« Car Talk: New Camera, Old Lens | Main | Pining for Gilda »

Wednesday, 13 November 2013

Comments

..."we're just fumbling about, flailing at conceptual will-o'-the-wisps while we try to wrap our heads and aesthetics around aspects of image quality we barely understand."

Time to find the Noam Chomsky of image quality to set us on the right conceptual course... :)

You've hit the wobbly nail on the head, Ctein.

Re: Your Example 2:

So I am not crazy or imagining things after all! For as long as I have owned it, I've been amazed at how Contax's 55mm/F3.5 lens consistently renders photos with a noticeably greater depth of field than either the numbers suggest is possible or by comparison to the performance of similar focal length lenses. Not only that, but it transitions from sharply focused to out of focus more smoothly than almost every other lens I own, both ahead of and behind the main subject, which is uncommon in my (admittedly limited) experience.

And now I know why all this is so ... thanks, Ctein!

As for developing a vocabulary to describe the optical performance of lenses and other photographic gear, be careful what you wish for, as high-end audio made a similar transition from "objective" measurement-based reviewing to "subjective" impression-based reviewing forty-ish years ago and look where it is today! 8^0

Would it make sense to try an empirical approach, having people "rank" the bokeh, say, of a series of images, then trying to determine a correlation between those rankings and the measured power spectrum? That's a crude description, and I'm sure the concept could be refined considerably. But some controlled means of objectifying the subjective would at least be a step in the right direction.

Ctien, I think you are on to something here. It is also the case that the tools, let's say a lens that worked well under its analogue constraints, don't work in an equivalent way on digital. Also, once there were a variety of films to choose from and a variety of papers -- developers too, but a relatively small number of variables. Also most of those components had a use-by date so you were carried along with materials innovations as papers, films and developers went in and out of fashion.

In the new digital world you have multiple generations of imaging chips with different processing algorithms, legacy lenses and ones "designed for digital," processed on different versions of multiple imaging software packages and edited on monitors for which there is no absolute standard. As hard as it was to master the old film-and-paper world, I think that the amount of potential variation in this evolving technology is quite bewildering. A vocabulary for discussing images made in this highly idiosyncratic manner would be most welcome.

Example 1 - different film formats.
Geoffrey Crawley came to a very similar conclusion to Ctein and Hicks. Also, that as the film format became larger (35mm to medium format to large format) it was the tonality that improved together with grain reduction for the same film speed/developer combination. Crawley formulated many good 'acutance' developers, which has an interesting affect on grain/sharpness/tonality.

One possible answer is to sidestep technical considerations and confine the concept of image quality to the realm of subjectivity. A picture is a good one if it pleases you to behold it. This has the virtue of overriding any conceptual hurdles just by exploding them. I’ve often wondered what makes a picture a good one – bar the subject and the inherent interestingness of the photograph – and every time I came to no conclusion. Sharpness doesn’t define image quality; if it did we’d never even look at some historical pictures. Derrière la Gare de St. Lazare has motion blur and Robert Capa’s most celebrated war photographs would have to be considered mediocre at best. What about resolution? In my view it most definitely doesn’t define image quality. A portrait isn’t better just because you can count the pores on the face of the portrayed person.
Ultimately the debate on image quality should be restricted to equipment. We can see the repercussions of that quality on the image, but that elusive concept of “image quality” should be left to our senses. One can measure a lens' quality through some parameters and express them on graphics and charts, but you can’t measure visual pleasure. That remains firmly in the field of sensations. You can’t put sensations into charts. There’s no measurement for pleasure and bliss. All we can do is confront the intensity of different sensations, but even that can’t be subject to exact measurements.
Of course, it would be quite comfortable if we were able to reduce image quality to a single word, a concept that would allow us to assert image A has better quality than image B. It just can’t be done, for we’re in the realm of subjectivity here. Some, however, have very firm beliefs about this: for them it all comes down to sharpness or resolution, but these qualities are just contributions to what we perceive as the overall quality of an image. Some folk feel no embarrassment to publish pictures of their cats on the internet – and I’ve even chanced to stumble upon several pictures of a cork board at DPReview! – to show the world what their cameras can do, but can we seriously say those are quality pictures?
Maybe we should just accept that “image quality” is a subjective judgement, and keep it at that.

Dear JG,

Boy, I sure do hate to look a gift compliment in the mouth, but I gotta tell you that, no, you do not now “...know why all this is so...”

I have presented a very tentative hypothesis, with no data or tests to back it up, just some casual observations and intuition. That isn't knowledge: that's a notion which might someday turn into knowledge. You've said that your lens has characteristics that would support that hypothesis. That's one possibledata point. That's not anything close to proof of the hypothesis or even that that is actually what's going on with your lens.

It's true that my notions give you an idea of what MIGHT be going on with your lens, which is surely a big improvement over having no idea at all. I'm happy to take credit for that. But having given you the answer? Oh, if only. I wish.

As for your other point, I'm not going to get into the audio debate, because I know next to nothing about the field (and don't much care). But, vocabularies are neither inherently objective nor subjective. It depends on the vocabulary. Furthermore, objective is not inherently superior to nor more useful than subjective. I will give you an example from photography and human vision:

Spectral characteristics are not the same as color. We can objectively describe the spectral emission of an object. For example, I could say that some object emits primarily in the 600 nm range of wavelengths. We have a separate vocabulary for color, which is entirely a subjective perception. The object that emits primarily at 600 nm may look red… or yellow… or green… or even blue.

Spectra are extremely valuable objective measures, don't get me wrong on that. But without a vocabulary of color, you cannot successfully or usefully describe what objects look like.

pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
======================================
-- Ctein's Online Gallery http://ctein.com 
-- Digital Restorations http://photo-repair.com 
======================================

I was trying to find a way to apply some of what you were saying to things I've noticed about lenses. I can't say about others, but I've noticed that wide lenses (28mm k-mount lenses) mostly have 5 non-rounded aperture blades. Is this because it produces busier bokeh or less appealing bokeh, but this bokeh produced sharper and harsher transitions between in and out of focus areas?

