Trying to Understand - Around PreSharpening

[rabanito]

I have been looking this AI Sharpen App/Topaz.

As I understand it, it should be useful for PreSharpening. I cannot imagine any other use and gladly accept corrections and teachings if otherwise.

I opened an image in PS (XY.NEF), duplicated it three times (3 copies) and processed them separately in Topaz AI Sharpen, NIK Presharpen (both w/default values) and through a HighPass Filter (Radius 1 pixel - Modus:Overlay) separately (see attached images at 200%).
The also attached original photo was not sharpened

I do not see real reasons why I should prefer one of the sharpened images to the others.
Are there reasons that I don't see? (Don't forget that I am very inexperienced)

Moreover, it would seem to me that NIK Sharpen Pro and the Highpass methode offer more controls than Topaz.
Is that right?

Thank you for any input

[Bart_van_der_Wolf]

I have been looking this AI Sharpen App/Topaz.

Hi,

As I understand it, it should be useful for PreSharpening. I cannot imagine any other use and gladly accept corrections and teachings if otherwise.

Yes, 'Sharpen AI' is mainly intended for 'Capture Sharpening', although it can also be used to correct unsharpness that's caused by resampling and distortion/perspective correction.

I opened an image in PS (XY.NEF), duplicated it three times (3 copies) and processed them separately in Topaz AI Sharpen, NIK Presharpen (both w/default values) and through a HighPass Filter (Radius 1 pixel - Modus:Overlay) separately (see attached images at 200%).
The also attached original photo was not sharpened

Good, although different Raw converters use different trade-offs that will already lead to small differences in sharpness. It helps if the Raw conversion did not change too much of the per pixel blur that was spread out to neighboring pixels. 

I do not see real reasons why I should prefer one of the sharpened images to the others.
Are there reasons that I don't see? (Don't forget that I am very inexperienced)

That partly depends on the further processing steps you anticipate as being required. Downscaling or upscaling require two obvious different approaches. Upsampling will magnify the visiblity of any shortcomings with sharpening. Artifacts will become more visible, and from larger viewing distances. And with down-sampling, it will increase the likelihood of creating aliasing artifacts. I'll address that in some more detail in a followup response.

Moreover, it would seem to me that NIK Sharpen Pro and the Highpass methode offer more controls than Topaz.
Is that right?

They do, but is that helpful or is it intended for other things than Capture Sharpening. I think the latter. Some of it is to hide artifacts that are formed as by-products of the operation, like 'sharpening' halos. And some of it is more useful for Creative 'sharpening' (more for local contrast and edge enhancement) than actual Capture Sharpening.

Cheers,
Bart

[Mark D Segal]

What is pre-sharpening? I haven't heard of this, so I'm asking. Do you mean by any chance "Capture Sharpening"?

Why examine the results at 200%. They can't be assessed properly (or at least I can't) at that magnification, because the pixelation of the excessive magnification interferes with an appreciation of the sharpening results. The results should be viewed at no more than 100%. And to be more pointed, hard to actually be certain on a monitor. It is best to view the outcomes in prints, if printed output is to be the end-game.

May I ask why you did not also try Photokit Sharpener 2? It's a free download from the Pixelgenius website; it will no longer be updated, but works well with current versions of LR and PS. To my mind, it remains the best technical solution to sharpening with maximum flexibility to achieve pleasant sharpness without apparent halos.

Looking at your photos, the halos seem excessive to me in the two right-most reproductions - but note my cautions just above.

[rabanito]

What is pre-sharpening? I haven't heard of this, so I'm asking. Do you mean by any chance "Capture Sharpening"?

Well, I thought that those were synonyms.
That is the name used by NIK for their "Sharpener Pro3:RAW Presharpener"

I found it also in articles like
"Image Sharpening" by Fraser/Schewe e.g. Page 108
"https://help.phaseone.com/en/CO8/Editing-photos/Details/Sharpening?p=1"
"https://blog.thomasfitzgeraldphotography.com/blog/2017/8/capture-one-tips-for-creating-a-better-default-image"

Sorry for the confusion. I really thought they were interchangeable terms.

Why examine the results at 200%. They can't be assessed properly (or at least I can't) at that magnification, because the pixelation of the excessive magnification interferes with an appreciation of the sharpening results. The results should be viewed at no more than 100%. And to be more pointed, hard to actually be certain on a monitor. It is best to view the outcomes in prints, if printed output is to be the end-game.

Again, I am not trying to judge sharpness at this stage but to illustrate why I don't think there is a significant difference among the three sharpened photos (a lowly amateur's opinion of course).
In the book of Fraser/Schewe there are also examples at 200 & 400 % as well. With the same or similar caveat you put in your response of course.

Actually I never judge "Capture Sharpening" on a print, this is one of the first stages in my workflow, often far away from Output Sharpening.

May I ask why you did not also try Photokit Sharpener 2? It's a free download from the Pixelgenius website; it will no longer be updated, but works well with current versions of LR and PS. To my mind, it remains the best technical solution to sharpening with maximum flexibility to achieve pleasant sharpness without apparent halos.

