Resolution wars (are pointless)

Maybe it's a party, or a wedding- some social event, in any case. A group of friends wants their picture taken. A tiny point-and-shoot camera is shoved into my hand: "Can you take one on mine too?"

Sure, of course I can.

"It's already set, just push here."

The great fat DSLR hanging from my neck, and the speedlight and 50mm prime lens tucked in my pocket, should be a sign that I know where the shutter release is. But yes, I appreciate the hint.

So I hold up the little thing, and the "16MP" symbol is showing. I sigh, take a few snaps, and return the camera.

Why the sigh?

Your 16 MP camera is not really 16 MP

"16 MP" on a spec sheet means that the sensor has (approximately) 16 million active elements. Unless you shoot with Sigma Foveons or that weird new monochrome Leica, your camera is using a Bayer array to separate out the colours. So 8 million of those pixels are green, 4 million are red and 4 million are blue. The camera then cheats with a bit of math to create an image with 16 million green, 16 million red and 16 million blue pixels.

Splitting the sensor into red, green and blue pixels also means that, for colours to appear correctly, every feature in the image must be spread over several pixels. So the camera maker adds a "low pass anti-aliasing filter" to the sensor- essentially, a crystal that spreads out points of light so that even the smallest feature will be spread over a red, a green and a blue pixel.

To fit a point-and-shoot's sensor into its tiny body, the lens and sensor both have to be very small. The fundamental limits of the optics, though, stay the same- diffraction, in particular, becomes a limiting factor with small pixels. On a DSLR, with its large sensor, a perfectly focused point of light might cover one or two pixels. The point-and-shoot has much smaller pixels, and that same perfectly focused point might cover five or six of them.

If you combine these three factors, it's obvious that most "16 MP" cameras can't actually produce 16 million pixels worth of resolution. The only difference between a 6 MP and a 16 MP file from a point-and-shoot is how much extra space is wasted to save image data that is blurred beyond recovery.

Pixel count is pretty much irrelevant anyway

Composition matters. Exposure matters. Focus and depth-of-field matter. Colours matter.

Pixel count does not matter, unless you're printing huge pictures to hang in a gallery.

I have a gamma ray imaging experiment running right now. The sensor on this thing is worth about $80,000. It tops out at one (1) megapixel, and that is usually too high- I downsample it to 0.26 MP for most work. The MRI and CT scanners at your local hospital usually work at 0.26 MP per image slice- maybe 1 MP if the doctor's looking for really fine detail. Here, where image quality really does mean the difference between life and death, we care about noise levels, contrast, blurring and image artefacts- but not about pixel count.

What resolution do you really need for photography?

  • For viewing at book distances, 300 dots per inch (dpi) is high enough that humans can't discern individual pixels. Beyond 300 dpi, you're limited by the texture of the paper fibres.
  • At arm's length, 200 dpi is plenty.
  • For a poster on the wall, 100 dpi only starts to appear pixelated if you lean in so close that you can't see the whole poster.
  • Most computer monitors are 72 dpi, although Apple's latest laptops can be had with up to 220 dpi- which is roughly what we'd expect for a device used at a bit less than arm's length.

Let's try relating dpi to megapixels, in convenient graph form, and see how big we can print. Remember that what our eyes perceive is linear resolution; when a camera manufacturer quotes a megapixel count, they're reporting the square of the linear resolution.

Or, if you're into ultra-high-resolution systems like 80MP medium format digital, or 4x5 film (which scans at 100 to 300 MP), we'll slap a logarithmic scale on there:

I think it's pretty clear, then, that there is no need to go higher than the following resolutions:

  • Posting online - 1 MP
  • 8"x10" prints - 6 MP
  • 11"x16" prints - 16 MP

Do you have an 11"x16" photo printer? No? Then don't shoot at 16 MP.

If you print larger than that, your viewers are likely to stand farther away- so you can reduce the linear resolution (dpi) of the print, without anyone noticing. Higher resolution won't hurt, but it also won't make any significant improvement at normal viewing distances.

Ridiculously high resolution cameras, like Nikon's 36 MP D800, really don't give you much of an advantage over their cheaper siblings. 36 MP is only (sqrt 36/16) = 1.5 times the resolution of a standard 16 MP digital SLR camera. Unless you're printing really huge, really detailed wall-size images (in which case you ought to look at large format film), there is no significant benefit to those extra megapixels.

Most of the 8"x10"s on my wall were shot on a 4 MP (2G + 1R + 1B) Olympus C-series. You can't see pixels. They're crazy sharp.

There is, of course, an exception to this rule for the specific case where you're going to digitally zoom waaaay in on the final photo, like in aerial surveys or forensics. For general candid and travel shots, though, there's no point in maxing out the megapixels.

If in doubt

Just set your point-and-shoot to 4-6 MP and leave it there. You'll be able to print up to 8"x10" with no problems, and 11"x16" with only the faintest hint of pixelation when viewed close up. And with the extra space you'll free up on the card, you can try more variations of focus, exposure and composition for each shot, thereby increasing the number of photos that are worth printing.



Add new comment