Conversely, I longer lenses I am familiar with tend to have a larger amount of aperture blades which appear circular when stopped down. I know that more blades and rounded apertures stopped down contributes to more appealing bokeh, but if my assumptions are correct, that this is because it produces smoother transitions between in and out of focus areas? This become more apparent with a shallower depth of field, and less apparent with a wider depth of field.

This is interesting article. It reveals the facts behind the assumptions and connections I made about the relationship between aperture blades, stopped down aperture shape and focal length. Thanks!

Good luck with that.

The effort will be useful in some ways, for some people, but I think the number of interacting factors ultimately will lead to failure.

I agree with you about grain. I have friends to whom grain/noise is a defining positive characteristic, especially for B&W. I only like it in say 10% or less of images I see.

An interesting thought experiment is to imagine that the original photographic processes were grainless. How many people would then like images with added graininess?

OTOH, I'm fairly convinced that one important difference between LF and smaller formats lies in the absolute effects of diffraction and distance of film from aperture. Certainly there is something in many LF images, especially portraits, significantly different than in FF, beyond just grain. The combination of softness with clear definition of details seems impossible to achieve in small formats.

I recently spent some time looking at a large (say 4x5 ft.) print of Karsh's portrait of Churchill. The combination of smooth tonality, ample small detail and lack of edgy sharpness is exceptional.

As to tonalities, I am at the moment mystified by some B&W images. A few landscapes with Fuji 645 and Ilford SFX, another landscape with Nikon D800 @ ISO 140 and an interior, informal portrait with Fuji X-Pro1 @ ISO 2500 with a Tele Elmarit 90 mm.

The image from the smallest sensor, highest ISO, but oldest lens design, has far the smoothest, most subtle tonal graduations, just creamy and delicious. The MF (almost), largest 'sensor', film images just don't have that quality. The vaunted D800, in the hand of a real expert, is somewhere in between.

Moose

As with all things perceptual, you can measure a bunch of stuff, and you can collect a lot of opinions, but making those two match up in any useful way requires very careful double blind testing, which is hard to do and people resist it mightily.

I am pretty sure that a lot of stuff people "see" isn't there. Sorting out what is there from what isn't is hard, and sorting out what isn't there but the perception of which is cued by things which are there sounds like a complete nightmare.

"Would it make sense to try an empirical approach, having people "rank" the bokeh, say, of a series of images, then trying to determine a correlation between those rankings and the measured power spectrum?"

My experience is that what one person finds attractive bokeh is often the opposite of what another likes.

Moose

Well maybe we should start by defining vague terms like "tonality" used in the typical sentence "This lens/film/paper/sensor has a great tonality," something I have been urging each and every Internet pundit to define to no avail for years.

I believe a show-and-tell approach could do something to help: find a picture with "great" tonality and put it side-by-side with a picture showing "bad" tonality, and try to pinpoint WHAT on the images is the cause for judgement and WHERE that is happening.

I find that most people talking about pictorial qualities tend to do so indiscriminately of any analytical ability, usually conflating the technical and the thematic, the materials, the form, and the content.

But for that it would require more effort than mere Internet punditry.

I agree that this is an interesting article, and addresses some important aspects of the evaluation of photography....evaluation of a certain kind. I'd also like to point out that this sort of engineering discussion seems to be reserved to photography alone among the visual arts, at least among enthusiasts. In the other visual arts, only conservators get into this sort of thing, in terms of evaluating an image. The art historian or critic who does is a very rare bird---less a species or subspecies than a mutation.

To me, that difference is the most notable thing about these sorts of discussions, and I have been thinking a lot lately why this should be so, and what it means.

Interesting problem developing a vocabulary when there are so many variables involved - and also, as you note, individuals have large differences in the way they perceive.

At least for your example 3, the offerings of the third party lens manufacturers give some chance of comparing different sensors in a fairly rigorous manner.
I would be very interested to see you let loose with the Sigma 18-35 on four different mounts, for example.

I'm not sure that I understand this article, but one thing I'd begin with is to get rid of the term "image quality." If you even *think* about his concept in terms of "quality," I believe you've already biased yourself in attempting an objective approach to your subject. As somebody else noted, Robert Capa's most evocative war photographs would fail on almost any measure of "quality" -- they are grainy, poorly composed, badly processed, and great.

Once you get past that issue (of "quality"), it seems that any good optical scientist could resolve most of the other issues by isolating each characteristic and giving them a number, or perhaps a position on an analog scale, so it would essentially be irrelevant whether your test was run on a test chart or pictorial subject. You would then get a list of numbers (or scale positions) for each lens or sensor, and by putting them together, you would have an accurate idea of what your camera/lens combination would output.

What would wouldn't get is what the eye would catch, or the mind would see. What DDB seems to want is some kind of hybrid of these things (lens/camera/mind) which I think is somewhat impossible.

Since you mentioned Roger Hicks, his wife, the photographer Frances Schultz (sp?) once wrote an article trying to explain why Roger got "Leica glow" and she didn't, though they used similar cameras and lenses. I can't remember the details, but it came down to differences in technique. In many ways, Roger's technique would be considered less desirable than hers (in technical ways) but that ostensibly poorer technique produced some very fine photos with "Leica Glow." As it turns out, in her opinion, "glow" didn't have much at all to do with the camera or lens.

So I'd ask this additional question -- is what DDB seeks restricted to the technical characteristics of the cameras/lens? You'd need to clarify that before you could make any progress in developing your vocabulary.

This is why I come to this site every day!

Thanks for a thought-provoking discussion. I have no answers -- but I now have some questions to ponder.

