Sharpening workflow question

I am new to digital RAW processing - finally gave up my film camera earlier this year. I did notice that shooting RAW puts weight on as I spend too much time on my computer. So less time spent trying to learn proper sharpening techniques is essential. Please help!
I've read Bruce Fraser's sharpening book and sadly I still am somewhat confused. It doesn't speak to my situation for I now use ACR 4.1 for initial capture sharpening and I don't print my photos, preferring to view them online at 1100 pixels (at it greatest length) on my NEC 20WMGX2 monitor (in case this is relevant).

My current workflow is a multiphase sharpening.
I use ACR to capture sharpen, using the methods described by Jeff Schewe in his Photoshopnews article. I use the sharpening mask and sliders.

My question is HOW to best sharpen the image after I have downsized it to a 1100 pixel length JPEG? Simple USM? Highpass sharpening? Smart sharpening? Recently I used high pass sharpening with a layer mask. This does require time to adjust the opacity slider and occasionally do localized adjustments with the mask. Time consuming but better results. I wonder if this is alot of work for nothing since the intended medium will be a Zenfolio web page rather than a print. Would using a USM PS action with predefined radius give quicker and indiscernibly similar results for a web based photo? If so what are good settings for the radius and threshold for a 1100 pixel length web based JPEG?

In case it is relevant, I shoot Canon 30D RAW files. I am a photo-enthusiast/hobbyist not turning out critical for client prints.

Thanks for your help.

Hi Phil

I have created an action to be used in (Mac or PC) CS, CS2 or CS3

you can download it:
here

and the new one:
here

The main difference is that

- the latter creates a layer not anymore of the background but on the current layer… (Thanks Bart!)

- By default the layer is set at 30% opacity (was 70), this is due to the new "clarity" function in both LR or ACR 4.1 as my action tend to create halos when used in conjunction with "clarity"

- Of course this action is to be used when your image is resized but still (hopefully in 16 bits)

Sharpening has been already discussed a lot here…
Links to some discussions:

http://www.openphotographyforums.com/forums/showthread.php?t=3299&highlight=sharpening+action

http://www.openphotographyforums.com/forums/showthread.php?t=3274&highlight=sharpening+action


http://www.openphotographyforums.com/forums/showthread.php?t=3013&highlight=sharpening+action

http://www.openphotographyforums.com/forums/showthread.php?t=2471&highlight=sharpening+action

http://www.openphotographyforums.com/forums/showthread.php?t=2621&highlight=sharpening+action

Enjoy!

Hi Phil,
the so called "capture" sharpening is to counteract the softness from the interpolation of the captured colors to produce RGB value for each pixel
If you shot in raw this step must be executed in the developping process. The strength depends on the interpolation quality.

Any other sharpening is an individual chosen and it depends on the image sharpness. There is not a general rule.
Adobe people suggest 3 sharpening. It may be good or not good.
If you want to print, for example, the need of sharpening depends on interpolation required (if any), from the quality of the interpolation and last, but not least, from the printer tecnhology.

Adobe resampling algorithms are not smart, so if you resample generally you have to sharpen.

Adobe people prescription are based on their products.

I know that using Adobe products better results are obtained using an "High Pass Filter".
But, perhaps, people don't know that USM is builded with some "High Pass Filter".
So, why Adobe USM is not so smart? The answer is: "algorithm quality".

Adobe is the industry standard, but that don't mean the Adobe algorithms are the best at every time .

This is not to denigrate Adobe products, but is to try to explaining the editing foundations.

Giovanni,

Good to hear from you again, I always like your posts!

Thanks for the input, but what scheme do you use and why?

Asher

Hi Asher,
nice to meet you.
I don't think any schema exists. Every shot is unique.
Yes there are some rules that can be followed as general guidelines, but prescriptions are very difficult.

I can suggest some guideline:

A good general rule for raw developping is to use the Raw converter of the camera maker (that is surely the best chosen for Canon). I verified that cannon DPP is very nice, and it cames for free when you by the camera.
Why?
The answer is very simple. I think that white balance is very important in a shot. The labs of the manufacturer knows the spectral response of the sensor and can transform the RGB captured data into the CIE_XYZ color space as no one can.
Every final color space (sRGB, Adobe 1998..) is builded from CIE_XYZ.

