Sharpness or not sharpness
Canaletto’s painting is a beautiful illustration of what I love about modern photography: seeing what my eye didn’t see when shooting.
Move from global to local. On the National Gallery website, I can get closer and discover all the details and the story the painter tells me. I can also appreciate the big picture by taking a step back. When I stitch together short focal length photos to build a large panorama, the resulting pyramid image allows me to wander inside to discover details and stories I never suspected.
Digital image processing reveals fine details that would not have been read without it. A few years ago I made postcards that I sent to friends for the pleasure of sharing photos and news. One of them was a photo of dunes (skyless) at sunrise. After too harsh HDR processing, what was my surprise to discover in this landscape a detail that I had not seen during the shooting. A double-tire track left by a human being in search of thrills (dune bashing). Often my friends didn’t see it, but I knew it was there.
All that to say that I like to find details in my photos and to be able to zoom in on very big photos.
I’d prefer rather resolution and processing to reveal the details than sharpness even if the last Aurélien module brings interesting additional information to the initial image.
Sometimes, I order A0 print too…
That is photo-illustration territory, similar to that of ads for new urban developments, comics or Jehovah Witnesses pamphlets. More cartoon than sharp.
I don’t fully agree with this statement,
people generally overdone with too much shadow/highlights/tone equalizer and local contrast enhancements.
Local adjustment are more prone to give an artificial look to the image, sharpen itself is the less offending in this category
For me, a sharp image is the one that contains the maximum amount of information (on the area where it is sharp).
No software, no algorithm can reconstitute information not recorded at the time of shooting.
So-called marketing thinkers sell (expensive) their soup by claiming they
can save rotten photos with their miraculous recipe of “sharpness improvement”.
Looking at it from a very far distance, a photo with very exaggerated local contrast can produce an impression of sharpness.
Some photographers(?) have abused this effect, like in the 19th century when they hung “cute” things around baby pictures, and now do exactly the same with Instagram.
Sure, but the information can be recorded but hidden, and needs software (algorithms) to show them. Google hubble deconvolution for example…
My creative/artistic answer:
The human visual system/HVS is very sensitive to “edges”.
Just like lenses with shallow depth-of-field can direct the viewer, local-contrast and sharpness can (as can blurring and negative-clarity make backgrounds less distracting). Shallow DoF is not what the HVS sees, yet we have learned to use the technique and even judge the quality of it (Bokeh).
Yes, these techniques (DoF, Sharpness, Blurring, etc.) are secondary to composition, lighting, production design and art direction, but I fail to see how they are not just tools to achieve something. As with every tool: if you use it wrong, you can break stuff. Since we are super sensitive to edges and how they look real to us, it is also super easy to break things. So if in doubt, don’t use it. If something looks off, dial it down.
My technical aspects answer:
When you know that your optical systems transfer function is a convolution of a-form-of-top-hat-function with real detail then a deconvolution algo can at least attempt a reconstruction if you can characterise or guess your form-of-top-hat-function. When where and how that fails is another question, but it’s not ludicrous to attempt. And as @heckflosse pointed out, it sure can work within certain limits.
I see USM-sharpening as a blind, way too simple attempt of mimicking a true deconvolution. It fails quickly and with horrible artifacts. Better sharpening algos exist.
Interesting discussion. I have definitely stopped caring as much about sharpening as my photography skills have improved and as I’ve realised that composition is far more important for an interesting photo. I think what a lot of people are chasing is actually local contrast rather than sharper fine details, although of course birders and macro photographers are probably looking for that amazing detail…
As a general rule, I’m always suspicious about assertions of what can be creative and what can’t. Surely there are no limits to how creative we can be and what tools can be used to achieve this creativity? Sorry, but this statement stands alongside your older bizarre assertion that we should be treating clouds as background elements and not be concerned about details in clouds. Why not if that’s what you want to do? I personally don’t use sharpening as a creative tool, but if someone wants to and likes what they create, more power to them.
I don’t know why sharpness per se would be so sought after. Personally I find images that are interesting and I’ve captured them without blur, properly exposed and focused, look “right”, here the gear can make a big difference. When properly used, a sharp lens like the 50mm Summilux or the like, produce wonderful images because they capture the image with fidelity.
I’m not very good at post-processing to “correct” flaws in the image. I will unabashedly use filters or manipulation to obtain what I think is an interesting, eye catching or beautiful, creative image.
However I’ve never had any luck with pp “sharpening” or fixing in any way the out of focus or overly or under exposed shots I took. In some cases I find “soft” images to be quite good and I keep them. Having gone “over the edge” long ago and purchased far too expensive (for my wallet) lenses and cameras, I’d not spend another dime of any significant time trying to “sharpen” blurry shots. You get what you shoot and sometimes those “blurry” images are the best.
Indeed. I am mostly shooting wildlife and macro, so fine details tend to be important to me. A lot of people who won’t give a thought when they see a common dove or sparrow in their yards are surprised by their subtle color and detail when they see these nondescript birds on display.
Last summer I was lucky to have this hummingbird tolerate my presence at minimum focal range with my 150-600mm zoom and the lighting was bright but diffuse so I could capture the feather’s iridescence without overwhelming the structure. This is probably as optimal conditions that I could wish for.