Interesting article and hopefully the beginning to a better understanding of an important topic.
I think there are two problems in today's photography world. In the old days progress was slow. People had time to take pictures and look at them. Time to experiment. You needed some skills and experience to make a 'good' image. Equipment developed slowly and in a logical manner. Glass plates to film, large format to smaller 4x5, then medium format which was for long considered too small for serious work. And then 35mm and even smaller. Lenses got better, gradually over decades. Twenty years ago all three formats were used side by side and a professional might have a system for each format (4x5, MF and 35) and use it as needed. Depth of field is less the bigger the format. It would have been easy to minimize depth of field, which seems to be an obsession today, but LF cameras used movements to 'try to 'maximise' the DOF instead, by moving the plane of focus so that important bits are in focus and unimportant air is not.
Now there is a huge range of equipment, sensor sizes, lenses from excellent to bad, and Internet where people share truths without any control on what do they actually know. It seems the most vocal ones have no time to take pictures, just try and compare pixels side by side. And cost of the body and lens are the only objective numbers there are for comparison.

Image quality, on the whole is fantastic these days...it's like clock speed in computers, everyone has more than enough. Defining a vocabulary may be interesting but academic, and probably as useful as it is in high-end audio. At the end of the day, you have to use your eyes or ears, respectively.

To paraphrase Justice Potter Stewart:

"I shall not today attempt further to define the kinds of material I understand to be embraced within that shorthand description ["tonality"]; and perhaps I could never succeed in intelligibly doing so. But I know it when I see it."

Thanks Ctein. Very thought provoking. My sense is that much of the differences between the look and feel of MF / LF versus 35mm is due to perspective and focal length. I suspect that even if you used a very high ISO film on 4x5 sheet film, the look and feel would be very different from a photo taking in the same scene on 35mm using very slow film. I.e its not just a resolution thing.

Enjoy your articles!

Antony

Terms for renderings, tonalities, etc. don't have to be inherently positive or negative; in fact at their most useful they are descriptive, not evaluative.

Being able to identify particular characteristics is a necessary precursor for saying "I like whatever", or that you don't. (Doesn't have to be one word; a phrase that we agree on the meaning of works fine. What's needed is many people familiar with the definition looking at a set of photos and agreeing on which ones do and which ones do not have that described or named characteristic.)

Right now, we agree roughly on what we mean by "grain" for film photos; mostly we'd agree on which photos were grainy and which weren't (especially if we simplified the question to ranking pairs; which is more grainy). This doesn't let is discuss whether a particular film grain pattern is more appealing than others precisely, but at least we can fairly quickly determine that I basically don't like grain at all, but some other people rather do. It's the agreed concept of "grain", I believe, that lets us reach that point of at least partial communication about our differences.

Similarly for "saturation" and "contrast"; we can discuss the "look" of Velvia relative to Kodachrome in terms that mean roughly the same thing to most of us -- but we won't discover that we agree on what we like. Still, being able to describe the differences in what we like remains valuable, useful.

"Bokeh" was itself a significant advance in this, and I think it's quite useful -- even though it's much less specific than any of the others. Bokeh means, basically, "the way the out-of-focus bits look". It does not mean any particular way they look; we don't yet have terms for the different ways they can look. Simply having a term for the general concept has helped people discuss it. And we find that some people care about it a lot more than others, and that those who care about it don't totally agree on what they like (despite a certain tendency for some people to say one lens has "good bokeh" and another has "bad"). Before "bokeh" became a common concept, people nearly never mentioned how the out of focus areas looked (unless they were discussing a mirror lens; that one was so remarkable that people remarked on it even in the absence of terminology). It just went into the big bin of non-specific reasons people probably liked one lens and not another. (Our host Mike Johnston played a major role in bringing this term into American, or more broadly English language, usage.)

We've also now got the word "micro-contrast", to describe little relative changes in regions of the photograph independently from the big overall changes, and I think that is starting to let us find some other ways both lenses and processing techniques differ.

None of these has been about defining "good" vs. "evil" (or "ugly"). They're all instances of identifying specific objective characteristics. We always, then, seem to discover that people have somewhat variant preferences, either overall or at least for particular subjects.

I suspect there are more image characteristics out there waiting to be identified and defined. (And I suspect that at least some of them are well-known to lens designers. I'm fairly confident that the designers at both Canon and Leica can describe quite precisely many of the ways current Leica lenses differ from current Canon lenses; lens lines wouldn't commonly have characteristic looks if the designers weren't able to control all those things.)

A good descriptive vocabulary is only a part of the problem, and to a considerable degree is a derivitive one. First its necessary to define 'image quality', which is highly subjective and subject to much dispute. The lover of razor sharp images will not evaluate quality the same as one who loves soft 'romantic' images. There is also the fact that there are physical attributes which correlate with -to varying degrees-image quality. Resolution, sharpness, dynamic range, color contrast, saturation, and on and on. Then its necessary to establish scales of 'quality', and determine how these relate to the perception of quality by humans. Ideally, at some future date, it would be possible to generate an algorithm (or set of) to describe 'quality' and predict how people will perceive it. I'm not sure the tools exist, but possibly something such as Fuzzy Sets might work. This whole issue requires a big research program in perception and sensory psychophysiology if it is to be formalized. You might look at the literature on the concept of 'beauty' to get an idea of what is involved. Finally, the very concept is subject to radical change. Look at the academic art world of Europe when the Impressionsts came along, and then again when Cubism arrived.
In the meantime we poor unquantified photographers will probably be doomed to arguing about 'image quality' for the remained of our careers. At least the discussions can be fun.

Re: Example 1: While I agree with the conclusion the you and Roger Hicks arrived at with regard to the presence or absence of grain being the distinguishing factor between film formats, I think that digital photography adds confusion to the discussion if it's not considered apart from film work.