Better sRGB, Adobe 1998, or ProPhoto? Any color space has pro and cons. So learn them before deciding.

Better 16 bit or 8 bit? My general answer is 16 is better than 8, but 8 is more than enough.

Before any other editing, look at the noise. If you think you have it, make a denoising action.
Before denoising, think at the final output. If you have to print, remember that the image will be compressed, unless you print a very large format, so noise will be less visible. In any case don't apply too strong denoising (rarely plastic is fine).
Denoising must be performed before sharpening. You don't want to highlight noise.

Look at the histograms, they can tell you if the dynamic range may be increased.

Look at the shadows and highlights. If the scene has a strong contrast, you can try to extend the dynamic range with specific tools.

.....and so on.

Every argument should be explained in deep. I hope to speak about these themes soon and I hope that others can say their opinions.

So my answer is, no general prescription.
Actions may be good if you hav to do it for job and you are in hurry. But every shot need all your attention if you want to extract the maximum from it!

Phil Marion said:

My question is HOW to best sharpen the image after I have downsized it to a 1100 pixel length JPEG?

Click to expand...

You've asked a question that's important for everybody who wants to present their work on a monitor or other display device in the best possible way.

IMHO, there are 2 elements that need to be considered in turn.

• Best downsampling method for the image subject at hand
• Most effective sharpening method that doesn't amplify or create distracting artifacts

It's important to have the downsampling create a good basis for the final sharpening. Any artifact that stems from improper downsampling, will be amplified by subsequent sharpening. That's why I made a webpage on the subject of down-sampling some time ago, and I am planning to do another one on the output sharpening itself.

To recap my findings in a nutshell, common implementations of resampling algorithms don't do a very good job at down-sampling in general, Photoshop is no exception. That leads to avoidable problems in the final sharpening stage. I said avoidable, because when we know what goes wrong, we can create a workaround. What's more, even if we cut a few corners for time or workflow efficiency sake, we can still improve the quality we would otherwise get.

Even if we want to create a quick-and-dirty action for the situation you describe, 30D images (3504 x 2336 pixels) to 1100 pixels wide(?) screen, it is possible to get good results. Of course, for superior results one has to work harder.

Consider the following scenario for an action;
1. An uncropped landscape oriented image will be down-sampled by a factor of 3504/1100 = 3.19, so any detail in the original image smaller than 3.19 pixels will be reduced to 1 pixel. That means we should blur the original by that 3.19 amount if we want to avoid aliasing artifacts. A simple Gaussian blur of approximately 3.19x0.25 should get in the direction needed. One could test that rule of thumb a bit further to get it more accurate, but such a quick-and-dirty approach won't be perfect, and the 3.19 factor would be different for cropped or portrait oriented images, anyway.

2. Resize with regular bicubic resampling to 1100 pixels wide, you can also set the File|Automate|Fit image... command to a 1100 pixel wide image (and appropriate screen height), just make sure your default resampling method is bicubic.

3. Create a duplicate layer and change its blending mode according to this example. Use a SmartSharpen filter (lens blur) on that layer, a small radius like 0.5 or thereabouts with a quantity of 100 (a matter of taste) should be good starting point.

Of course there can be lots of improvements on this general scenario, but the general idea remains the same, blur, resize, Smart sharpen.

Bart

Bart

thanks a lot; I've been looking at bad downsampling with bicubic sharper, suggested by Adobe for PS, of pano's as well. I'll give the recommandations of your site a try!

Yikes!!!!

I was naive to think there was some magic number and sharpening method!!

I do use ACR for capture sharoening, I manually evaluate each image to determine the sharpening setting and use the mask, when appropriate.

Since all that was remaining was the output sharpening required due to the image downsizing I was hoping that one type of sharpening worked best for this type of sharpening....especially in view of the fact that it is for the web. But who though digital imaging should be so easy.I guess I am a hopeless romantic dreamer.

Thanks to all for your replies.

Up to now, I have been too busy to really sink my teeth in to any of the responses gathered thus far. I have had a chance to do so and have specific questions to those kind enough to have helped me thus far (thanks).

