my Kodak books outlining the process any more but I seem to remember they also referred to unsharp masking. Of course, it was in the context of contrast control.
I never knew about unsharp masking for sharpening until I started PS, so I never was a problem in my mind. But, I can see now that considerable trouble would have resulted if I asked for an unsharp mask from a NYC lab without specifying exactly what I meant!
Both sharp and unsharp masks for contrast were done, and the unsharp prevailed as it required no precise registration. I made many of them myself.
Lawrence I havent used CS, and dont see me upgrading before I land my next big job at least..
I do not know what magic S/H pocesses.
These layer based techniques in <PS7 has some drawbacks. Generally they only work in 8-bit mode, although they can be rewritten to 16 bit with actions.
But if you combine the methods described in my two techniques above, use advanced layer blending with, say Contrast Masking, you can control the layer to effect only the shadows of the underlying image, for instance. Or the midtones, or the highlights or..
It gives me good control.
But generally, I like to do my adjustments before 8 bit these days.
Just wanted to say I love these threads where you image pros hash out the pros and cons of different methods of achiving a common (usually desired<g>) result in PS!