Tuesday, November 21, 2006

The Truth About Digital Cameras

David Pogue is right that a 5 megapixel image will print well, but wrong about the implications this has for choosing a camera. First, generating a 5 megapixel test image by downsampling a higher resolution image will produce a smaller image with artificially high quality—unless you’re careful to use a really bad downsampling algorithm. Second, camera sensors and lenses matter. If you set a 13 megapixel camera to shoot at 5 megapixels, it will produce a better image than a 5 megapixel camera shooting at its maximum resolution. A more carefully designed test would show whether today’s printers are bad enough make up for these two factors.

7 Comments RSS · Twitter

>If you set a 13 megapixel camera to shoot at 5 megapixels,
>it will produce a better image than a 5 megapixel camera
>shooting at its maximum resolution

Not necessarily. There are many factors involved here - lens, ccd/cmos size and quality, and so on... If a camera goes for the lower resolution, but instead includes a better lens, you might get better pictures in the end. I think modern digital cameras are so complex that it's almost impossible to make general statements like these.

As a general rule, I think you're right, but I stand by this specific example. How many 13 megapixel cameras do you know of where these other components aren't high quality compared to what you'd find on a 5 megapixel camera?

There's also a consideration of the printing process used. Many labs don't make the greatest of prints. And all the images were uprezzed pretty heavily. It might have been a different story if they stopped at 11x14. It's really hard to say what the limiting factor is in this equation, but there are plenty of good reasons to want more resolution, cropping being one of them.

The methodology used here is deeply flawed, but I think the point is mostly sound. If you're mostly planning on taking pictures of people or snapshots of places you've visited, you're probably not going to notice a huge difference between a 5MP camera and a 10MP camera, especially when you're uprezzing them to larger sizes.

Um, there are 4-5 MP DSLRs from a few years ago that beat any 13 MP point and shoot from today by a wide margin. The lenses on lots of recent cameras can't resolve enough detail for their sensors. I think on most point/shoot cameras, 4-5 MP is about the limit of what they need, and there's no reason to perpetuate this idea that more megapixels == better.

by "lots of recent cameras" I mean the absurd tiny 10+ MP cameras, with lenses nowhere near big enough to resolve that kind of detail.

Yes, I agree about older DSLRs beating newer point and shoots. But when I said "13 megapixel camera" I wasn't thinking of point and shoots. I don't think there currently are any with that resolution--at least I couldn't find any when searching, or when browsing Canon's product line.

I think the basic conclusion--that megapixels beyond 5 don't matter that much for most people--is sound (though I think it does help when cropping). In his comments, Pogue says that his point was that the other camera variables matter more than megapixels. Good. It's a shame the article didn't make that clear and that the experimental design was sloppy.

Although late to this blog...David Pogues comments are frighteningly accurate. I have just taken a series of images, at the same subject, with flash (for least camera shake) with: finepix 4900Z ( eqiv. 3.3MP) finepix E550 (equiv. 8.5MP) Olympus E-20p (5.1MP) KonicaMinolta D5 (6.1MP) Canon EOS 350D (8.0MP) Result: Without a doubt (and it isn't something I'm glad about!) the finepix 4900Z is the best image; on all fronts, by a fair margin. What can't speak can't lie.
As I say I wish it wasn't so!

Leave a Comment