Merci Nicolas, I have not yet downloaded your script yet. What sharpening method does it use? USM, smart sharpen, high pass? I downloaded the sharpening scripts from the lightsright website but can't seem to get them to work. Hopefully I will be more successful with yours!

Giovani, thanks for your reply. I have read Bruce Fraser's sharpening book. My sole output source for digital shots is the web so his book didn't really help me much with output sharpening. I don't print. I was hoping for direction as to which sharpening technique would be best for web-based photos. I use the highpass method so maybe I am already using the best method. Thanks for the tip for setting the bicubic default. Not too sure if I'll go back to DPP, I HATE it's interface and crude controls. ACR is a pleasure to work with though it's colours are not as good as DPP - though pretty close now that I use presets in ACR.

Bart, your reply was also instructive. I hope you could explain why it is important to blur before sharpening after one has resized as this seems so counter-intuitive. Is this similar to the reason why one applies anti-noising (essentially blurring) prior to sharpening? Why the " x.25 " element in the equation to determine the amount of gaussian blur? Can you explain why you feel the smart sharpen method in PS is the best one to use after the blur has been performed?

Phil Marion said:

I hope you could explain why it is important to blur before sharpening after one has resized as this seems so counter-intuitive.

Click to expand...

No, you need to blur before downsizing. The reason is that the original data has, say, 3.19x3.19 times more pixels than the result requires. So, what happens to that detail when you replace every 3x3 pixels by a single one without a pre-blur (or a more elaborate 'low-pass filter')? I'll spare you the scientific expanation here (here's a free book about it), but my examples of the test target show what happens (you can try it yourself by downloading the full 1000x1000 pixel size (1294 KB) test target and resize it). By blurring the original size, you create a weighted average of all 3x3 pixels so the resampling will not have enough detail to cause aliasing artifacts. Aliasing is caused by (re)sampling more detail than can be represented in a single discrete sample of that detail, that's also the rationale behind the AA-filter used in DSLRs.

Why the " x.25 " element in the equation to determine the amount of gaussian blur?

Click to expand...

That is my empirical rule of thumb for Photoshop's Gaussian blur implementation. You can see how valid it is by creating a sharp B/W edge, and then apply such an amount (3504/1100 = 3.19 pixels, and 3.19 x 0.25 = 0.8) of Gaussian blur. Then count the number of pixels it takes to transition from black to white. In your scenario, it would produce something close to the required 3.19 pixels. When that edge is down sampled by a factor of 3.19 (to 31.3%), it will become something close to a 1 pixel transition. While that is smoother than the original 1/2 pixel transition, it also avoids stairstepping on diagonals or increased noise. A subsequent sharpening step will regain the intended sharpness in the result, but with fewer artifacts. My rule of thumb factor can be reduced to 0.2x the downsampling factor, which will obviously create a lower blur radius and requires less final sharpening, but it will be less effective in avoiding artifacts. It depends on the subject which setting produces the best trade-off.

Can you explain why you feel the smart sharpen method in PS is the best one to use after the blur has been performed?

Click to expand...

It uses a more advanced method of sharpening which leads to fewer edge halo artifacts than e.g. High Pass or USM sharpening. In it's advanced mode it can also be tweaked for even better results (but that's something for a less generic approach). Nicolas Claris' action also uses it, in addition to improving the overal tonality (local contrast), although his action is originally intended for full size images so it doesn't avoid down-sampling artifacts.

Bart

after using Bart's hints for downscaling big panoramas, about 15000 pix to 7000:

Thanks Bart, it works perfectly for avoiding downscaling artefacts. It takes a bit of time to do it automatically, but it's worth it.

Michael Fontana said:

after using Bart's hints for downscaling big panoramas, about 15000 pix to 7000:

Thanks Bart, it works perfectly for avoiding downscaling artefacts. It takes a bit of time to do it automatically, but it's worth it.

Click to expand...

Glad you found it to be of help. And I'm sure the subsequent sharpening is much faster on the reduced size version .

If one frequently has a large volume of images that need down-sampling, one can save a lot of time by installing ImageMagick and use a batch file to process entire (temporary) folders full of images. IM's down sampling routines produce the best down sampled output quality.

