NEW YORK Despite widespread advertising hype that all things digital are better, most engineers know better. There's good digital and bad digital, just like everything else. And there are purists, or diehards, who won't accept digital substitutes until they are clearly better than the analog technology they replace. There are audiophiles who still insist on vacuum tube amplifiers, and photographers who continue to prefer film over bits. Which brings me to image stabilization.
Last month I got a taste of real hands-on engineering with image stabilizers, at a workshop run by The Mathworks (of Matlab fame) at the TI Developer Conference in Houston. About twenty of us sat in a PC-at-each-desk classroom, using their new Video and Image Processing Blockset to build image stabilizer for the Texas Instruments TMS320DM642 Digital Media Processor using a TI EVM DM642 evaluation. It worked great, though resolution was less than standard definition, and we never got to see the results on a TV with interlaced scanning. Which started me thinking: In the real world of digital camcorders, will optical stabilizers -- the purist's choice -- ever really be outdone by digital stabilizers?
I remember back to the Canon press conference in 1991, when the groundbreaking L1 model was being introduced -- the first camcorder with interchangeable lenses. Of course the lens for this $4,000 camcorder cost more than most camcorders at the time, but this lens included, for the first time on a consumer model, an optical image stabilizer. A clear fluid prism physically tilted the image up, down, left, or right to compensate for camera shake, controlled by tiny motors. Sony introduced this technology on their high-end Hi8 models almost simultaneously, and before long the feature was trickling down to models in the $1,000 to 1,500 ballpark. But it was still expensive, and the feature was clearly desirable for more mass-market priced models.
Enter digital image stabilization, also referred to as the electronic image stabilizer -- DIS or EIS. Using an image sensor with more pixels than are needed for a video image, the system creates a "video window" within the image sensor that can move up, down, left or right to compensate for shake.
Digital stabilizers worked reasonably well, and cost than optical stabilizers, making the feature affordable on practically all but the least expensive camcorders.
Over time, camcorders began incorporating image sensors with larger and larger numbers of pixels -- not for video, but for capturing still images. Still, a side benefit was that with more pixels, the range of correction for digital stabilizers could become ever bigger, allowing for compensation of even stronger camera shake.
One would think, given this backdrop, that digital stabilizers would ultimately eliminate any need or desire for the old-fashioned optical stabilizer technology. One would think.
Not so. Most of the higher-priced (above $2,000) "pro-sumer" camcorders -- the kind that the independent filmmaker wannabes, and wedding videographers use -- continue to incorporate optical stabilizers. And the difference, though subtle, is apparent to even an untrained eye. Finding words to describe this difference can be tough -- some describe the digital system as looking snappy, jerky, less smooth -- but regardless of the words, the visible difference is most definitely real.
Panasonic recently confirmed this reality by introducing optical stabilization on a new 3-chip camcorder that probably offers the best video picture quality value today (see Panasonic adds optical stabilizer to under-$1,000 3-chip camcorder.)
As I wrote that piece, one question kept going through my mind: With all the recent advances in digital technology, why do optical image stabilizers endure? Why do they continue to provide tangibly better moving image picture quality?
And, a bit more philosophically, does this speak to a broader question about the limits of digital technology in an analog world? Or is it just a matter of time before the digital systems fully catch up, a matter or processing power, price, and portability?
I have my own theories, but I'm curious to hear yours first. Click on the Video/Imaging DesignLine Forums to post yours. (A very painless free registration is required, it's minimal, and takes just forty-two seconds, I clocked it.)