Why sharpness ?

Processing
1

This is where I tell the story of my life, so don’t expect information.

I have been working on deblurring algorithms for 4 years now. I tried deconvolution, the blind and the non blind kind, unsharp masking with fancy guided filters, and before that the usual high-pass (actually an unsharp mask with gaussian blur) then wavelets contrast boosting. The latest has been a diffusing process and you will get that in darktable as the diffuse and sharpen module.

And yet, over the years I have gradually stopped using sharpening methods.

Why ?

I have imprinted in my brain the classical digital look. It was all the rage in the late 2000’s and some people have not evolved from it : saturation + 100%, contrast + 200%, sharpness + 500%. That makes for the caricature of an image, not for an image. All you see in there are the effects, they hide the content and the intent. That’s bad. It’s like a young musician than tries to play loud and fast to show off, and only produces noise in place of music. No content, no intent, only effects.

Perhaps, having 12/10 on each eye makes sharpness not that appealing to me, it’s just business as usual. Or perhaps it is that, having discovered analog film and its subtle softness, I find it more expressive because it’s less realistic. Less surgical. Or even that, with a 4 K screen, I have practically disabled any sub-pixel anti-aliasing in the first place, so that display doesn’t lie to me anymore.

Cartier-Bresson himself said “sharpness is a bourgeois concept”, and he knew a couple of things about the bourgeoisie, being born in industrial upper-class. There is something we need to ask ourselves about sharpness : what does it means ? What does it convey ?

And the truth is it’s meaningless. It’s a technical quality. It is neither beautiful nor ugly, it’s just optics. It reproduces reality 1:1. Yes. And then you find yourself having more skin retouch job to hide all those blemishes nobody saw 30 years ago. Honestly, who gives a flying shit about reality ? We already live in it, isn’t that enough ? Who said that reality had to look real anyway ?

Or perhaps it’s hubris. Sharp lenses are expensive. Producing sharp pictures makes you look like someone who owns expensive gear. Again, no meaning, only metadata.

But my real problem is sharpening looks bad. Whatever method you choose, as soon as you want it to be noticeable, it’s almost impossible to get organic and good-looking results. You will, again, get a caricature : fringes, reversed gradients, non-blended transitions, chromatic aberrations. The truth is sharpening sucks, because there is no really good and clean method to do it, and the ones that come close… well, you computer is not ready for them, but they will sure keep your room warm in winter. Yet it seems that people don’t see it. Whenever I see someone posting a “pleasingly sharp” image, I see edge issues and fake sharpness, and seem to be the only one.

The only real issue with lack of sharpness is with focusing mistakes. Front and back focus are disturbing since your subject is actually not advertised as a subject anymore. So refocusing is a real need, but it won’t be a thing until high-end GPU become “affordable” again because it’s once again expensive and still not perfect.

But for everything else, smoothness is more important than sharpness. Reality is not a line drawing either, objects blend smoothly into each other. All the more since the atmosphere diffuses light a bit.

And it’s a funny thing that painters would spend a significant amount of time to render blurs, hazing and sfumato where photographers spend time to suppress and remove them.

So, why is it that sharpness is one of the most sought-after and discussed property of an image ? What do people see in it that I completely stopped caring about ? Or is it that people care about it because they feel like others do, so they should as well ?

33 Likes
2

I personally don’t use any additional sharpening at all; it’s not even available in Filmulator yet (though the tonemapping does add a little edge enhancement).

That said, I do have very sharp lenses so the image isn’t flawed from the start, letting me not have to sharpen noise or chromatic aberration.

For me, as long as the softness isn’t a noticeable flaw, I don’t care about the difference between very sharp and razor sharp.

2 Likes
3

For me, the answer is that my eyes are not blurry, so I don’t want my images to be. If people have blurry eyes, they get glasses. If people have blurry images, they get sharpening algos… or better lenses.

However like you I think being undersharp is better than being oversharp, and many of my favourite photographic images - autochrome - are soft. Therefore when I use it, I try to keep it subtle. The threshold mask in dt really helps a lot with this. It also gives a more artistic sharpening, with broader details remaining blurred. This leads me to:

Artists are in the position where they can control the blur in whichever fashion they choose, using it as a technique to draw attention to certain areas, and away from others. This is not just a technical quality, but an artistic one. I like to think of sharpness and blur in the same way.

