Dave Pogue of the New York Times talks about some fun he had having people on the street compare a 5, 8, and 13 megapixel version of the same photograph.

As he predicted, this post gets a lot of people “riled up.”

[Some commentary after the jump]

I totally believe the results he got. Anyone who has seen some up-resed photos knows that the detail differences due to pixelization is really hard to spot. The worst photograph is 5 megapixel which makes it about 180dpi. At a difference of two feet, there would be no way even theoretically for an eye to distinguish it from one shot at infinite resolution. At a one foot distance, it’d be theoretically possible.

Of a studio shot baby picture shown behind a pane of glass in a cold New York November, the results sound totally realistic (You have a 1 in 6 chance of guessing right). Today’s commercial printers are so good at upres-ing photographs that I’d bet telling a 180dpi from a 300dpi might be very difficult unless you break out the magnifying glass, if even then.

If the subject was different, you might stand a better chance: unmoving, lots of high spatial frequency detail (patterns and such), taken from a tripod, with a cable release, mirror lock up on. Those are some extreme conditions, but you need pretty extreme conditions to make the original print rich enough to be distinguishable.

Imagine for instance, if the subject was a photo test target. Detecting the differences would be trivial because the target has a set of really high contrast black and white converging lines.

Which is sort of the point Pogue is making: theory blows out the window when you are talking about a consumer or working pro buying a camera.

Showing your ignorance

First, the brain isn’t doing any smoothing here. It’s the eye’s natural digitization that is the problem. The eye does a pretty awesome job of preserving all that acuity in the eye when it gets back to the brain.

It’s absolutely embarrassing reading some of the commentary on the blog entry. Most of these people didn’t bother even reading the pithy post injecting their ad hominem: “Exactly what I would expect from a tech-writer.”

Many people asked him to identify the three cameras involved. The answer is the three of them are a Canon 5D, Canon 5D, and Canon 5D.

How can I guess this, because Dave tells you that it’s the same photographed downressed to the different resolutions! The Canon EOS-1Ds Mark II is the most affordable camera with a greater than 13 megapixel resolution. But the number “13 megapixel” makes me think that it was shot with the more affordable 12.8 megapixel Canon 5D. (The Nikon D2Xs falls short at 12.2 megapixel.) He’s not testing cameras differences, he’s testing output.

Then there are the ones who read enough to figure out that it was the same photo in all three cases and then say he should have taken a photo of the same subject using three different model cameras. He’s not testing sensor size or optics, he’s testing output.

And native resolution isn’t going to mean a dot of difference here. More megapixel means a smaller photosite all other things being equal. And a smaller photosite is going to mean more noise. Testing output is exactly the right test to be doing.

Understand Pogue is talking to the consumer. If you are faced with three nearly identical cameras: say the latest Canon SD, whose only claim over the older Canon SD may be the number of megapixels on it and a higher price tag. Well you’re buying into the “megapixel myth” see? (Now this is a bit harsh here because the some of the latest Canon SD cameras have things like a titanium finish and image stabilization to distinguish themselves from older models. But I think you get the idea: my pocket camera has more megapixels than my D70, it’s definitely crossed the line into “too much.”)

Under the condition where the megapixel is the defining factor in the purchase (as it often is), Pogue’s test is completely valid. The point is, are you buying the Canon EOS-1Ds Mark II because it has 16 megapixels? Or are you buying it because it has a larger sensor?

Did I buy my Nikon D200 because it had 2 more megapixels than the Canon and 4 more than my old Nikon? or did I buy it because it’s the speed, balance, metering, and sound of the shutter click is simply unsurpassed?

Many people do things for the former reason. But the real reason lies in the latter.

As for the “more megapixels is for cropping.” That’s some bullshit used by Canon people when faced with the hard evidence that there’s going to be no realizable difference in their 8 megapixel 8×10’s vs. Nikon, Pentax, and Konica-Minolta’s 6 megapixel ones. It is the same shit that went the other way when the tables got turned: Nikon, Pentax, and Sony went to 10 megapixel while Canon stayed at 8 until recently. Crop, my ass. This test shows that you can crop almost 50% in each direction and still retain enough megapixels where the prints are indistinguishable for a regular subject! Read the damn article and pull out a calculator!

No offense, but cropping a photo only occurs when you’re too cheap or lazy to buy or mount the optics you wanted to have had to have taken the photo right the first time. Or, to cover up framing mistakes (which I make all the time). In both cases, it’s the photographer, not the camera here. You have to be comparing a 2 megapixel shot to an 8 megapixel before the “cropping” justification starts carrying any water.