Output sharpening calculations for optimum settings.

[rogue tog]

Hello everyone!

Having been a Lightroom user for the past years, I was well used to the ease of the automated output sharpening, the app offered. However, making a decision recently to move to Capture One, I realised that I needed to learn more about Output sharpening settings and to how optimise them.

The software, like many others, offers three sliders for Output sharpening.

1. Amount
2. Radius
3. Threshold

Reading this article made me come to the following conclusions.

1. Amount --> Subjective, depends on user preference (similar to the effect "Low, Standard, Strong" settings in Lightroom output sharpening have, in a very rough maner?)

2. Radius --> Objective (seems to be the one parameter that can be defined based on mathematical terms and depends on several factors such as resolution, viewing distance, etc.

3. Threshold --> Masking off less areas of lower contrast. Subjective.


The opinion that radius can be calculated for output sharpening, was based on the fact that the page offers an actual calculator for that exact purpose! However the engineer in me, would like to know how it works and how I can expand and use it in other cases.

What are the equations that govern the radius for optimum output sharpening? How can I calculate
the radius for:

1. Screen viewing ie downsize for web?
2. Printing?

Thanks and apologies for the geeky post

[xpatUSA]

The opinion that radius can be calculated for output sharpening, was based on the fact that the page offers an actual calculator for that exact purpose! However the engineer in me, would like to know how it works and how I can expand and use it in other cases.

What are the equations that govern the radius for optimum output sharpening? How can I calculate
the radius for:

1. Screen viewing ie downsize for web?

Welcome,

As you've likely found out, the subject is huge with varying opinions, terminology and methods!

As to downsizing for the web or your screen - this article may offer a clue or two:

http://kronometric.org/phot/iq/Down%...%20methods.htm

2. Printing?

Can't help you much there, I don't print. Printing often involves re-sampling either deliberately by your good self by a method known only to Adobe or the printer driver. Re-sampling changes the sharpness, smoother for up and sharper for down. No one equation fits all, sad to say.

Thanks and apologies for the geeky post

No need for apologies, there's a fellow geek or two here.

[DanK]

All three can be expressed mathematically, but the best amount for all three is subjective.

[pnodrog]

Good luck.

I use the Lightroom options for all my output sharpening which it seems to do fairly well. Obviously the formula will have to take into account pixel pitch related to overall image size and assumed viewing distance. Not too dissimilar to acceptable DOF calculations. In general the radius for prints needs to be significantly larger than that for screen viewing (300/72 = apx 4.1 or 360/96 = apx 3.7) provided the viewing distance/angle is similar (ban pixel peekers.)

I think I will reiterate good luck....

P.S. Another of many articles on the net.

[dje]

Hi Rogue Tog

I don't know the exact detail but I think the calculator referred to in your link is calculating a pixel radius guideline by dividing the viewing distance by dpi (or probably more accurately ppi) and mutliplying by a constant. I believe this is a calculation of the distance in an image (related to pixel size) that can be resolved by the human eye at a certain viewing distance.

Just to complicate things, bear in mind that the pixel radius setting used in a sharpening algorithm is often related to the standard deviation of the spatial distribution curve that is used to distribute the filter values about the central pixel. The smaller the SD, the finer the distribution and consequent sharpening effect. But the relationship between pixel radius setting and SD would vary from one algorithm to another. It should only be regarded as a guide.

Hi Paul, could you please enlighten us as to the significance of the values 72 and 96?

Dave

[pnodrog]

Hi Rogue Tog

I don't know the exact detail but I think the calculator referred to in your link is calculating a pixel radius guideline by dividing the viewing distance by dpi (or probably more accurately ppi) and mutliplying by a constant. I believe this is a calculation of the distance in an image (related to pixel size) that can be resolved by the human eye at a certain viewing distance.

Just to complicate things, bear in mind that the pixel radius setting used in a sharpening algorithm is often related to the standard deviation of the spatial distribution curve that is used to distribute the filter values about the central pixel. The smaller the SD, the finer the distribution and consequent sharpening effect. But the relationship between pixel radius setting and SD would vary from one algorithm to another. It should only be regarded as a guide.

Hi Paul, could you please enlighten us as to the significance of the values 72 and 96?

Dave

Typical screen PPI as apposed to typical print DPI.

[rogue tog]

Thanks for all the answers guys!

I realise this (as expected....) is way more compliated than I hoped it would be... Should I just use my eyes and judgement then?

I mean I could, but for someone who likes numbers this can have detrimental effects on my psychology :-P

[Manfred M]

Thanks for all the answers guys!

I realise this (as expected....) is way more compliated than I hoped it would be... Should I just use my eyes and judgement then?