1 Like
4

I agree with @Soupy. I would add that some digital cameras (like mine) have a rather heavy low-pass filter in front of the sensor. Some amount of sharpening is often necessary.

I agree that sharpening is one of the most abused processing tools today. But, just like people these days don’t seem to oversaturate as much as they did a few years ago, maybe in the future we’ll see more natural, carefully sharpened images too.

1 Like
5

My feeling is that what we really want but often have no idea how to describe, let alone quantify, is corporeality; a convincing sense of light playing on real, solid objects with tactile surfaces and 3-dimensional textures. Of the many factors that contribute to this elusive quality “sharpness”is the easiest to recognize and quantify. It’s deeply unfortunate that in everything from lens and sensor design to post-processing software design/usage the obsession with sharpness so often leads further away from the more subtle hope for depth and richness.

7 Likes
6

The smartphone jpegs are the worst offenders… between the aggressive sharpening and tone mapping they do ensure that every little detail everywhere is clearly visible, but sometimes the overall image ends up not looking very realistic. The raws do need at least a little bit of sharpening though.

1 Like
7

@NateWeatherly Your comment was what I was trying to express in my presentation at the live meeting.

One needs to grasp the contents of the image well in order to capture it properly within the frame. Technology may help gather and reassemble the missing pieces. But we need to know what the actual problem and subjects are first.

What must accompany the Why.

8

Surely when we talk about sharpness, shouldn’t it be about fine detail?
Fine detail is the result of many things, starting with demosaicing.

9

You are quite right about Cartier-Bresson’s abhorrence of the preoccupation with sharpness; throughout his life, his concern was always with the quality of a photograph as a picture, and its mere technical qualities did not interest him. As far as he was concerned, if the photograph was a poor picture, no amount of technical excellence should rescue it from the bin.

This argument about the importance of technical excellence (particularly ‘sharpness’) versus ‘artistic’ merit has persisted since the earliest days of photography in the 1840s, when photography was an expensive persuit involving considerable expertise in chemistry and optics. For some, photography was almost just a record of experiments in optics and chemistry; whereas others were intent on making photographs emulate paintings.

The tension between ‘scientists’ and ‘artists’ underlay, for example, the scism in the Photographic Society of London (later to become The Royal Photographic Society) that led a large group of the ‘artist’ faction to break away and form “The Brotherhood of the Linked Ring” in the 1890s, which was mirrored in the USA by Alfred Stieglitz’s “Photo-Secession” movement.

Is it too much of a stretch to say that the preoccupation with ‘sharpness’ (and other technical and scientific aspects of photography) is one side of a continuing tension between photographs as pictures and photographs as a manifestation of scientific principles?
:wink:

3 Likes
10

1, I don’t surely know either, maybe because we all look first “is the photo in/out of focus” and assume “sharp is good”. So a trained perception?
2. Again trained perception from “good jpegs”?
3. That would be “deeply human”. We can only emphasize: get your personal standpoint on that.

Very few photographers are artists i would say. Every child paints - every child sings - everyone does photos. Only a few children get musicians or painters. Why should this ratio be different with taking photos?

I am definitely not an artist, i like to shoot photos, maybe 10% are worth to be shown to others but art? No.

4 Likes
11

A very interesting contribution. I can already say that.
But I have to think about it quite seriously. Obviously, I have lost my way there. A maximum sharpening to be achieved, which should remain undetected if possible, i.e. should not be detected as artificially unmasked. And why? To whom did I want to prove what? Probably that I am better than my camera with the small sensor? I will deeply question that.
Now already many thanks to @anon41087856
Micha

12

Sharpening, sharpness and softness are just tools. Tools to be used with intent at the curating or shooting moment. I think that goes without saying. Those who don’t use their tools with intent should stop consider and look (litterally ) at what they are doing.

I like the Bresson quote mostly because it gives the property owning class a boot. Something that doesn’t happen frequently enough today.