When comparing prints made from 35mm negatives to those made from a (reasonably competant) digital source, the "analog" images often look more "digital" (in the classic sense of the word, where digital means "made up of discrete bits of data") due to the presence of grain. In the absence of obvious aliasing or pixelation, a well made print from a digital source more closely approximates continuous tone than a print from a similar format film source. I think that in large part, this explains why many film photographers are put off by digital-based prints. "Soulless", "without character", and "too perfect" are some common descriptors that are used to argue the inferiority of one medium against the other.

The two media need to be considered separately. While grain may be the most obvious distinguishing factor between film formats, it's not the defining difference between the various digital formats. There is an observable qualitative difference in the output of digital sensors of varying size. A camera-phone, a 2/3" compact, a 1" mirrorless, a Micro 4/3, an APS-C, a "full-frame", and a medium format digtial back, each with the same pixel count, will all yield images with visibily different characteristics, even when exposed with the same lens.

It would be simple if the primary difference were the presence of grain or noise. Unfortunately, it's not so simple. Where are the words that we can use to define those differences? I submit that they don't exist in any quantifiable, objective sense. Words like "micro-contrast" and "tonality" are at best a poor attempt to objectify qualities that aren't any more measurable than the sweetness of one violin over another, or the difference in timbre between two pianos.

I agree with Manuel, in his comment above, "Maybe we should just accept that “image quality” is a subjective judgement, and keep it at that."

Ah, if only photographs sank or floated based on technical parameters that were so easily quantified and discussed in scientifically precise discourse. Such a "vocabulary" might be essential for, say, a printing lab service. But beyond such a limited task-specific realm not so much, particularly since fewer and fewer photos are ever viewed in print form today.

The challenge of creating and appreciating "good" photography begins and ends between the ears. It really always has. We've reached a point where any camera can do a fine job for most purposes, from the least expensive (or phone) to a $70,000+ medium format kit.

Young people would be better served learning some of the well-established observational skills and language of the art world. "Awesome" or "Sucks" are certainly workable polar positions but there sure are many other very descriptive words and phrases in between.

It is reassuring somehow to read that even great photographic minds are not sure what defines subjective image properties. I don't understand the "better tonality" argument for medium format much. I have thought that one big difference between the look of different formats, regardless of grain or tonality is the depth of field for the same aperture and field of view. A normal lens on an APSC digital is about 35mm, vs. 75-80 mm on a medium format 645 or 6x6 camera. For the same aperture and composition, the medium format shot will have a shallower depth of field. Even adjusting apertures for the same depth of field, perhaps the rate of defocus as you move out of the in-focus zone is different; I have not looked at the theory. I know digital photogs have been developing techniques to get the out-of-focus background look with wide angle fields of view by using telephoto lenses and stitching. I prefer to just shoot old film cameras. But even then, I do so mainly because of the tactile fun; I can't always say the look is much different.

Darn it, is photography an art or a science?

I think the answer to your question, Grasshopper, is more philosophical than empirical.

By which I mean, the best lens is the one that makes me go oooh. When you de-obfuscate the nature of "oooh" in less than 10,000 words, I will personally buy you a month's supply of tea.

Though of course, only I will judge your success ;-) Such is the subjective nature of artistic merit and (to quote Southpark) kwaliteh!!!

This subject has been confounding me for some time and you are right; without an accepted vocabulary we simply don't get the opportunity to have a meaningful conversation.
Having just read through Gregory Heisler's 50 Portraits I find myself wondering how I got by without an 8x10 or even his 11 x14 view camera for my own commercial and commissioned editorial portrait photography. I have made do with a 'full-frame' digital camera for the last ten years and was feeling quite smug about it!
The beautifully printed examples in his book and his thoughtful and articulate commentary is a great starting point for a conversation about why larger format (film) does far more than add sharpness and detail. There is a great passage on pp 158-9 where he marvels at how the 8x10 neg though not as sharp as the same frame on a smaller digital file, could be enlarged further and look more satisfying. His words, " (the subject's) skin looked alive on film, with blood pulsing beneath its surface. In comparison, the digital images looked flat and pasty. His color on film was more complex, more human."

This is the same food fight as back in the 70s trying to explain why a contact print of a 4x5 inch negative looks so different from a 4x5 inch projection print on the same paper, and why different enlargers can have radically different looks, and why for instance Richard Avedon's eight foot x ten foot prints seem sharper than his eight inch x ten inch contact prints.

Tone and sharpness are size related. What seems like good tone and sharpness at one size will be different at another size.

For instance here is a picture http://hughcrawford.com/gigapan.large.php?id=82664 where if you view it small the masonry looks well defined and the trees not so much, but zoomed in (it has the same interface as google maps) the trees have all the tone and sharpness and the masonry is sort of a noisy mush. It looks pretty good printed 7 feet tall but doesn't really work in a little 16x20 print.

Similarly in this http://hughcrawford.com/gigapan.large.php?id=128850 the tone is dark and mushy is over all unsharp in a little 16"x20" , but at 5x10 feet it's sharp and has almost too much contrast.

Many 35mm B&W photographers accentuated the grain in their images because it made them look sharper when printed at about 8x10 inches, but further enlargement made the subject matter less and less sharp at the same time the the image of the grain was getting sharper. If you tried to make a conventionally sharp 35mm image you had you choice of good "detail" or good "tone".
All this whinging about DOF and Bokeh and aperture blades is a obsfucation, the real issue is do you want local contrast and fidelity (AKA sharpness, acutance, more high-frequency power and low noise) or global fidelity (AKA tone, enough detail everywhere, good shading, and paradoxically more noise)

The high ISO low noise craze is so wrong headed. High ISO simply means that it takes less light to get to the clipping point of the sensor, and low noise just means more aliasing. One reason medium format digital cameras have low ISO ratings is because of the increased capacity to record light, the same as when in film days some photographers preferred the slowest, largest, and thickest emulsion film that they could use. That's why you see piles of those 2400 wattsecond power packs in old studios.

