Blur Busters!

Shake? Fuzzy shots? Bad vibrations? Who you gonna call?

A fundamental truism of photography is that the single most effective accessory for sharper shots is the one most universally despised.

The tripod. Your lip curls just reading the word, doesn't it? Clumsy, unpackable, finger-pinching pests that are heavy if steady, flimsy if light. Damn!

While there's no substitute for a rock-steady camera support, a magic bullet called image stabilization (or vibration reduction, or optical stabilization, or anti-shake) can help you get sharper handheld shots at shutter speeds that would have once mandated a tripod. Since these systems were first introduced in 1994 (on a Nikon point-and-shoot) and 1995 (in Canon lenses for EOS SLRs), image stabilization has become mainstream technology. So we thought it was worth a thorough look at the current state of the art in anti-shake.

The state is pretty good, and getting better. As we speak, the two behemoths of the SLR world, Canon and Nikon, offer a range of stabilizer lenses, and a top independent lensmaker, Sigma, has just come out with an Optical Stabilizer lens. All these can be used with film as well as digital SLRs. The 8-megapixel Konica Minolta DiMAGE A2 EVF camera employs a unique "chip shift" Anti-shake system. And three digital compacts, the Canon PowerShot S1 IS plus the Panasonic Lumix DMC-FZ10 and DMC-FX5, use in-lens image stabilizers. Let's, um, shake down all of them.

How we tested

To some extent, any realistic test of image stabilization has to be subjective: Human beings are going to hold the cameras and lenses in the real world, and so human beings-not "shake machines"-should hold them as part of the test procedure. We therefore had four editors, all experienced SLR shooters, handhold five representative lenses and three cameras, and averaged their results over 3,500 shots.

But the actual measurement of image blur was quite objective. Using our new DxO Analyzer software, we were able to measure precisely how much test images of a pattern of dots deviated from those of a crisp image.

A typical baseline "blur factor" (about 1.0) is the average score of a camera set to 1/100 sec, with a 100mm lens, mounted on a sturdy tripod, and fired by remote release. This is what can be expected in a handheld shot, stabilization off, by a very steady shooter-an Excellent rating. A blur factor of 2.0 represents the minor fuzziness you'd get by applying the Blur filter once in Photoshop. Up to about 2.5 we consider a Good rating. By blur factor 3.5, the image will be noticeably unsharp, but could be usable in modest magnifications after some sharpening was applied-a Marginal rating. A higher blur factor will give you an unusable image (unless, of course, you want blur), and we call that Unacceptable.

This test is also considerably tougher than the one we used back in 2001 ("Canon Image Stabilization vs. Nikon Vibration Reduction," September) in which we shot the test lenses at infinity. Because this time we used a test target at relatively close distances, the angular displacement-the movement that makes the shake-was much greater than at infinity. The test distances were about what you'd use for a wildlife portrait with a 400mm lens, a head-and-shoulders portrait with a 100mm lens, and tight available-light candids at wide to normal focal lengths-typical real-world shooting distances.

The pictures at bottom of the facing page show the four blur levels on our test targets, and corresponding shots of our "test butterfly."