This first shot is a complete edit without any sharpening:
And this is the same edit with Diffuse and Sharpen, using the unmodified Sharpen Demosaic (AA Filter):
I was very judicious with noise reduction to preserve as much native detail as possible, but I believe the sharpened image is improved, particularly looking at the detail of the feathers and the eye.
So in my mind, I don’t give a hoot about sharpness, but detail matters a lot.
PS - I do shoot lots of other critters than hummingbirds 
Lovely photo @Dave22152. The second one is certainly sharper, but honestly, I’d still be very happy with the unsharpened one. Sometimes I think sharpness or lack of it only becomes apparent when we have a comparison to make.
I love hummingbirds. I should try to capture one here in BC but I’m waiting for the Fujinon XF 70-300 to come back in stock, which looks like never!
It is first of all a stylistic choice. Commercial pictures (e.g. for DaVinci Resolve) are more sharpend for obvious reasons.
I think the best way to use (post) sharpness is using it only locally, to help focus.
Another reason for sharpening might be (web) compression.
The problem with sharpening is that it adds something that is not there. If you take a photo of a blurry photo, software can make yours look “sharper” than the original, creating details that never was.
Having said that, I want to add that Aurélien your contributions in this field are appreciated
Why sharpness?
Because most images are viewed on displays with 1920x1080 pixels or slightly more or less. Then sharpness is more about perception of details.
A print doesn’t need the same sharpness to look fine, since just very few people looks at prints on a detail level. If you’re sitting about 40-60cm in front of a screen thats a different thing - you get sharp text, ui elements and that impacts the perception.
So i agree, if your intention is to create pictures for prints, you don’t need to bother about sharpness except for having the subject in focus.
@anon41087856 has provoked great thoughts in my image processing. He has taught me that depending upon the histogram is like painting by numbers. He has highlighted the trend to seek ultimate sharpness as a technical pursuit rather than an artistic achievement. But he has also taught me to respect the look. I listen to Aurelien and I consider and respect his views, but I will always go for the look that satisfies me. For me the correct amount of sharpness is about the correct amount of detail. A portrait generally requires softer skin while a landscape a little bit more sharpness in the foreground. I have regularly judged photographic competitions for a calendar. In my experience 25% of the images are oversharpened to the point that we exclude them, 50% of the images we say would be improved with a little more sharpening, but we do not exclude them. The remaining 25% have a nice level of sharpness applied. When I teach my photography students about sharpening, I teach them my rule of halves. By that I mean adjust the sharpness level to what we like and then only apply half that amount because many of us tend to be too heavy handed.
BTW, I love the new AA sharpening preset that Aurelien has created in his new diffuse or sharpen module. We are so lucky to have a developer like Aurelien working on Darktable to make it distinct from the commercial image processing products out there chasing our dollars.
I guess I don’t feel the goals of a digital image to necessarily be the same as a printed photo or a classical painting, and it’s best to give the artist a best-in-class tool kit and leave creative choices to the artist.
My own viewpoint is that a digital photo only lives in the same family tree as a printed photo because it can be printed. Otherwise they are notably different animals. Once you have a print, you have created a thing to be enjoyed at the scale it has been created. However a digital image is a more liquid medium, certainly in terms of viewing.
A digital photo can be enjoyed at many scales, and technology makes this extremely easy. Just zoom in and out. I think it is a very reasonable statement to say that when you are zoomed into parts of a digital image, the areas that are supposed to have high detail in fact have high quality detail. I’m not saying that sharpness and detail are the same thing- I would say that detail is a property of the composition, and sharpness is about resolving that detail. When I look at the image zoomed out, or after it has been rendered as a printed picture, I expect to lose at least some of that sharpness. Certainly one can point out that a grossly over-sharpened image looks creepy even when “zoomed out”, but to me that’s not a great argument for never using sharpening. That would be throwing the baby out with the bath water.
Perhaps it is an obtuse statement to make, but I think I’ve overused sharpness if it looks over-sharpened. It has a “look”. I think the key is in training our eye/mind to recognize and avoid it, and this comes with practice and experience. A little bit bit can go along way, but every composition is different.
To speak purely about merit as a technical tool, high quality sharpening is helpful, if for no other reason, to allow creative cropping to be used while giving up as little detail as possible as you effectively create a “forced zoom” into areas of the image via the cropping operation. Again, obviously an over-sharpened image won’t look good zoomed in either, but that is not my point.
Sharpness is a function of contrast.
Once you grok that, this whole thread - including the initial post - becomes mostly moot and everyone would be talking about conceptual and creative usage of contrast. Most of that would be through lighting and gradients and falloff and edges and whatnot.
Still, the algorithms to detoriate the image from the original RAW file even more are needed. Someone having spent the better part of the last years on exactly those algorithms should know better. Photography is the transformation of light into something else. It is never about reality but only about the remembrance of a possible perception of a moment in space and time. Another hard concept to grok.
This does not necessarily have to be the case. When you lose sharpness, you first lose the contrast at the edges before you lose detail. Thus, if you restore the contrast on the edges that must not mean that you add something that was not there.
@Thomas_Do you’re not adding the same thing that was lost though - only something that attempts to reconstruct the appearance of it — which is fine but it’s no the same thing.