Actually I never tried it. Thanks for the idea. I think it can still be downloaded 

[Bart_van_der_Wolf]

[...]
Again, I am not trying to judge sharpness at this stage but to illustrate why I don't think there is a significant difference among the three sharpened photos (a lowly amateur's opinion of course).
In the book of Fraser/Schewe there are also examples at 200 & 400 % as well. With the same or similar caveat you put in your response of course.

I also see no problem with close inspection on zoomed-in regions-of-interest (ROIs). Our displays are a poor proxy of what the final result will be, due to its (in many cases) poor resolution and potentially high contrast. It also doesn't take final printed output resolution into consideration, which would require zooming-out or downsampling to achieve actual output size. And then it's still lacking output sharpening at this stage anyway.

Actually I never judge "Capture Sharpening" on a print, this is one of the first stages in my workflow, often far away from Output Sharpening.

At the stage of Capture Sharpening one should focus on restoring Capture losses, and nothing more. Anything else will potentially lead to issues with Creative 'sharpening', and with resampling for the final image size.

Cheers,
Bart

[texshooter]

This source explains why it is better to capture sharpen (aka pre-sharpen) at the end of the workflow rather than at the beginning.

https://m.youtube.com/watch?v=_x8KqIYD5Qs

[Bart_van_der_Wolf]

This source explains why it is better to capture sharpen (aka pre-sharpen) at the end of the workflow rather than at the beginning.

https://m.youtube.com/watch?v=_x8KqIYD5Qs

His premise is kind of that because people don't know how to correctly Capture sharpen their images, they shouldn't do it as a first step.
 
I'd say, learn how to do it properly, and do use it as a first step (together with noise management).
But if you don't master it, well yeah, maybe delay doing it until later in the process.

As I said in the post before:

At the stage of Capture Sharpening one should focus on restoring Capture losses, and nothing more.

Cheers,
Bart

[hubell]

Looking at your photos, the halos seem excessive to me in the two right-most reproductions - but note my cautions just above.

I agree. The two right-most crops have that oversharpened, "crunchy" look that I personally find very unattractive. The AI Sharpen crop looks way better to my taste.

[texshooter]

I'd say, learn how to do it properly, and do use it as a first step (together with noise management).

Are there advantages to capture sharpening (I`m only refering to deconvolution in this context, not unsharp mask) at the beginning instead of at the end?  Specifically,

1) Do programs pano stitch or focus stack photos better if a little deconvolution capture sharpening is applied first?
2) Does applying deconvolution capture sharpening using plugins like Topaz and Raw Therapee after all other PS edits, such as color and tone adjustments, cause unwanted artifacts?
3) Do deconvolution capture sharpening algorithms (whether used by Camera Raw or third parties like Topaz) do a better job if applied early rather than later?

The reason I ask is because if there is no real advantage these days to capture sharpening at the beginning,  I would prefer to wait until all my PS edits are done and instead do all my sharpening at the end (including capture, creative, and output sharpening). It just seems more streamlined that way, plus I can use masks to apply capture sharpening more precisely.  And because I can always use the new Camera Raw sharpen filter in PS, which is non-destructive.

Something tells me the old wisdom of capture sharpening in Camera Raw is a relic of the past before we had the Camera Raw sharpen filter inside PS CC.   Times do change.

[rabanito]

2) Does applying capture sharpening after other PS edits, such as color and tone adjustments, cause unwanted artifacts?

Fraser/Schewe recommend  in their book "Image Sharpening" performing any major tonal correction before doing any sharpening or noise reduction because these can easily undone
Not because of possible artifacts, at least in that section

I guess that that is why in LR Sharpening an NR at almost at the end of the workflow "develop".

[texshooter]

I guess that that is why in LR Sharpening an NR at almost at the end of the workflow "develop".

I suspect you mean something differently than I do with the term "end of the workflow."  I don't use Lightroom, so the end of the workflow for me does not mean any single panel inside Lightroom (or Camera Raw), but rather inside Photoshop and after all edits have been made, except sharpening and resizing.

[rabanito]

I suspect you mean something differently than I do with the term "end of the workflow."  I don't use Lightroom, so the end of the workflow for me does not mean any single panel inside Lightroom (or Camera Raw), but rather inside Photoshop and after all edits have been made, except sharpening and resizing.

Well it was meant just as an example.
The main point in my posting is the quote of Fraser/Schewe
Sorry for the confusion

[Bart_van_der_Wolf]

Are there advantages to capture sharpening (I`m only refering to deconvolution in this context, not unsharp mask) at the beginning instead of at the end?  Specifically,

1) Do programs pano stitch or focus stack photos better if a little capture sharpening is applied first?

That is likely the case, depending on the specific application.

2) Does applying capture sharpening after other PS edits, such as color and tone adjustments, cause unwanted artifacts?

Not necessarily.

