16 July 2009

Compact Digital Cameras: This Is Progress? (Part 2)

We all know that the resolution of digital cameras is measured in megapixels, and a pixel is a single point in an image. Each pixel reproduces a single point of color. Aggregated together, those points of color comprise a photo. These pixels are generated by a photo sensor, onto which these pixels sites are packed very tightly together. Their proximity causes them to interfere with each other, increasingly more as exposure time or sensitivity (ISO) is increased. This interference is registered as noise in your photo, sometimes called “digital grain.”

Except, whereas photo film grain can be rather aesthetic, digital noise is plain ugly. In fact, one of the biggest differences between camera models and brands is how effectively the manufacturers manage image noise. They expend enormous resources to engineer noise out of their cameras, in part because they are constantly building more noise into them.

The dilemma is this. One really expensive part of a digital camera is the imaging sensor, and the smaller the sensor the smaller the lens and camera overall, all of which drives costs down dramatically. But the only way to shrink the sensor is to shrink the pixels and limit or reduce pixel count, or else crowd those pixels closer together. Making more noise.

Since raising pixel count is a vital marketing tool, due to consumer ignorance (everyone likes a simple better-than metric), the only option is to make those pixels smaller and pack them tighter to fit ever more on those chips. And the smaller and denser the sensor, the less dynamic range it has and the more noise it produces. Sensor size and pixel density are critical bits of information, but manufacturers do not advertise them. When you do find the sensor size for a camera, the specification is arcane, leaving you to crack the code.

I mentioned dynamic range (DR). This is the total range of tonality, between the lightest and darkest parts of an image, that a camera's sensor can record. Image data that exceeds this range is clipped, meaning lighter details turn to white and darker details to black. The tiny sensors and lenses of compacts gather and record a much narrower dynamic range than DSLRs. This is a critical issue for image quality. More than anything, compressed DR gives compact camera images that compact camera look. In addition to high noise and reduced DR, compacts also tend to exhibit more chromatic aberration (purple fringing of highlights) and optical defects, due to their little, cheap lenses which have to resolve ever more optical information.

So now we have 12 megapixel compact digicams with itty bitty sensors packed with itty bitty pixels that generate almost as much noise as clean data. Camera makers have therefore become wizards at managing noise. They smear it around and apply all kinds of clever algorithms to disguise it. Given the garbage these sensors produce, what they extract from them is pretty amazing, at least at low ISOs and short exposures. But the overall trend to make more and worse pixels is tremendously discouraging to those who value image quality, or want to do less-than-perfect-light photography, even with compact cameras.

DSLRs take the beautiful, clear, noise-free images that we all want, because they have big sensors with large pixels that produce superb pixel-level fidelity. Couldn’t compact cameras be made to do the same thing, maybe even at terrific compact camera prices? Yes and no. (Part 3)

No comments: