To Sharpen or not to sharpen ~ that is the question...

[revi]

Sorry, but for me, these show clear signs of over sharpening, especially around the head of the bird; there's a very nice and smooth halo outlining the beak and the head. The tree also shows oversharpening, especially the side opposite of the bird (this seems actually to have gotten worse in the reworked images). And I realise that Jiro and Colin had to work from the small JPG posted, so they had a 'small' handicap

Best solution if you started from RAW: redo from start... If it's a camera JPG, you could try lowering the in-camera sharpening.

regards

Remco

[tameigh]

OK, I redid it from RAW



This one still looks a touch soft to me, but I am wondering if I just don't have the right expectations, don't recognize the cues. Also it may be the subject, a "downy" woodpecker. Naturally soft.

[Colin Southern]

OK, I redid it from RAW



This one still looks a touch soft to me, but I am wondering if I just don't have the right expectations, don't recognize the cues. Also it may be the subject, a "downy" woodpecker. Naturally soft.

Hi Tim,

How many sharpening passes did you use, and what were the settings?

[tameigh]

I did one pass at a radius of .5, 400%, and a second pass at radius of 3.9, 40%, then reduced the image for display using bicubic sharper.

Edit: I just read the FAQ on uploading pics, and neglected the final sharpening step, but still, I don't know what to look for. Here is the raw file from another shot of the same bird, if anybody would be so kind as to look at it. This shot has the bird in a little patch of sun. Like the first, oversharpened version.

This version includes a resize using bicubic, and a final pass at sharpening of .5 r and 50%



_MG_0088.CR2 - 9.4 MB

[Colin Southern]

Here's how I would do it ...
Tiny brightness adjust in ACR

USM 300% @ 0.3
USM 40% @ 4.0
Down-sample to 900 x 600 using bicubic smoother
USM 100% @ 0.3

[tameigh]

On your set of 2 examples, they do look "over-sharpened" because they are bright and with high-contrast.

I think that the high contrast nature of the tree bark, and the black and white of the bird, which is in a little patch of sunshine in the picture, and so a harsh light, do contribute to the perception. This is something I have been trying to understand. I guess that I thought of sharpness as a universal standard of clarity, but I can see that there is a lot of room for "artistic" adjustment. I think I will print a test pattern from my photo printer, and photograph it and then try to restore it to it's original as closely as I can to get some kind of quantitative idea of what sharpening works best with my camera.

[Dave Humphries]

OK, I redid it from RAW



This one still looks a touch soft to me, but I am wondering if I just don't have the right expectations, don't recognize the cues. Also it may be the subject, a "downy" woodpecker. Naturally soft.

Hi Tim,

What radius are you using to sharpen?
When are you sharpening, what amounts?

That radius still looks a bit wide to me (e.g. 1px or more), hence even though, as you say, it looks a bit soft, it has a halo around the head and beak.

For this kind of subject matter, I rarely use more than 0.3 or 0.4px radius, I know that's smaller than what Colin recommends, but it works for me (and I learnt all I know about sharpening from Colin).

EDIT: Ah, I see everyone has been busy whilst I was watching TV
I should have refreshed the page before answering

Yeah, I'd go with what Colin did, maybe missing the 4px @ 40% out though

Cheers,

[tameigh]

Here's how I would do it ...
Tiny brightness adjust in ACR

USM 300% @ 0.3
USM 40% @ 4.0
Down-sample to 900 x 600 using bicubic smoother
USM 100% @ 0.3

Thank you.

[tameigh]

Hi Tim,

What radius are you using to sharpen?
When are you sharpening, what amounts?

That radius still looks a bit wide to me (e.g. 1px or more), hence even though, as you say, it looks a bit soft, it has a halo around the head and beak.

For this kind of subject matter, I rarely use more than 0.3 or 0.4px radius, I know that's smaller than what Colin recommends, but it works for me (and I learnt all I know about sharpening from Colin).

Cheers,

What are the visual cues that a "radius looks wide"? I did use a 3.9 at one pass, but I am following a cookbook right now, so to speak, I have no idea why any of the steps are done the way they are.

Thank you everybody for your time, btw.

[Dave Humphries]

What are the visual cues that a "radius looks wide"? I did use a 3.9 at one pass, but I am following a cookbook right now, so to speak, I have no idea why any of the steps are done the way they are.

Thank you everybody for your time, btw.

The way I look at it is that 4px (or 3.9) applied to the full size image will downsize to 1px, if the downsize is at that same 4:1 ratio (and it often is), which is why I often miss that step out with subjects lik this.

As people don't say my pictures look soft, I figure I get away with it most times

Occasionally, if the capture is soft I will do one or more* passes to try to get some edge in it, but it's usually best to not publish if it's that bad

* e.g. 50% at 2px + 60% at 1px + 70% at 0.4px, but again I may vary those amounts and radii to suit the picture content, I also might miss the last step if the downsize isn't great, due to excessive cropping and just do that as the final sharpen at say 70-100% at 0.3px

Cheers,

[tameigh]

ok,

I took jiro's advice here and brought down the highlights a little bit with the burn tool. I cloned out the halo on the beak and head on pixel at a time, and I ran one more pass of sharpening at .5 r, 45%, all of this on the shrunken image.

I think that the head is softer than the body, and this may be a DOF issue with the capture. I wrote off the tree, as not really part of the exercise.

I am very happy with the results, compared to what I have been getting, of course I would be interested in any comments regarding my effort here, but I feel like I have raise my game a lot from this thread today.

[Dave Humphries]

I am very happy with the results, compared to what I have been getting, of course I would be interested in any comments regarding my effort here, but I feel like I have raise my game a lot from this thread today.

Hi Tim,

That's better in every respect except noise, I'd suggest a third party add-on like Neat Image; kill the noise (straight after ACR) before doing anything else* in CS5 - * including the 0.3px at 300% capture sharpen - that way, you can get away with 0 or 1 threshold when sharpening, here it looks like it could have done with being a bit higher.

What you're learning will stand you in good stead for years to come, keep at it,

[Colin Southern]

I did one pass at a radius of .5, 400%

I'd say that's 9/10 of your problem right there. Occasionally there's a benefit to capture sharpening @ 0.2 @ 400%, but never at 0.5 pixel. 0.5 @ 400% would be almost guaranteed to give frosting on high-frequency components (like the finer bark detail).

[jiro]

Good Job, Tim. Take Colin's advice - do some noise reduction to make your image look better. A very good improvement indeed.

[Colin Southern]

Take Colin's advice - do some noise reduction to make your image look better.

Um, that's Dave's suggestion - I never use noise reduction

[jiro]

My bad. Sorry for the mistake.

[Colin Southern]

My bad. Sorry for the mistake.

That's OK - people get us confused all the time, although I'm not sure why, as it's easy to figure out ...

... Dave is the good-looking one!