High pass/Overlay vs Unsharp mask

[Dmitri]

any opinions?

 

[Garbz]

Now there's something that I haven't through of. What is the difference? I went a head and did a quick test to see how each reacts in photoshop by way over applying them to gradients. Check it out:

On the left is High pass with a radius of 10, and overlay with a 100% opacity.
On the right is unsharp mask with a radius of 10, tolerance of 0 and 100% strength.

Now a quick analysis shows that on the outside edge the effect looks incredibly similar, actually if I removed the centre you wouldn't tell the two effects apart when they spread out onto a solid colour. This result would actually favour the unsharp mask since it is far more configurable. You can have strengths greater than 100% and the threshold to limit the effect slightly. This however can also lead to abuse and unfortunately over sharpening and ugly halos. The upper limit on the effect of highpass/overlay may drastically reduce the chance of this oversharpening unless it's applied several times repeatedly.

One thing I am surprised about is that the inside of the gradients appear much smoother on the high-pass filter. This would tend itself to a more linear effect against varying contrasts, however that would lead to a reduced "perceived" sharpness.


Now just not looking at images yet I would say from this that high pass / overlay would be incredibly well suited for really large scale sharpening to increase local contrast, such as a radius of about 150pixels and an opacity of about 30% or so just to give the image a little more pop since the results would seem to be more linear and consistent. I'll have to try this at some point since I've previously used unsharp mask for this purpose.

That said for final sharpening and creating more edge definition the more aggressive application of the unsharp mask may favour more. I would imagine it would give a "sharper" effect at a lower radius, along with it's customisability it would best suit the actual overall sharpening of the image to increase the microfine detail.


This is interesting, I'll eager to hear what other people think. I'll have to go out and take a photo tomorrow to try.

 

[KmH]

Both, plus a couple of other techniques. It depends on the image.

 

[Big Mike]

They are the same process....just different ways of going about it.

I read an article a couple years ago that explained how/why they were the same.

 

[NateS]

Don't know why, but anytime I use High Pass/Overlay my image turns into total crap. Even when keeping it minimal, my results are not pleasing. However I can do a decent amount of USM without it looking jumbled and oversharpened. I'd guess it has to do with my High Pass technique or lack thereof.

 

[CraniumDesigns]

i tend to lean towards high pass, as it can be done non-destructively, unlike unsharp mask. i take my base layer, duplicate it twice, set the top layer to "hard light" then do a smart high pass filter on the middle layer. the bottom layer is only to preserve the original image.

 

[Overread16 4]

But you can do unsharp mask nondestructivly by the same method - duplicate the base layer and run the unsharpen mask on the duplicated layer. Heck you can even get fancy with layer masks so that the effect is not applied to the blurred areas (thus the blurred areas from the base layer show through)

 

[Garbz]

Actually with the advent of smart layers anything can be done non-destructively these days. You can even go back and change your sharpening afterwards.

 

[freixas]

Hate to reply to an old thread, but I put your clever little test image to good use.

And I can now answer the question of how HighPass sharpening differs from USM. They both start off the same way: by subtracting a Gaussian blurred version from the original. But the USM adds the difference directly back to the original, whereas the High Pass sharpen has to use a blend mode to do the work—and the formulas are not the same.

You can view all the gory detail in my paper at

http://freixas.deviantart.com/gallery/#/d2rib58

 

[brianT]

Hate to reply to an old thread, but I put your clever little test image to good use.

And I can now answer the question of how HighPass sharpening differs from USM. They both start off the same way: by subtracting a Gaussian blurred version from the original. But the USM adds the difference directly back to the original, whereas the High Pass sharpen has to use a blend mode to do the work—and the formulas are not the same.

You can view all the gory detail in my paper at

freixas's Gallery

Wow, excellent paper! Looks like a lot of research went into it. I read most of it, but don't really understand the math formulas as they're beyond my understanding.

I've used a lot of different sharpening methods. What I tend to use the most is:
1. Convert to LAB mode
2. Duplicate layer to be sharpened
3. Sharpen entire image (of the duplicated layer) using USM on LIGHTNESS channel to 100% 1.0 radius, sometimes 200%, and sometimes other larger radius
4. Add layer mask to duplicated layer
5. By painting with a brush, add or subtract areas of the layer mask so areas of an image are sharper than others. To me this is the most important step. For example, in a clean blue gradient sky and don't want any sharpening because it will just add noise. In small detail foliage and don't want much sharpening because it creates too much contrast and noise which draws the viewer's eye in where I don't want them to look. The areas that get most of the sharpening are the details of the subject. Of course, it depends on the subject. Naturally softer surfaces like people's faces or maybe flower pedals don't get too sharpened. But rough stuff like concrete looks better with more aggressive sharpening applied. So in the end it's an artistic choice.

On a flattened image I sometime repeat the above steps in an additional sharpening pass. I do the same steps but use a much larger sharpening radius (like 15). This is usually to very slightly sharpen clouds in the sky, and only those clouds -- nothing else in the image. This sharpening is very, very subtle. So it makes a very small noticeable difference but adds some 'punch' to the clouds, if desired.

I should also say that my method is used for images to be viewed on the monitor in original size, or smaller size to be posted on the web. I don't do any printing, so maybe printing would need a different sharpening method.

 

[katy625]

Wow i LOVE LOVE LOVE learning new things with Photoshop. I have had PS 7 since last year and I didn't even know anything about the Highpass filter....i use USM but I tested this out on a pic and combined USM then Highpass together and the results were great!! Thanks ya'll!