Bart

I've downloaded Nicolas' sharpening action and it works like a charm. It really improves local contrast as well.
One thing I've noticed about it - compared to other sharpening plugins I've used/created - is that the opacity slider isn't as responsive as them. Which is a good thing. I hate trying to gauge the optimum opacity percentage in the blending menu because many many times when I go back to a file the next day my eyes tell me I erred with either too much or too little sharpening. Maybe I need glasses.
Anyhow, I notice that more often than not the 30% opacity level that Nicolas' plugin defaults is pretty darn close to the optimum level (as far as my eyes can tell).

Do you find you need to tweak the blend opacity level much?

P.S. thanks for the sharpening action!!

Hi Phil

I'm glad you like and enjoy the action!

In fact the opacity of the final layer may vary depending what previous sharpening you have applied or not during the RAW processing.

It also depends, for users of Lightroom or ACR if they used the "clarity" function.
My sharpening action as quite the same effect (in the midtone contrats) and when one is added to the other, the result may become to much contrasty…

The level also needs to be adjusted depending of the resolution, the less resolution: the lower % of sharpening.

I noticed that this action works very well also on the old film scans;-)

Nicolas Claris said:

Hi Phil

It also depends, for users of Lightroom or ACR if they used the "clarity" function.
My sharpening action as quite the same effect (in the midtone contrats) and when one is added to the other, the result may become to much contrasty…

Click to expand...

I use ACR to convert my RAWs. Do you recommend that I NOT use the Clarity function if I will be using your plug-in to sharpen for output? Or is it better to use the Clarity slider at a low level?

Phil Marion said:

I use ACR to convert my RAWs. Do you recommend that I NOT use the Clarity function if I will be using your plug-in to sharpen for output? Or is it better to use the Clarity slider at a low level?

Click to expand...

Bonsoir Phil

no recommendation!
However, if you feel that clarity is needed, use it, maybe at a slower rate as you would do and be prepared to lower the layer opacity of the sharpening action.
There is no truth about sharpening, do what you feel works best for you, after all you may do some trials with different sharpening combos…
On my own, I mainly use C1 as raw converter, but, when it happens me to use LR or ACR, I never use the Clarity slider above 15 or 20%.

Phil Marion said:

I use ACR to convert my RAWs. Do you recommend that I NOT use the Clarity function if I will be using your plug-in to sharpen for output? Or is it better to use the Clarity slider at a low level?

Click to expand...

Clarity is simply a Midtone contrast tweak.

Andrew Rodney said:

Clarity is simply a Midtone contrast tweak.

Click to expand...

Absolutely and when my sharpening action is used over Clarity, midtones are aver contrasted…

Nicolas Claris said:

Absolutely and when my sharpening action is used over Clarity, midtones are aver contrasted…

Click to expand...

Well I can't comment on YOUR sharpening action. Its not (I would suggest) shouldn’t be an issue. Clarity is a tone rendering adjustment. Sharpening should adjust edges to produce the appearance of sharpening. Now I'll admit that one has to have some understanding of tone adjustments with sharpening such as not pinning subtle and reproducible highlight detail too close to clipping because sharpening may, may push them to clip. But I don't think anyone should be worrying too much about using tone and contrast controls for rendering their images with respect to what the sharpening might do later in the process. That is, IF clarity looks good to the current rendering of your Raw, I'd hate to have to back it off because some sharpening routine later might not 'like' these settings.

Bonsoir Rodney
this action has been written long time before "CLarity" came-up in both LR and ACR…
It performs different levels of USM and accentuation on separated highlights/midtones/darktones areas.

It is ended with an adjustable layer, set by default to 30% opacity so one can adjust this to taste.

As already said above, I just don't use Clarity, as I use C1 (which doesn't have the clarity function)…

For those using ACR or LR they can use a bit of clarity if they wish, but it is not necessary if they perform the sharpening action…

Kind regards

All I can tell you is that Clarity is a great and useful rendering tweak, like many found in LR and CR. The sharpening routine I use doesn’t care if its used or not. It doesn't treat rendering of images this or that way, it just sharpens (no top) of what I feed it.