William Mortensen's book "Mortensen on the Negative" (literally) sheds quite a bit of light on the subject particularly in the last chapters. He explores a lot of what today we would call "tone mapping" but he called "tone light" vs "texture light". Mortensen was a proponent of Gama Infinity processing where he would extend negative development times for days to get more adjacency effect and local contrast at the expense of increased grain which was a non issue with large format film.

William Mortensen is sort of the antithesis of Ansel Adams , but I came to find that Mortensen's approach was much more relevant to my work in terms of technique if not aesthetics, and I think in the context of Photoshop and tone mapping is more generally relevant today.

Found a good thumbnail description of Mortensen's photographic practice here:
http://unblinkingeye.com/Articles/Mortensen/mortensen.html

Dear folks,

Waall, what ah think we have heah is a certain failiah ta communicate. [s]

“Image quality” is not a measure of merit, it does not refer to whether a photograph is a good photograph or a bad photograph. It certainly is not a synonym for artistic or aesthetic worth. It's an entirely independent dimension. Actually, a whole set of dimensions, which was one of the points of my column.

If it makes you happier, substitute the phrase “qualities of an image” or even “image characteristics” for the phrase “image quality.” It's all the same thing.

I could elaborate, but instead I commend you to read DDB's comment. He got this exactly right.(Which does not surprise me in the least.)

Now, for those of you who want to discuss artistic or aesthetic merit, great! There already exists substantial vocabulary, developed over centuries of aesthetic analysis and criticism. You are not obligated to avail yourself of this vocabulary, but it's not one that needs inventing.

For those of you who think that artistic merit is a pointless exercise, that's also fine. Understand, though, that without a vocabulary, it is completely impossible to discuss the artistic merits of a photograph. If your discourse is limited to, “I like it, therefore it's good” or Ken Tanaka's “awesome/sucks,” that's the end of the conversation. Nothing's left to be said.


pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
======================================
-- Ctein's Online Gallery http://ctein.com 
-- Digital Restorations http://photo-repair.com 
======================================

Dear folks

“Currently indefinable” is not a synonym for “subjective.”

One thing that's important to understand is that while our preferences are subjective, the image qualities we're talking about are not. They have objective reality. That does not mean, though, that we have the descriptors for the important characteristics of that objective reality, which is the point of the column.

Here's a more refined example of this, around grain (you can create analogous paragraphs for noise, if you like). We have measures of overall graininess that are pretty robust. We have weaker descriptors for the characteristics of that grain, such as smoothness, basically the distribution of grain sizes, but we know this is important to people's evaluation of how grain looks to them.. We have even worse ways to describe the outliers, the things that are especially noticeable––the occasional “boulder grain” or “shot noise.”

None of this goes to the aesthetic question of what kind of grain one likes, or how much or how little. But without a reasonably complete description of what grain looks like, what its image qualities are, it's hard to talk about why we do/don't like the grain in a particular medium (or even individual photograph) or what kind of changes to the grain would make it more to our liking.


pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
======================================
-- Ctein's Online Gallery http://ctein.com 
-- Digital Restorations http://photo-repair.com 
======================================

Thanks for this post - a simple question leads to some difficult and complex issues and areas for exploration.

The different "looks" I see between my lenses have a lot to do with how "gentle" the fall off appears to out of focus areas if that makes any sense. So it's not just the in focus and bokeh, but the transition from one to the other. Admittedly, a poorly described sense/criteria, but some lenses, say my modern Canon L glass, can be harsh in these transition areas, leaving a sort of flat impression to some photos. In contrast, my Canon LTM 50 1.2 is a difficult lens to use in most cases, but it has such a beautiful fall off from focus area to out of focus area - you could call it a certain glow maybe? It's certainly distinctive for me compared to my other lenses.

Dear Moose,

Regarding smaller vs. larger formats, I'm totally in agreement that there other things going on besides granularity. Really important things. It probably would have been more accurate to say that what Roger and I tweaked to was that granularity was a more important dimension of image quality than resolution in people's perceptions. Not that it was necessarily the most important dimension. That's an argument I'd be willing to indulge, but I wouldn't feel confident I could prove it.

But there is absolutely, definitely something else going on, and I haven't pinned it down.

I see the ineffability that you're talking about, but I don't think it has to do with depth of field or the contrast between in focus and out of focus areas. In my work I endeavor to make out of focus areas completely irrelevant, and I see what you're talking about. In particular, I've seen it in 6 x 7 cm photographs I've done on high-speed, grainy film of subjects at effective infinity, so depth of field or differences in perspective are not a consideration, and the grain is as bad or worse than I would get out of 35mm. Vis:

http://ctein.com/Chicago_Skyline.jpg

Nonetheless, the photographs still look fundamentally *different* from 35mm photographs. And I can't tell you why. I lack the vocabulary!

One notion I've been noodling with is that there's a bandwidth bottleneck in the chain; essentially, that larger formats (and especially the larger lenses) allow more information to propagate through the system, and that this is actually a visually-meaningful difference. It's a question to throw informational optics at. I find it intriguing. I would very much like to do a proper analysis of this in my copious free time.

So far I've been thinking about this for about 15 years and I've yet to find that time. I am not holding my breath [sigh].

But, you know, there's probably a PhD thesis in it for some enterprising young grad student who wants to run with the notion. Hint. Hint.


pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
======================================
-- Ctein's Online Gallery http://ctein.com 
-- Digital Restorations http://photo-repair.com 
======================================

Please provide very specific, concrete evidence for this statement:

" There's a huge amount of internal signal processing that goes on before we get our grubby mitts on the photograph (even with RAW). Overall, it makes the photographs look better; we'd be a lot less satisfied with our cameras without it. But it doesn't make any pretense of linearity"

I know Nikon throws away information that represents negative values. Of course many brands offer lossy and lossless compression. But lossless compression can not affect linearity. I doubt you can find any data to support these claims.

