There’s been much jawing about David Pogue’s claim in the New York Times that 5 megapixels is enough. He didn’t just claim it, he claims to have proved it. I think he may have a point, but the quality of evidence is crappy. First of all, he doesn’t say where the 13MP picture came from (probably a pretty high-end camera), nor how he photo-reduced it (PhotoShop and friends have all sorts of smart anti-aliasing), nor how the print enlarging was done (are the results comparable to the ones you’re going to get at your local print-shop?) It would have been way more interesting if he’d taken a two carefully-composed pictures, one with a 5MP camera and one with a bigger one, had them both photo-enlarged at the nearest drug-store, and compared those. Also, he ignores a real advantage of having more pixels; you can crop a picture way down to focus on the interesting parts and still have enough pixels to look good. Weirdly, he also fails to point out the downside of bigger pictures; they’re slower to transfer, edit, and display on-screen. So this is hardly a triumph of journalism. In this week’s other photo-news, the much-ballyhooed Leica M8, maybe the world’s most expensive consumer digicam, turns out to have quality issues. I still want one; but that would be unforgivably self-indulgent, I’ll only award myself one as a prize, say if I manage to bring peace to the Middle East, or (even tougher) make my Ruby parser go fast.
Comment feed for ongoing:
From: aiddy (Nov 25 2006, at 00:00)
Another downside of higher MP - noise becomes a bigger issue
From: Eric Schneider (Nov 25 2006, at 00:28)
Yeah, David Pogue's right that, at the moment, 5 megapixels is something of a sweet spot for resolution. It's certainly good enough for most photos (unlike what Pogue's article seems to be saying, 150dpi for color digital prints tends to be 'good enough', though higher resolution does help if you're converting your color digital photos to B&W.
Despite the transfer and storage issues, higher resolution is almost always a good thing, mostly for the cropping flexibility that you bring up. Still, I'd have to say that if the higher resolution cost more, I'd probably be more likely to spend the money on faster lenses (which I need for what I do) or travel money to fun places to take photos. *grin*
Still, with most manufactures moving to higher and higher resolutions for no additional cost, upgrading as equipment wears out doesn't seem to have much of a downside.
<a href="http://www.flickr.com/photos/eschneider/295758590/" title="Photo Sharing"><img src="http://static.flickr.com/111/295758590_60e491cba9_m.jpg" width="240" height="180" alt="Trail01" /></a>
From: Chris Rijk (Nov 25 2006, at 03:51)
This "13MP" camera is probably the 12.8MP Canon 5D, which costs about $3000 depending on where you get it from. You're certainly right that comparing a print from a "true" 5MP camera with the "13MP camera" would be much more meaningful than comparing a "normal" and downscaled (then re-upscaled) print. This is particularly the case for cameras which use a Bayer filter, which is almost all of them (including the 5D):
While the Canon 5D has an output image size of 4368 x 2912, in a resolution test using line-pairs (of a pure B&W subject), you'll get a measured resolution of 3500 x 2000 or so, based off the resolution tests from here:
Putting it another way, you could downscale the 12.8MP Canon 5D image to about 7MP (3500 x 2000) without losing much actual resolution (even more if the subject is colour resolution limited). This is not to knock the 5D - all Bayer camera images will be about the same in this regard, and most would actually be worse - the 5D has larger photosites and doesn't need as aggressive an anti-alias filter as lower-end cameras. (An AA filter reduces measured resolution, but also helps prevent nasty aliasing effects).
This sort of thing reminds me of using MHz (or GHz) as a way to compare CPUs from different companies - it's better to "measure" actual performance (especially ones relevant to your intended usage) than to simply compare some "counted" value. And like with comparing CPUs, there's more to camera choice than mega-pixel count.
However, I wish there were a wider variety of photo resolution "benchmarks" out there - the world is colour, not B&W, and resolution tests should include that.
From: Sander (Nov 25 2006, at 05:17)
aiddy hits it exactly - the big problem of more megapixels in current day point 'n shoot cameras is that they get incredibly noisy. The pixelcount might double, but the sensor stays just as tiny, which makes each photoelectric sensor need to be even smaller, thus even more sensitive to light, and thus overall much noisier. Yet manufacturers know that megapixels (still) sell, so they apply hyper-aggressive noise-reduction algorithms to the current batch of 8-10MP point 'n shoots, which makes them lose detail, have saturation problems, etc, etc.
It's seriously a better proposition right now to buy a 6MP point 'n shoot from a year ago than it is to buy a 10MP point 'n shoot that was just released.
Of course all of this is specifically talking about point 'n shoots with their tiny sensors. DSLRs with their much larger sensors still have quite a bit of room for growth in the megapixel count, and although anything in the 6-12MP range is good enough for printing almost anything any non-pro would ever want to print, the possibility of cropping does still make some extra megapixels worthwhile, there (if not nearly as important as build quality, focusing program e.a.).
From: Anthony B. Coates (Nov 25 2006, at 14:00)
I had some photos taken at a track day a couple of years ago, and I asked the photographer about his digital camera. He told me then that it had a 12Mpixel sensor, but they only used the equivalent of 5Mpixels for printing. That was something that made me think.
There is, of course, a lot being written about how the current crop of 10Mpixel cameras lose their extra resolution to the extra noise from the sensors, but that's a separate issue.
From: Lennon (Nov 27 2006, at 09:45)
You don't want an M8 -- it's over-priced, buggy, and *way* over-hyped (think first-gen Apple product, and then distribute that fervor over a much smaller, more fanatic group of customers). In particular, it's low-light performance has little to recommend it over much less expensive DSLRs, and the CCD was manufactured with an anemic IR filter, requiring you to either add an IR filter in front of the lens, or shoot exclusively in B/W to avoid weird (and often uncorrectable) color casts in your exposures.
If you really want a digital rangefinder, go for the Epson RD-1. It accepts Leica lenses, and the actual physical interface is apparently much better. Personally, though, I'd much rather drop $3k on a Canon 5D (even though I'm a Nikon shooter presently) and get a full-frame sensor, since I'm constantly butting up against the 1.5x crop factor of Nikon's sensors.