Different types of photography require different tools. I’m both Düsseldorf and are bure boke

2 Likes
13

Sharpness and contrast just look good to me, when you try some last gen high-end lenses the pictures have some magic to them, they are sharp and smooth at the same time if that makes sense. That said I agree that softwares don’t seem to do the trick (it’s just a hard/impossible problem), although I haven’t tried some of the new ML-based ones.

On the creativity side it’s much easier to blur a sharp image (or produce a blurry image with a sharp lens) than sharpen a blurry image.

14

In one of @anon41087856’s presentations, he introduced me to the idea of processing photos in a far-zoomed-out view. Trying this I noticed I started caring more about composition and lighting and colors; less on technical merits such as sharpness.

Around the same time I noticed that some of my all-time-favorite photos are woefully blurry. Of course I can’t know whether I would have liked them better if I had nailed focus and shutter speed. But… I think perhaps some of them are actually better for their imperfections.

Similarly, I never once was dissatisfied with the sharpness of my film photographs, even though they are clearly less sharp than my digital files.

I also noticed that I stopped caring about anything pixel-level when I updated from 12/16 MP to 24 MP. The only thing single pixels show is noise! Come to think of it, that is why I vastly prefer Darktable’s richly versatile contrast equalizer over Capture One’s restrictive sharpness/texture/clarity sliders.

I think my post production got better by caring more about the big picture than the pixels. At least I am happier, so it’s a win either way. I wonder how many perfectly usable photos I culled for some irrelevant technical imperfection…

4 Likes
15

One does not have to be a good artist, in order to be called an artist.
I hear this argument all the time, but it never washes. If that were so, most music in the top 100 could not be said to be recorded by artists, and many works of modern art could not be said to be art :wink:
So, all photographers are artists in my book, whether purposefully or not. It just so happens that some produce better art than others.
However, I think this opens up a debate unintended from the original post, so perhaps its best to leave it be.

2 Likes
16

For me, this . More often then I like to admit :slight_smile: .

Be it older glass (with auto focus issues ), older cheap cameras where the focus array is a bit misaligned, etc… I often mis focus in my shots by a hair.

‘sharpening’ is then a tool to try to fix issues . You see what you get , and you do an A-B (better/worse/better/worse) and I’ll take what i can get.

But if I do get good focus (more often then not on my cheaper CDAF m4/3 camera or my analog gear ) i don’t need to do anything.

I do think that there is this idea that an aliasing filter adds some blur , and they try to ‘remove’ that blur to get back to what in your head sounds more like a ‘ground truth’, how it should be. I start with something , something adds blur, I want to go back to before they blur . I understand the rational, if i care a lot is something else.

Another thing can be if you see very small details in your original high res image that you like and want to be noticeable in a downscaled version. You use some sort or sharpening to let details be visible they you noticed and liked in your original version. Again, i understand the rational. And I do use something like this most of the time.

I was also quite surprised with the amount of details still (clearly) noticeable in simple 35mm film negatives. No it’s not as contrasty as an high res digital shot, or they don’t stand out they much … but they are there.

Single hairs for example . Everytime I have a digital picture with an analog equivalent , I see some details in the digital one that are almost 1px wide and I think ‘no way this would be visible on the analog shot’. And then I look… Yes, there it is. It jumps out less, but the same detail is visible. And it actually ‘sits’ better in the surroundings often.

(Frizzy hair with a sunny backlight for example, the way the sunlight merges with strands of hair just looks fine straight out of the box from analog, where the digital shot would present something like 'wow look at that hair '. It all depends what you want as an end result )

17

Very good discussion…

I never do sharpening on full-sized out-of-the-camera images. Messed with it a few times, but never saw a marked improvement over the selection of a decent demosaic algorithm.

I do always add a bit of sharpen to downsized renditions. Once the image is crunched down to a size viewable on monitors, there’s a definite benefit to increasing the acutance (illusion of sharpness) in such renders. With a convolution kernel, I find 0.5 to do the trick.

I’ve not done any serious printing yet, but I may play with sharpening there in determining how to best render for that medium, along the same lines as for acutance…

2 Likes
18

That’s an interesting point. I have to ask if we should make a distinction between sharpness and detail, where the former is an attribute alongside contrast, saturation, etc. and the latter being the information we’re trying to convey. You might need some level of sharpening to bring out that detail, but push too hard and you’re introducing false information that detracts from the photo.