I'm not much interested in attempts to measure the hell out of everything, but as usual with TOP the range of comments provide lots of food for thought.

I'm with JC on the need to step lightly around terms such as image quality, otherwise people might think you're suggesting we can measure whether one photo is better than another (which you're not, of course).

There's a lot of wisdom in Ken Tanaka's comment too. I think it would be great if people concentrated not only on developing a more nuanced vocabulary, but also the ability to articulate their sensibilities and judgements. The art of good photo critique was probably always elusive, but in today's internet age it is even more an endangered species just because of the amount of images out there that never receive any more than Ken's "Awesome!... Sucks!". I'm afraid that extending our ability to measure and compare everything will just lead to a "price of everything, value of nothing" mentality.

In trying to understand how technical measures translate into image quality, I cannot recommend the follow Zeiss publications highly enough:
CLN35: Depth of Field and Bokeh — http://www.zeiss.com/C12578620052CA69/0/4FAB9EF851C018C5C12578D200405960/$file/cln_35_bokeh_en.pdf (Example images: http://lenses.zeiss.com/content/dam/Photography/Camera%20Lenses/international/zip/cln_blog/Understanding_Bokeh.zip )
The discussion of bokeh starts on p.25 and explains quite well how iris shape, correction of spherical aberration, and stopping down influence bokeh, and how spherical aberration influences perception of depth of field (see particularly example image 13).

CLN30: How to Read MTF Curves I — http://www.zeiss.com/C12578620052CA69/0/B8E453399DCB8418C12578D200415458/$file/cln__30_mtf_kurven_en.pdf
CLN31: How to Read MTF Curves II — http://www.zeiss.com/C12578620052CA69/0/A6DCEB1C6ACCE37DC12578D200390F82/$file/cln_31_mtf_kurven_2_en.pdf (Example images: http://lenses.zeiss.com/content/dam/Photography/Camera%20Lenses/international/zip/cln_blog/MTF_images.zip )
The example images from CLN31 do a wonderful job of illustrating how varying MTF for different spatial frequencies translates into real images. There is also some discussion of digital sharpening relevant to Ctein's example 3.

The table comparing different formats in CLN31, p.10 may be relevant to Ctein's example 1: part of the difference in the "look" of various formats may be that, for a fixed output size & viewing distance, their differing degrees of enlargement mean that correspondingly different ranges of spatial frequencies for the lens/film system are relevant to perceived image quality. In other words, with less enlargement required for larger formats, lower spatial frequencies are relevant to perceived image quality and MTF measures are generally higher at those lower spatial frequencies.

Some things are better left to the senses. Finding words to describe how one sees would be similar to explaining how the brain processes visual stimuli (boring). Sometimes it not about the numbers, but the colors.

@David Dyer-Bennet: "Bokeh" was itself a significant advance in this, and I think it's quite useful -- even though it's much less specific than any of the others.

There is actually quite a bit more terminology for people who really care about bokeh. People will often talk about smooth or harsh bokeh, usually referring to whether out of focus points are rendered with bright centers and dim edges (smooth) or dim centers and bright edges (harsh). Neutral bokeh is intermediate, with blur circles more or less evenly illuminated. It's very common for lenses to have different effects on out of focus points in front of and behind the plane of focus, so people will talk about front and back bokeh. And there are other terms used to describe specific effects, e.g. "nisen" bokeh refers to the way some lenses double out of focus lines (frequently an effect of "harsh" bokeh) and "cat's eye" bokeh refers to the way some bright points are rendered as pointy oblongs toward the corners of the picture (common in fast lenses that vignette strongly at wide apertures).

I am stepping in over my pay grade here, but I do have a quandry. I have been a photographer since 1970, sometimes as a pro and the rest as an eager amateur. I have had great photographic friends, therefore great photographic discussions. Whenever we discussed this subject I always thought that what defined the quality for me was the word clear. The prints that I love best are clear, a word I really can't define but that I can feel. I have seen pictures with this clarity from all formats, film and digital. I have achieved it upon occasion, and strive to get it every time I print. It is elusive. I just feel that I can see further "into" these prints than others.

As some have said, this doesn't seem to be a measurable quality but worth pursuing.

What I'm looking for is for people tho talk about various characteristics to be able to present clear examples of them. I'm not asking for complete technical explanations; just really clear examples.

Ideally, I want to see A/B pairs of photographs taken under identical conditions of the same subject with different equipment (format, lens, or whatever) that show differing amounts of whatever characteristic it is we're discussing. (Doesn't have to be different equipment; when we talk about the characteristic "depth of field", we can show it by taking two photos with the same lens at different apertures. I've seen such photos, and in fact we have a pretty good understanding of the physics of depth of field, so when we talk about this one I'm sure we're all talking about the same thing.)

I want the terms of the discussion to be better defined, and more clearly understood, is what I want.

Virtually ever style of photography has a standard bearer. An image could be said to have the qualities of (insert artist and image here).
As a tiny example, I recently took an interesting shot from the inside of my car and converted to B&W and called it an homage to Friedlander. And IMHO it was.

Take this notion to FAR more specific extremes and I think you could A)develop a new language and B)encourage people to "study" a bit more of the history of photography. Just a thought.

Example 2: "the transition from acceptably-sharp to not-acceptable will be affected by that power spectrum. I have a suspicion that this is all very well understood by the best lens designers."

Enthusiasts now discover the qualities of old lenses on cameras like the Fuji X series. Maybe lens manufacturers will design lenses targeting one or another such quality. Modern lenses seem to aim at one ideal of technical perfection, merely trading off cost for achievement of it. An exception would be the Zeiss ZM 50mm lenses. The Planar is one animal, the Sonnar C another. The latter embraces the imperfections of the Sonnar design.