Yes, Clarity

is a great and useful rendering tweak

Click to expand...

but as all tools it must be carefully used, and checked viewing at 100% in the raw processor, as it very easily produces halos when overused (as many contrasting and/or sharpening techniques)…

I do prefer to adjust midtone contrast and sharpening, after all WB/color/resize process just before converting to output space and switching from 16 to 8 bits…

Yes I'd agree you can go overboard with clarity and make an ugly image. But you can do that with curves, Vibrance and just about any rendering setting provided.

Yes , of course, but Clarity seems so easy… just a slider to push and viewed at somewhat 25% the image looks nice… then when printed…

Rodney, I'd be interested in your input to my query posted in the original post.

Thanks for your help.

Phil Marion said:

My question is HOW to best sharpen the image after I have downsized it to a 1100 pixel length JPEG? Simple USM? Highpass sharpening? Smart sharpening?

Click to expand...

This being the original question?

Capture sharpen the master high rez file. You can down sample and sharpen for output based on that size and the output device. How? I use PhotoKit Sharpener. I'm biased, but I'm lazy too. Bruce did all the hard work of figuring out the appropriate values. I pick what sharpening I need, hit the OK button and move on.

Andrew Rodney said:

This being the original question?

Capture sharpen the master high rez file. You can down sample and sharpen for output based on that size and the output device. How? I use PhotoKit Sharpener. I'm biased, but I'm lazy too. Bruce did all the hard work of figuring out the appropriate values. I pick what sharpening I need, hit the OK button and move on.

Click to expand...

Thanks for your reply (though it didn't help me). I've read Bruce's sharpening book, and I DO use ACR for capture sharpening. I love the mask creation. I am also familiar with the PK plug-in and its ease of use (and the fact you are friends with the creators of the plug-in) but since I don't print my shots I see little value in spending $100 for a plug-in that is great for those who print on a variety of papers and printers. I too am lazy and was hoping someone could tell me the 'magic numbers' for output sharpening on my NEC 20WMGX2.
I've done some testing.
Nicolas' plug in works best on files where I don't use any 'Clarity' adjustments in ACR. Sadly ALL of my processed RAWs have had some 'Clarity' adjustment done. So using his plugin would require undoing the Clarity on my thousands of shots. Too much work. Going forward I will use his plugin and lay off the 'Clarity' tool. But I have thousands of past photos that could use sharpening improvements.
I find that the best results I get so far on those shots where I've performed a 'clarity' adjustment is with using High Pass sharpening. It is better than Unsharp Mask. I guess I should try to master the Smart Sharpen technique and compare but...as I said, I am lazy.

Here's hoping some other big-brained digital photographer with a big heart can save this digital noob a lot of time and trouble and suggest an optimum sharpening technique for photos that do not go to print but are only used for viewing online.

Bonsoir Phil

Could you give a try on your "clarityfied" file with this action and tell us if it works better than the previous…

Sharpness_Nicolas_Claris_Clarity

Thanks for all!

Andrew Rodney said:

Yes I'd agree you can go overboard with clarity and make an ugly image. But you can do that with curves, Vibrance and just about any rendering setting provided.

Click to expand...

Hiyas,

Jumping in, perhaps only testing waters with the toes, but... isn't "Clarity" a sigmoidal contrast algo? Adobe is scant on the "formulae" for its functions(to wit, USM has a "black&white" kernel, and the "blur" is applied uniformly across the entire image data, IIRC). Therefore, "Clarity" only alters contrast, an item often mistaken for "sharpness", and having nothing of the USM effects... thus, I suspect should be applied /before/ USM.

rgds,
Dave

Giovanni Brembati said:

Hi Phil,
the so called "capture" sharpening is to counteract the softness from the interpolation of the captured colors to produce RGB value for each pixel
If you shot in raw this step must be executed in the developping process. The strength depends on the interpolation quality.

Click to expand...

My understanding is that capture sharpening is needed to counteract the results of the AA filter, not the bayer de-mosaicing.

I also suspect that capture sharpening can also effectively be done after the raw conversion (assuming the 16 bits have been preserved.)