3) Do deconvolution capture sharpening algorithms (whether used by Camera Raw or third parties like Topaz) do a better job if applied early in Camera Raw rather than later In Photoshop?

Yes, most deconvolution algorithms do best when the data is as Raw as possible, just the photons (shot noise) and in linear gamma space, 16-bits/channel or better is preferred.

The reason I ask is because if there is no real advantage these days to capture sharpening at the beginning,  I would prefer to wait until all my PS edits are done and instead do all my sharpening at the end (including capture, creative, and output sharpening). It just seems more streamlined that way. And because I can always use the new Camera Raw sharpen filter in PS, which is non-destructive.

I kind of agree, but deconvolution requires as Raw data as possible. Once one e.g. changes contrast, the relationship between photons and different sources of noise in adjacent pixels is distorted, making it harder to unscramble the omelet, as it were.

Something tells me the old wisdom of capture sharpening in Camera Raw is a relic of the past before we had the Camera Raw sharpen filter inside PS CC.   Times do change.

They do, and that's why new applications like Topaz Sharpen AI spark my interest. These kinds of applications function on the basis of pattern recognition, not on absolute photon counts like deconvolution does.

Cheers,
Bart

[kers]

I just tried my old way of sharpening to Topaz Ai sharpening and (again)Topaz does a good job, but goes wrong in some areas.
i started of with a 64 asa  100% perfect sharp image on the houses.
The trees above the roof get some kind of pattern- ( i used Sharpen in Topaz Ai)
If i print the details at 150dpi (- whole image= 140cm wide without upscaling) it appears also visible in print.
so i keep to my old ways of sharpening... also saving me 3 minutes /image on a mac.
PS this was done on the bases of a NEF developed in LRcc- other ways of developing might make a difference

[bjanes]

Fraser/Schewe recommend  in their book "Image Sharpening" performing any major tonal correction before doing any sharpening or noise reduction because these can easily undone
Not because of possible artifacts, at least in that section

I guess that that is why in LR Sharpening an NR at almost at the end of the workflow "develop".

Lightroom and ACR use parametric editing. When you make adjustments in editing, the pixels in the raw file are not changed but rather the adjustments are stored as instructions in metadata. The on screen preview is altered, but the actual pixels in the file are not altered until one opens the raw file in Photoshop at which time the recorded adjustments are applied in an order that ACR considers optimal rather than in the order of the edits. The order in which you make the adjustments is not material. In the case of Lightroom, the adjustments are made when the file is exported or printed.

Regards,

Bill

[rabanito]

... the order in which you make the adjustments is not material. In the case of Lightroom, the adjustments are made when the file is exported or printed.

Regards,

Bill

Thanks 

[digitaldog]

Fraser/Schewe recommend  in their book "Image Sharpening" performing any major tonal correction before doing any sharpening or noise reduction because these can easily undone
Not because of possible artifacts, at least in that section

Further, they suggest doing all the work on layers, so you can turn them on or off, alter opacity and never bake them into the underlying data until you print or save out a flattened iteration.

I guess that that is why in LR Sharpening an NR at almost at the end of the workflow "develop".

What Bill wrote. The order makes no difference. It's not a 'do at the end' concept whatsoever. Capture sharpening is used to overcome softness in capture and isn't output specific (it's output agnostic).
Output sharpening which is somewhat limited in LR/ACR is based on the output and resolution of the data being output. AND based upon the capture sharpening already applied earlier in making parametric edits.

[rabanito]

Further, they suggest doing all the work on layers, so you can turn them on or off, alter opacity and never bake them into the underlying data until you print or save out a flattened iteration.
What Bill wrote. The order makes no difference. It's not a 'do at the end' concept whatsoever. Capture sharpening is used to overcome softness in capture and isn't output specific (it's output agnostic).
Output sharpening which is somewhat limited in LR/ACR is based on the output and resolution of the data being output. AND based upon the capture sharpening already applied
earlier in making parametric edits.

Thanks

[Lightsmith]

I only sharpen as my very last step after having resized and image and then exported it in a chosen format. I apply more sharpening for a 400 x 400 pixel web image than for a 3000 x 4500 image that is going to be used to create a wall print.

My first step is to do a Levels adjustment and the second step is to adjust Contrast. A Contrast adjustment makes an image look sharper without the degradation that results from sharpening.

To evaluate different sharpening applications I will download 3 of them and then take an image of a head shot and make 4 different 4x4 inch size images of a small section of the face - eyes and mouth and cheek, and one image sharpened manually with Photoshop. With all four brought in as layers I then combine them into a single flattened image which I can view on screen and also make a 8x10 print.

It takes under an hour to download 3 sharpening applications and install them on a computer and create the 4 image files and a composite. I have found that this composite makes it easy for me to see the level of detail preserved with the eye and eyelashes and any issues with the color fidelity that has resulted from the sharpening.

A common complaint with some cameras like the original Canon 5D was that people tended to look like plastic dolls and this was the result of excessive noise suppression in the camera with very high ISO settings. Sharpening applications often do the same.