@David Dyer-Bennet — If you haven't done so already, do have a look at the Zeiss papers and example image sets I linked to above. They do exactly what you're asking for in a number of their example images. You can skip much of the technical explanation in the papers and just focus on their explanations of those examples.

Bokeh is a quality with a large parameter space. To understand it as a design issue you need to get into the nitty-gritty of aperture functions and Zernicke polynomials, but if you don't want to rise to that level of mathematical complexity you can learn a lot by reading spot diagrams - plots of where nominally-focussed light rays end up on various surfaces lying fore and aft of the focal plane. Of course, it is not necessarily easy to find, calculate, or measure the spot diagram for the lens under consideration.

I think the issue of sensor/film size is mostly baffling because photographers are used to thinking in terms of simplified versions of the technical reality. Grain is an issue, a very important one for me, but the LF look stands out even at very small reproduction sizes, even when there are no time-based cues such as running water, so I have come to the conclusion that there is something optical going on.

The way I explain it to myself is a mixture of abberations and the finesses of depth of field. First, the residual aberrations of fast lenses for small formats are different to those of longer focal lengths forced to cover larger areas. My 43 mm f1.8 on APS-C, gives a very different look to my 240 mm f9 on 4x5, despite them being roughly equivalent in angle of view and depth of field (I crop square). Aberrations affect the look of a busy background, but also the depth and character of the region where in-focus transitions to out-of-focus.

The second factor is how the circle of confusion expands as you move away from the plane of best focus. Even in an aberration-free lens this expansion is non-linear with subject distance - a fact which is obscured if you just plug in the conventional formulae for depth of field limits. The longer focal lengths used for the same angle of view in larger formats do produce a different distribution of blur - a more linear one, if you like.

There is also the fact that the limit of the blur disk from defocus is the size of the exit pupil, and on smaller formats this tends to be larger in proportion to the format size - you end up with the mega blobs in the background most commonly seen in backlit insect macros and the like.

Finally, my personal opinion is that aliasing is responsible for lack of grace in the way digital cameras handle fine detail. My day job for twenty-odd years was a form of microscopy (STM) in which the best and scientifically most trustworthy results come when you explicitly tune the response of the instrument to optimise the sampling rate and take care to eliminate aliasing. I am sensitised to aliasing, and see it everywhere in the sort of blobby, high-contrast detail which gets called 'resolution' in camera reviews. It's a question of taste, not morality, but I prefer the gentle decay of visibility found in analogue film, or over-sampled digital.

An article with which I can resonate; a welcome to the land of knowledge and choices. You're absolutely right that articulating is underrated; if I look back through photography text-books I'd see advice on DIY b/w film development along the lines of "experiment with several, nail the technique, choose one or two combinations of film and developer and get to know their qualities" with little naming of said qualities (grain, edge-acutance, the specific qualities of staining developers) and no hints of practical application.

Further, I've been increasingly aware of the qualities of bokeh: if it's going to exist in a shot then one has to balance overall and local contrast with an appropriate spatial frequency that retains the creamy softness of highlights and gives local sharp contrast without rendering the bokeh edges crassly.

Two thoughts, however, to bake the noodle.

1A: if a characteristic of the digital race is to reduce noise, then it was no less a desire in film's heyday to reduce grain; advice was always biased toward larger format and slower film.

1B: both film and digital are only start-points, a means to an end: all photography is a journey from a capture medium via artistic choices to an output medium. Whether you do digital->flickr or film->print, the time in post-processing or darkroom is a modulation from film-or-digital towards realism-or-arty-vision. Even if a workflow is nothing but scanning Velvia and applying some tone-curve adjustments, they are still chosen for realism and/or vision.

2: investigate bokeh-panorama, aka "Brenizer" technique. If some LF film results can be characterized by very narrow DoF for the scene at hand and lots of megapixels with low intrusion of grain, can you get a similar "look" with stitching lots of digital images for a wider effective aperture? If not, what qualties are missing?

I think ultimately your quest may be pointless due to the fact that you are using your brain to analyze the image. The way you see the image and how much enjoyment your get out of it is, I think enormously affected by your individual brain and what you expect, not the actual image. I am not saying that you are unable to tell one level of grain from another, or one degree of color from another. I think most people can look at images comparing those things or resolution and agree on which one looks which way or better. I am talking more about the overall look or quality or feel arguments that people espouse with certain equipment.

Studies in wine tasting have shown some amazing things about how your brain affects what you taste. A study in france with PHD students in wine tasting (who should know a thing or two about taste) were given white wine at room temperature and colored with food dye to look like red wine and a huge percentage (close to 40 percent if I remember correctly) could not tell it was white wine. Their brain saw the red color and the warm temperature so they tasted it as red wine.

Studies using the identical wine in two different bottles have shown that people will give much better ratings to the sample that they perceive to be more expensive. When they did MRI imaging of the brain while tasting, the pleasure center of the brain was more active in that same more "expensive" sample. So even though it was an identical taste they enjoyed the one much more than the other.

I suspect this is often the case with the indescribable look of certain lenses cameras etc. I think people genuinely like it better and enjoy it more but it has as much to do with your expectations as it does with the actual image characteristics. It would be very interesting and easy to set up a study where you take two pictures taken with the same camera and tell people one was a Leica or phase one or film or whatever and the other was a less loved camera lens combo and I would be surprised if people didn't like the one better than the other. It would I think you could make it especially easy to fool people if you made different adjustments in photoshop. Alternatively you could take pictures with the two different cameras and switch the captions. Hey that sounds like the kind of thing you could do on this website! Have people rate characteristics on a scale of 1-10. It would be very interesting. Brian

I think we are making a general philosophical error when we try and define positives. It's like trying to define beauty. You could start down the regular features, symmetry and good skin root and end up with a mannequin with no personality.

But we can define qualities that are negative easily. Blur, noise, ca, etc. The best lens or camera is the one with the fewest faults, right?