I mean I could, but for someone who likes numbers this can have detrimental effects on my psychology :-P

Yes. If you read Fraser & Schewe's well regarded book on the subject "Real World Image Sharpening with Adobe Photoshop, Camera Raw, and Lightroom" that is exactly what they recommend. Each image is different, so there is no single formula that can be applied.

For output sharpening, I rarely sharpen for display to screen (as we get some "inherent" sharpening from the downsampling process), but I do when printing and there the paper type and paper finish come into play. A matte paper needs more sharpening than a glossy paper due to the ink absorption characteristics of the different finishes, as well as the way the paper itself reflects light.

I do assume that any import and in-process sharpening has been carried out ahead of time.

The reason I put the word inherent in quotes is that downsizing does not actually sharpen the image at all, but the image seems sharper. This is the same phenomenon we see when we look at an image on the screen on the back of our camera (at screen size) and when we get back into PP the image is not as sharp as we had thought. Screen size matters when assessing sharpness, so downsampling does make our images look a bit sharper.

[JBW]

Hello everyone!

Having been a Lightroom user for the past years, I was well used to the ease of the automated output sharpening, the app offered. However, making a decision recently to move to Capture One, I realised that I needed to learn more about Output sharpening settings and to how optimise them.

The software, like many others, offers three sliders for Output sharpening.

1. Amount
2. Radius
3. Threshold

Reading this article made me come to the following conclusions.

1. Amount --> Subjective, depends on user preference (similar to the effect "Low, Standard, Strong" settings in Lightroom output sharpening have, in a very rough maner?)

2. Radius --> Objective (seems to be the one parameter that can be defined based on mathematical terms and depends on several factors such as resolution, viewing distance, etc.

3. Threshold --> Masking off less areas of lower contrast. Subjective.


The opinion that radius can be calculated for output sharpening, was based on the fact that the page offers an actual calculator for that exact purpose! However the engineer in me, would like to know how it works and how I can expand and use it in other cases.

What are the equations that govern the radius for optimum output sharpening? How can I calculate
the radius for:

1. Screen viewing ie downsize for web?
2. Printing?

Thanks and apologies for the geeky post

There are actually two sharpening options in Capture 1. One is found listed as 'Sharpening'. The other is hidden in the 'Lens' tool.

Sometimes I use both sharpening options but many times I find he one hidden in the Lens tool is all I need.
Brian

[rogue tog]

Yes. If you read Fraser & Schewe's well regarded book on the subject "Real World Image Sharpening with Adobe Photoshop, Camera Raw, and Lightroom" that is exactly what they recommend. Each image is different, so there is no single formula that can be applied.

For output sharpening, I rarely sharpen for display to screen (as we get some "inherent" sharpening from the downsampling process), but I do when printing and there the paper type and paper finish come into play. A matte paper needs more sharpening than a glossy paper due to the ink absorption characteristics of the different finishes, as well as the way the paper itself reflects light.

I do assume that any import and in-process sharpening has been carried out ahead of time.

The reason I put the word inherent in quotes is that downsizing does not actually sharpen the image at all, but the image seems sharper. This is the same phenomenon we see when we look at an image on the screen on the back of our camera (at screen size) and when we get back into PP the image is not as sharp as we had thought. Screen size matters when assessing sharpness, so downsampling does make our images look a bit sharper.

That is interesting. I agree with you on print sharpening, but when it comes to screen, I always thought that downsampling made things softer.

In Capture 1 this becomes really noticeable when using the proofing view.

Furthermore the article linked in my first post mentions:

"Even if an image already looks sharp when viewed on-screen, resizing it to less than 50% of its original size often removes any existing sharpening halos. One usually needs to apply output sharpening to offset this effect:"

[rogue tog]

There are actually two sharpening options in Capture 1. One is found listed as 'Sharpening'. The other is hidden in the 'Lens' tool.

Sometimes I use both sharpening options but many times I find he one hidden in the Lens tool is all I need.
Brian

JBW I believe the sharpening options you are referring to, would be suitable as Capture and Creative sharpening stages.
The app also caters for Output sharpening, which is a process that has to do with the final size / medium used for the final image.

While I am sure that one may use the Sharpening tab to achieve same goal, I believe it would be much easier to do so throught the Export procedure.

[xpatUSA]

That is interesting. I agree with you on print sharpening, but when it comes to screen, I always thought that downsampling made things softer.

In Capture 1 this becomes really noticeable when using the proofing view.

Furthermore the article linked in my first post mentions:

"Even if an image already looks sharp when viewed on-screen, resizing it to less than 50% of its original size often removes any existing sharpening halos. One usually needs to apply output sharpening to offset this effect:"

This might be of interest. The author is a noted authority on re-sampling.

http://kronometric.org/phot/processi...%20methods.htm

Here's a file to play with:

http://kronometric.org/phot/processi...gs_lg_orig.png

HTH,

[Manfred M]

That is interesting. I agree with you on print sharpening, but when it comes to screen, I always thought that downsampling made things softer.

