> From: "Nick Clark" <[EMAIL PROTECTED]> > > Is there any reason IS couldn't be implemented in software? >You could produce a 5MP image from a 6MP sensor by using the extra pixels >to shift the image. You'd need to measure the movement of the camera, >which could be done using a sensor of some sort in the body, or could >conceivably be done by measuring the movement of the image on the CCD. >This could mean that IS could be added to the *istD by a firmware >upgrade. > > This is all speculation, and I could be talking rubbish. > > Any comments?
I doubt you could get fast enough response from the computerized parts of current DSLRs. You could more reasonably implement panorama-tools-like mathematical correction of lens flaws in firmware too, but apparently that is still too difficult an operation to get the cameras to do on the fly. My limited understanding of IS suggests that you will get much better results by implementing the stabilizing in the optical path rather than at the film plane whether you are moving the film plane mechanically or electronically. It's rather amazing how much you CAN'T do to correct flaws in a photographic image by computer. Unsharp mask, for example, does not in fact correct for bad focus--it just compensates for it by increasing local contrast. DJE