Well perhaps not. Perhaps it is the subtle flaws that make things special. Just a little noise, a little too warm, etc. After all isn't good photography all about using such effects artistically?

Isn't real beauty the imperfection upon the perfection? The quirks that set things apart? Wasn't the old argument about Zeiss and Leica all about clinical sharpness vs. glow and contrast?

Dear Brian,

It seems to me that your comment is exactly WHY a proper vocabulary is desirable and necessary. You haven't said anything that I can discuss with you, because I don't have any idea what particular image characteristics you're talking about. Unless you're asserting that ALL of them that we don't yet have a good vocabulary for exist solely in the mind of the beholder.

In which case, vocabulary would still be desirable. It's entirely possible to talk about subjective qualities when one agrees upon the language with which to talk about them.

It would not surprise me at all if some of what you say is true. I would be extremely surprised if all of it is true. But we have no way of pursuing the question without a vocabulary.


pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
======================================
-- Ctein's Online Gallery http://ctein.com 
-- Digital Restorations http://photo-repair.com 
======================================

Ctein takes Brian to task a little... but..

"It's entirely possible to talk about subjective qualities when one agrees upon the language with which to talk about them."

Tell me what you think green is like? And what vocabulary (or language) you prefer. You are a colour photographer I gather?

Roger, thanks! The DoF one gets into exactly the sort of examples I'm looking for (these are big enough and dense enough that going through them is fairly slow for me the first time...and I'll need to go through several times to really learn what they're teaching).

These seem to be special editions of Camera Lens News. The index page for what's available is at http://blogs.zeiss.com/photo/en/?page_id=18

Dear Richard,

Thank you for bringing this up. I think color, in fact, is an excellent example of just what I'm arguing for. Color is an entirely subjective phenomenon; it's the interpretation our brain gives to what it sees, but color doesn't correlate at all to spectral luminosities. You can take the same reflected spectrum and, depending on circumstances, it may look like almost any color in the spectrum.

Conversely, a whole bunch of different reflectance spectra can look like the same color. That's why color constancy works–– we see a ripe lemon as looking yellow whether we're looking at it under blue sky light or candlelight. But the spectra we're seeing are entirely different.

If we did not have an agreed-upon vocabulary for color, it would be difficult, if not impossible, to have useful and practical conversations about the spectral characteristics of objects in a scene. Oh yes, you could get there with some extremely complicated analyses and charts and plots, but it would be really, really hard and wouldn't be very useful for photographers. But because we have established an agreed-upon vocabulary, we can investigate color.

We can discuss things like color constancy,people's exceptional sensitivity to the rendition of skin colors (which is what gives film and digital camera designers the biggest fits ––that whole flesh-neutral balance thing). We can even investigate how different people's perceptions of color are not, in fact, exactly the same. There are multiple genetic variations which subtly shift color perception (I'm not talking about color-blindness). Because we have a vocabulary, we can investigate those. We can even investigate color illusions; perceptual situations that produce results that we know are contrary to fact (e.g., fruit that look the “wrong” color), optical illusions of color.

A meaningful discussion of most art and color photography would be utterly impossible if we didn't have a vocabulary for this subjective realm.

pax / Ctein

The one keyword I haven't heard come up yet (it is hinted at obliquely in the Mortensen reference above) is Textural Detail. Grain interacts with this. All that chemistry interacts with grain. The size of the film gives a smaller or larger area to represent that detail - density of information. Analogous to tape speed in audio recordings - 78 rpm get the most "units" (area) to represent a given detail.... I hope this is at least suggestive. I love a fine metaphor in the morning....

Dear Richard,

I just want to say that I agree with everything in your first post. It may not appear to be the case, but I'm interested in making photographs than in measuring stuff. I've done so much “measuring” in my career because people would PAY me! Except for a couple of issues that personally smacked me upside the head -- the focus problem with variable contrast papers, the silvering-out of RC papers, how good (or bad) were my zoom lenses -- I doubt I would've done much of the other research I've done, 'cept I could write reviews about it.

Which doesn't mean I don't think tests aren't valuable, it's just that, like you, I'm happy to let someone else do it and use my time to make photographs instead.

Given the tendency of the Internet to take everything out of proportion, I also have some qualms about opening this Pandora's box. Along with the concerns you raise, I'm positive that I'll see some of the very tentative hypotheses and musings in my column quoted somewhere else, down the line, as statements of absolute fact. I've learned to never underestimate people's ability to take material out of context.

pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
======================================
-- Ctein's Online Gallery http://ctein.com 
-- Digital Restorations http://photo-repair.com 
======================================

Dear Lance,

I *think* that this is what is concerning Oren Grad when he looks at digital photographs, exactly how the “textural detail” (to use your phrase) is being rendered. Not entirely sure, but I think so.

I think that for Oren, though, the absolute amount of detail is less important than the nonlinearities–– the distortions, the dropouts, the weird (to him) “equalizations”.**

Related to this is, indeed, how we respond to that detail. I know several photographers who, when TMax 100 came out, much preferred Tri-X. Not for reasons of tonality or curve shape or speed, but because Tri-X looked crisper. With a decent lens, decently used, you had detail all the way down to the grain/resolution limit of the film. TMax 100 had such high-resolution and was so grainless that you always saw the “rolloff” in edges, due to the natural resolution limits of even the best of lenses. Purely in terms of information content or detail, TMax 100 would win out, but it made evident that photographs were never perfectly sharp, whereas Tri-X masked that.

Some people will immediately get what I'm talking about from this description. Others will not; I'm having to use an awful lot of verbiage to describe something that we lack a good, concise vocabulary for, even though it's a perceptual phenomenon that is widely recognized.

(**Which harks back to the point, which is that lacking a clear vocabulary, we're often not sure if we're even talking about the same things.)


pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
======================================
-- Ctein's Online Gallery http://ctein.com 
-- Digital Restorations http://photo-repair.com 
======================================

The comments to this entry are closed.