Some forms of photography like sports, wildlife and macro benefit from a level of detail that might be actually be distracting in other forms, so the trick is to apply the right amount of sharpening without over doing it.

I think where people get in trouble is pushing sharpness as an attribute beyond the level of information than the image can support, or trying to compensate for poor focus, motion blur or poor lighting. In that case the image begins to suffer from artifacts or looks artifical and over processes

19

For me I think its a perceptual thing…I never question these things on print…if it’s blurred or crafted to be hazy or foggy it seems far more acceptable on first sight than things that are fuzzy or blurred on a screen. I still think also for me that screens don’t do a very good job of conveying depth and texture in a perceptual way that occurs when things come out as a painted work or printed on paper…

1 Like
20

Sorry, I do not know how to be brief. Anyone who’s been in a live meeting where I’ve got to talking probably knows that though. :slight_smile:

Back in the manual focus days there was some skill involved in getting a sharp image (by that I mean in-focus) but today unless you have faulty equipment it’s really not that impressive. There was also the fact that most of those lenses we designed/built/traced without the aide of computers. Sample variation was relatively large so getting a sharp piece of glass was as much luck of the draw as it was showing off that you had fat stacks of cash. I remember doing optical physics by hand back (as an exercise) back in university and the fact that compound lenses were able to be mass produced with any measure of quality before computer aided design was widespread was really a feat of craftsmanship.

“Back in the day” stories aside this is like any other gear wankerteering hobby or profession out there and people are looking for ways to say “mine’s better.” Sharpness is a relatively easy concept and visual indicator of “quality” for the 200lb gorrillas operating the cameras to get excited about. Monkey do like shiny. I agree with @hannoschwalm in that most photographers are far from artists and there is a huge amount of gear cultism going on. I think with modern imaging systems though sharpness is a base expectation. I haven’t really gotten excited about it in more than a decade as it’s an expected feature of any lens or camera in the modern era like autofocus or auto bracketing. I dunno, maybe the trends will reverse and someone besides LensBaby will introduce a blurry lens again.

In my own I really only use sharpening in post processing for one of two reasons. Firstly, I have a couple of older cameras with AA filters where some sharpening is recommended. Secondly, overcoming atmospheric effects in some scenic landscape photos or astro photos. For the latter I really just use some minor local contrast adjustment as a sharpening technique. Otherwise I don’t really worry about it.

All that being said I treat sharpness like any other technicality of a photo. You can’t sharpen or color grade your way out of a bad composition or emotionally unconnected photo. Those are the icing on the cake IMO. A boring photo is a boring photo no matter how sharp or how hard you apply the blue/orange look present too it. Maybe you don’t even want sharpness and it’s counterproductive to the theme you’re trying to show. Instead of spending so much time in the digital darkroom playing with settings or looking at the B&H catalog go out and try to get good or maybe find a different thing to spend your time on. Personally my rule is a I shoot JPG+RAW these days and if I’m not OK showing the JPG SOOC to someone I don’t bother editing the RAW. There’s usually some fundamental flaw in the photo at that point. It’s garbage and goes in the bin. I’ve got too many TB of data to wrangle already without worry about keeping sub-standard crap on my drives and paying to have it backed up, etc.

I think I’ve become more harsh in recent years though. :slight_smile:

Don’t get me wrong, tools are neat and I do enjoy a well designed camera that’s enjoyable to use. I used to be (still am to some degree) an auto mechanic and a good well thought out tool can make a lot of difference in how your day goes. But good tools in hands of someone with no idea or vision is useless.

As an aside what’s 12/10 vision? Here in the US we use 20/X, my eyes are 20/17. The larger the denominator the worse your vision. In my case it means I can read from from 20 feet what most people need to be 17 feet away from to see. If you have 20/30 vision it means you need to be 20 feet from a chart to read what most people can read at 30 feet away. I’m assuming your number works the same sort of way just on a different scale? Since you’re in Europe I’m guessing that’s meters and your vision is good enough to read a chart at 12 meters what most people can see at 10 meters? Do I have that right?

1 Like