A recent question prompted me into one of my "thinking sessions," though I found myself taking a different path than I expected to.
The question is this: how do you define image quality?
As many of you know I've been promoting the notion of collecting optimal data for decades now. That's sort of my modern digital take on the Ansel Adams analog approach I learned while young. Optimal data collection provides the most (and best) options for creating a final visual image.
Thus, normally you would expect me to answer the above question like this: collect optimal data.
In thinking about that this time, though, I ended up going a different direction, partly because I was trying to figure out what prompts the question in the first place.
Some of you might say something akin to this: maximum image quality is when you fully capture the reality in front of you (or capture it correctly; substitute your own word here).
Almost immediately I recognized that this is one of the reasons why so many attempt to make everything in a scene in focus (e.g. use hyperfocal distance or focus stacking). That's because their definition of image quality is "rendering every item with maximum fidelity".
That, of course, is not how our brain works, and one of the reasons why I'm not a fan of hyperfocal distance. Crisp and detailed generally tells our brain something is near, while fuzzy tells us something is far away. As I like to put it: "if you can see the individual whiskers on the lion, it's close; if you can't tell whether the brownish blob on the horizon is a lion, it's far away."
I don't like taking the third dimension (depth) out of my photos. I like putting it back in ;~). Hyperfocal and focus stacking techniques work totally against that. I might focus stack a critical region (virtually always close), but I don't like near-to-far crispness, as it works against depth cues.
I did learn a slight exception to my Galenesque norm of near sharp focus, middle moderate focus, far less focused, and that came when examining some excellent Ansel Adams prints closely. Adams had a tendency to use a telephoto-ish lens to bring something far towards us in his compositions. The "Mount McKinley and Wonder Lake" (prior to the peak being officially renamed Denali) has his focused point well towards the peak, letting the near trees go a tiny bit softer (he may have tilted the lens, but not quite perfectly). If you've ever stood in that spot you'll know why: that peak is 27 miles away from you yet dominates your vision, particularly with even a modest telephoto lens. Ansel was exaggerating that impact.
Moving on, color is another problematic concept when it comes to "image quality." Color isn't exactly what most people think it is (a fixed value). Color isn't fixed. For instance, if you've ever used a polarizer filter you probably know that this changed "colors" in the scene when, in fact, nothing in the scene actually changed color. All you did was filter out some perpendicular light waves. The light was deformed (typically by reflections) from the object to you, so is the ultimate image quality recording that polarization properly, or taking it out?
Moreover, color is perceptual. If I walk into a scene wearing what you know to be a white t-shirt, you'll see it as white, even though the light I'm in may be rendering it warmer (redder) or cooler (bluer) in terms of the digital capture.
You can see where this is going: reality—which a lot of folk are attempting to capture in their images—is not something that is easy to define, let alone capture.
Image quality has been attempted to be defined by others. For instance, math and photographer guru Norman Koren presented a formula for image quality, which uses the Shannon Information Theory construct (which is part of Information Theory). Basically, in this form, image quality is related to a formula that uses resolution, signal, and noise to define quality. Imatest, Norman's testing evaluation software, produces a number for this, typically expressed as the number of bits per pixel. More bits per pixel means higher image quality. Of course, many of you will be upset to find that using any form of sharpening (or JPEG compression) will reduce the bits per pixel number, implying the image is of lesser quality ;~).
Which brings us to this: what most people are actually trying to figure out is "how good does that image look to me?"
I can't tell you that, because I'm not you ;~).
But I do know this: if you don't know what you're trying to say (show) with your image in the first place, whether you manage to achieve your goals will happen randomly. That's a common theme I hear from readers. They'll tell me that they once got an image that was "perfect," but they can't consistently repeat that. Most of their images are highly imperfect.
That's akin to one of two things: (1) even a stopped clock records the time correctly twice a day; or (2) when they pay attention to setting the clock and constantly monitoring it, it does better telling you the time than when they stop paying attention.
To me, it all comes back to this: what's the picture trying to do? Who's it for? What did you want it to say?
Some approach photography a bit like trying to write a novel before they have a character or plot. Some random words might read interestingly, but ultimately the work fails because it's just random words in search of character or plot that the reader might enjoy.
Apertures, shutter speeds, ISO values, camera settings, those are all like the words you use in writing: they support the actual idea, they're not the idea itself. Yet most of you spend all your time on this visual equivalent to spellchecking and grammar and ignore the plot. What should your photo convey? What does it need to contain? Tell me those two things and we can set up the camera, lens, and get the timing right. If you can't tell me those two things, then your photo probably won't convey what you want it to, and it will contain things it shouldn't (and maybe leave out things it should).
And if it contains things it shouldn't, your image isn't as high quality as you think it is ;~).
