As I have been telling friends for quite a while now, megapixels are no longer a key metric for deciding between different digital cameras. For relatively inexpensive cameras with small lenses, there is probably no advantage to having any more than about six megapixels, in terms of the quality of images you will get. Indeed, having too many pixels crammed onto a small sensor can start to decrease image quality, as pixels that are overwhelmed by the amount of light hitting them ‘leak’ into neighbouring ones. Images from sensors with unnecessary megapixels also clog up memory cards and hard drives, and mean that you need a very powerful processor to deal with large numbers of them at a time.
Akira Watanabe, manager of Olympus’ SLR planning department, has declared that twelve megapixels is adequate for all consumer purposes. Furthermore, he has declared that Olympus will now focus their attention on other issues, such as dynamic range, colour reproduction, and low-light image quality.
From a photographer’s perspective, this is very good news. I have taken plenty of great photos with a 3.2 megapixel camera, and subsequently blown up some of them as large as 11 x 14″. At the same time, most small digital cameras have poor performance in low light and problems addressing dynamic range. At this stage, improvements in those areas are a lot more valuable than cramming more pixels onto their sensors.
Of course, it will cause a bit of trouble for people selling cameras in big, non-specialist retail outlets. When I worked for Staples, I was never given any information on products beyond what was written on the little information card in front of it. While it is easy to say: “This camera has eight megapixels, that one only has seven,” it is a lot harder to test, understand, and express more subtle photographic characteristics.
Perhaps companies could compete to bring down the price of full-frame sensors.
Dynamic range is the big issue I’d like to see worked on. In the lab, we would very much love to see the emergence of something called a gradient camera: See here, here.
Instead of measuring the intensity of a pixel, it measures the intensity difference to neighbouring pixels. This allows very large dynamic range across the image, while still preserving fine detail locally. The downside is that recovering the image from the sensor output involves solving a differential equation. So something like a digital view-finder might be hard to implement. It’s quite a radical design, but if the computation can be dealt with, it seems very attractive. An idea I’ll be following with interest.
CAMERAS What you’re told is important: megapixels.
Somehow, the industry has managed to convince consumers that having more dots means better photo quality. And that may have been true in the early days, when two-megapixel cameras roamed the earth. Enlargements made from a four-megapixel camera’s shots could indeed look sharper than a two-megapixel camera’s shots.
But that visual difference evaporated once cameras hit five or six megapixels. Nowadays, even six megapixels is plenty even for enormous, poster-size prints. On small cameras, the experts will tell you, cramming more megapixels onto those tiny chips can actually reduce image quality, because the chip heats up and causes colored speckles (“noise”) in the photo.
What’s really important: Sensor size. A bigger light sensor in the camera means better light sensitivity, which means the shutter doesn’t have to stay open as long, which means fewer blurred shots.
But the camera companies don’t want you to know this statistic — it’s not on the box, it’s not in the ads — because it’s easier and cheaper to goose the megapixels than the sensor size. You can look up a camera’s sensor size on the Web, but even then you’ll be presented with goofy, hard-to-understand, impossible-to-compare measurements like 1/2.3 inches (for small cameras) and 16 x 23 mm (for S.L.R. cameras). They never, ever appear in simple diagonal inch measurements, the way TV screens (and even camera screens) are measured. No, that would be too simple.
More on megapixels and image quality.
Sir,
Your recent article (“Dotty but dashing,” 20 April 2010) claims that incorporating quantum dots into cell phone cameras will alow them to “leap… to a photographically respectable 12 megapixels.” Megapixels are an increasingly misleading figure, when it comes to digital cameras. While it is quite possible to make an attractive 12″ x 18″ print from a 3.2 megapixel file taken from a camera with a decent lens, cameras with 6 megapixel plus sensors can still produce terrible output if they have bad lenses or noisy output.
Twelve megapixels worth of mediocre data is not photographically superior to three megapixels of superior data – a point that should be borne in mind by consumers and journalists alike.
Milan Ilnyckyj
Ottawa