In Capture 1 this becomes really noticeable when using the proofing view.

Furthermore the article linked in my first post mentions:

"Even if an image already looks sharp when viewed on-screen, resizing it to less than 50% of its original size often removes any existing sharpening halos. One usually needs to apply output sharpening to offset this effect:"

That really depends on the downsizing algorithms that the software is using. I've seen some where the halos seem to look enhanced, so it's hard to make a hard and fast call on that as the downsizing algorithms will be written into the software that you are looking at to display the images.

That is somewhat beside the point; we start with an image that is in the tens of megapixels in size and the displays that we load the images on are often use a window that is in the 2MP range. The images are being viewed on anything from a large computer screen (I happen to use a 27" model for as my main editing screen) and the display can be anything. A more size standard screen, a laptop screen, a tablet or a phone. What size of display are you going to sharpen to?

As I don't know, I just don't bother with an output sharpening step. My input sharpening / in-process sharpening seems to be "good enough" for all of these displays.

[dje]

Yes I think it depends on the algorithm. For example, I found PS's Bicubic Sharper included too much sharpening for my taste and was introducing halos whereas Bicubic was smoother and i used this with very mild output sharpening. Now that i am using Affinity Photo, it includes Lanczos as a downsizing option and it gives results that are just right for my taste!

Dave

[dje]

JBW I believe the sharpening options you are referring to, would be suitable as Capture and Creative sharpening stages.
The app also caters for Output sharpening, which is a process that has to do with the final size / medium used for the final image.

While I am sure that one may use the Sharpening tab to achieve same goal, I believe it would be much easier to do so throught the Export procedure.

I think that's the best way to go. It allows you to use the "main" sharpening adjustment for your master image and then the Export sharpening is used for a specific output.

The Lens correction sharpening in Capture One is for diffraction correction and also for sharpness falloff near the edges. Never tried it though.

Dave

[xpatUSA]

Yes I think it depends on the algorithm. For example, I found PS's Bicubic Sharper included too much sharpening for my taste and was introducing halos whereas Bicubic was smoother and i used this with very mild output sharpening. Now that i am using Affinity Photo, it includes Lancros as a downsizing option and it gives results that are just right for my taste!

Dave

That would be Lanczos I think, Dave, please pardon my pedantry . . .

Lanczos is my favorite for my own purposes. There is a version "Lanczos2" which is a bit sharper; don't know if Affinity offers it, though.

I played with some target shots today (144-cycle Siemens Star) which showed quite a lot relevant to this discussion.

But, since I apparently "tend to do this" and since I "generally will bring up points that to some will appear to be fairly minor", I won't bother to post my "fairly minor" results.

[dje]

That would be Lanczos I think, Dave, please pardon my pedantry . . .

Lanczos is my favorite for my own purposes. There is a version "Lanczos2" which is a bit sharper; don't know if Affinity offers it, though.

It would indeed be Lanczos Ted, thanks for picking that up. Pedantry welcome

Affinity has two versions of Lanczos

• Lanczos 3 Separable
• Lanczos 3 Non - separable

Don't know much about how they work though

Dave

[xpatUSA]

Affinity has two versions of Lanczos

• Lanczos 3 Separable
• Lanczos 3 Non - separable

Don't know much about how they work though

Dave

Andy of Affinity summarizes them here:

https://forum.affinity.serif.com/ind...&comment=53943

And provides two links that'll make your eyes water.

Non-separable looks "better" to my pedantic eye . .

[Cantab]

I played with some target shots today (144-cycle Siemens Star) which showed quite a lot relevant to this discussion. But... I won't bother to post my "fairly minor" results.

Ted, you have wetted my appetite. I'm curious about your results -- even if few others may be interested.

[xpatUSA]

Ted, you have [whetted] my appetite. I'm curious about your results -- even if few others may be interested.

OK Bruce, just for you . .

The 144-cycle Siemens Star target is by Bart van der Wolf and is sinusoidal . . also comes with a couple of slant-edges for MTF and Edge Response measurements.

The star is 4" diameter, shot from about 4' under warm LED lighting. Camera focused manually, magnified live view.

3 crops follow . .

Out of Sigma's converter, no adjustments:



Already some artifacts because I shoot that camera in Sigma's 2x2 on-sensor binned mode (1/2 size pics).


Sharpened "to taste" in RawTherapee (de-convolution method):



A lot of sharpening just for the purpose of illustration!

Reduced for posting here - ratio 1600/2336:



Chose Nearest Neighbor deliberately to exaggerate artifacts caused by downsizing.

Gotta love that moiré, eh?