Unsharp Mask vs Smart Sharpen

[alecdann]

Is there a reason, or use cases, to use Unsharp Mask rather than Smart Sharpen?  Smart Sharpen seems to have more tools.

I'm interested in multi-pass sharpening.  is there a reason to use one over the other in that scenario?

Alec Dann

[patjoja]

Is there a reason, or use cases, to use Unsharp Mask rather than Smart Sharpen?  Smart Sharpen seems to have more tools.

I'm interested in multi-pass sharpening.  is there a reason to use one over the other in that scenario?

Alec Dann

For digital photographs, Smart Sharpen is the way to go.  A google search will yield a treasure trove of articles comparing the two.

Regards,

Patrick

[digitaldog]

http://creativepro.com/out-of-gamut-thoughts-a-sharpening-workflow/

[BobDavid]

Sharpening is the last step in editing a file. ... One trick is to change the profile of the file to LAB (be sure to take a snapshot of the file before the conversion). Then in LAB apply a sharpen adjustment to the luminance channel. Observe how the result looks at 50%, 66%, and 100%. When finished, take a snapshot. Then covert the picture (or not) to the original color space.

Now there are two versions, or snapshots, of the file: sharpened and not sharpened.

1) Use the globally sharpened file exclusively;
2) Paste either "snapshot" above the other and modify the opacity of the top layer to taste;
3) Create a mask for the top layer, fill it black or leave it white, and use the brush tool to holdback or to reveal the bottom layer.

There are many variations of this technique. It's simple and intuitive once you get the hang of it. The trick to sharpening is to ensure the photo doesn't look like it has been sharpened.

One thing to consider about sharpening is to keep in mind the size of the final output.

[TonyW]

Rather than risk the quantisation errors introduced by converting to LAB you will probably find that staying in your preferred RGB editing space and applying sharpening to a new layer masked as appropriate and setting the layer blend style to luminosity.will do an equal job without the data loss risk.

I cannot see any reason not to use Smart Sharpen for all work

[BobDavid]

Rather than risk the quantisation errors introduced by converting to LAB you will probably find that staying in your preferred RGB editing space and applying sharpening to a new layer masked as appropriate and setting the layer blend style to luminosity.will do an equal job without the data loss risk.

I cannot see any reason not to use Smart Sharpen for all work

I've not run into quantization issues. Sharpening the luminance channel is clean. Try it, you'll like it.

[TonyW]

I've not run into quantization issues. Sharpening the luminance channel is clean. Try it, you'll like it.

I have tried it and from memory this was something that came from Dan Margulis many years ago. 

Actual testing of images revealed nothing to be gained by stepping in and out of LAB over staying in RGB and applying sharpening to luminosity layer add the risk of the loss of data being visible and I made the decision to not risk it.

I must say that my testing limited to comparing only a few images and I never came across any overwhelming evidence to support the use of LAB sharpening over luminosity.  If there is any that demonstrates this I would be very interested to see.  To be fair I think the risk probably greater working in 8 bit rather than 16

I guess at the end of the day to quote (misquote?) Deng “it does not matter as long as it catches mice” 

[BobDavid]

I shoot RAW and develop files as 16-bit, aRGB TIFFs (C1 for the A7rII and the Olympus RAW developer for the Pen F).

I've learned from Dan. His book "Photoshop LAB Color" is an excellent guide/tutorial on LAB. I've read several of his books. Dan is a controversial figure. Some dismiss his concepts as being outdated or off the mark.

Dan says R, G, B, C, M, Y, K, L, A, and B channels are available to work in concert. The key to working with 10 channels is to create a master file: 16-bit aRGB (my preferrence). From there, if adjustments are called for in other spaces, copy the master file, convert to whatever color space provides access to specific channels, make an adjustment, and then convert the file back to aRGB. For example, use the CMYK>aRGB file as a layer and the Master aRGB file as another layer. Then through masks and blending modes, adjust accordingly.

Of course, one needs a calibrated monitor capable of accurately displaying aRGB (think Eizo Coloredge).

I expect to hear descent from some of the frequent and loyal LL readers and contributors. The color science "guru" engineers may come out of the woodwork and go to the trouble of presenting XYZ and XY graphs to dispel Dan's concepts. To them I say:

Color science/management involves quantitative (by the numbers) and qualitative (subjective/intuitive) assessments. The quantitative aspect is extremely important. It provides a standardized point of reference by which to build upon. True: often times playing by the numbers is the first and only way to achieve desired results. However often times, it's necessary to make adjustments that lean towards "qualitative."

So long as the end result is legal, within a specified color space: aRGB, sRGB or CMYK, and it is free of artifacts (especially under intense scrutiny), all is well.

This is an aerial overview. There are nuances I've not brought up.
 

[Doug Gray]

I have tried it and from memory this was something that came from Dan Margulis many years ago. 

Actual testing of images revealed nothing to be gained by stepping in and out of LAB over staying in RGB and applying sharpening to luminosity layer add the risk of the loss of data being visible and I made the decision to not risk it.

I must say that my testing limited to comparing only a few images and I never came across any overwhelming evidence to support the use of LAB sharpening over luminosity.  If there is any that demonstrates this I would be very interested to see.  To be fair I think the risk probably greater working in 8 bit rather than 16

I guess at the end of the day to quote (misquote?) Deng “it does not matter as long as it catches mice” 

Lab is pretty awful in 8 bits. Quantization problems doing any sort of editing are worse than 8 bit ProPhoto. However, it's not an issue at all in 16 bit Lab. And always convert any space to 16 bits before converting to Lab.

I do most of my repro type stuff in Lab. Way easier to spot check colors in Photoshop. Also, make sure you select 32bit display mode for the info panel. It displays Lab in decimal with extended precision.

[TonyW]

I have and still do use LAB and find it worthwhile for some things see this post
http://forum.luminous-landscape.com/index.php?topic=116641.0

What I am saying is that personal experience indicates that claimed advantages to RGB>LAB>RGB sharpening. are not that great and certainly not necessarily better than staying in your editing space and using luminosity layer or fade to luminosity.  Results from the two not exactly the same but I have never seen a repeatable example that shows superiority of one method over another - I would welcome seeing such examples

From what I understand the losses in 8 bit around 30 levels from 256, losses in 16 bit representing a much smaller percentage and some say negligible.  I have found myself often in the situation restoring images having been presented with 8 bit images without any option for 16 bit acquisition. 

So my view may be coloured, nevertheless I feel the relative safe option is to stay in editing space for either 8 or 16 bit and only step out for those that may be very difficult to do otherwise and then only within 16 bit workflow

[BobDavid]

I don't do any image processing in 8-bit mode. The trick to sharpening is to do it so that it's unnoticeable. For me, that's the litmus test for high quality sharpening.

[Peterretep2]

Unsharp Mask or Smart Sharpen? After some adjustments to the luminosity layer I'll use both sharpening methods. Using an unsharp mask somewhere in the Amount 15 - Radius 35 range gives a nice boost to contrast and clarity. I follow that with Smart Sharpen and like the results quite well.

Peter

[digitaldog]

Lab is pretty awful in 8 bits.

Yup, awful and unnecessary. But it's not our data to worry about