Hi All! Let me contribute my 0.02 in your favorite currency. I don't know how the actual system works in the new Pentax camera. Also, I hope I didn't miss somebody's e-mail that already described everything more concisely and elegantly.
There are two types of IS: optical (hardware) and electronic (software). Some hi-end system implement both. The way I read and understand the description posted in the Pentax press-release for A10, they use the former one. 1. There are two ways of optical (hardware) IS. a) In the first one is this: There is a gyro-based sensing system that detects the motion of the sensor. This (via CPU) provides the feedback to the micromotor attached to the CCD of the camera, and moves that relative to the camera to provide constant position of the sensor relative to the "absolute" coordinate system (i.e. crudely speaking, the Earth). This is the first approximation. If the feedback and response time were zero, it would've worked just like that. However, due to a finite response time, this system needs to be predictive. So, it calculates not only the change in the coordinates, but also the instant velocity and acceleration. This way it can work as a predictive system. Now, I should say, that this first method is similar to how the IS in K/M DiMAGE A2 works: http://www.dpreview.com/learn/?/key=anti-shake b) For some digital camcorders and cameras, a similar by idea, but slightly different in the implementation method is used: "a liquid- or gel-based prism mounted in front of the lens that bends the light so that the image falls on a fixed point on the CCD" (quoted from CNET glossary). It appears, this is the way Panasonic does it in Lumix cameras: http://panasonic.co.jp/pavc/global/lumix/technology/index.html (see their "quick pictorial guide") I am not sure which of the two methods is implemented in the A10. The first one could be applicable for an SLR, the second one would be harder to use in an SLR (and unlikely with the today's SLR concept), as it requires an additional optical element between the sensor and the lens. c) If it was not a question of the size/weight, - the best way of stabilizing the sensor would be to mount it on the gyro-based mount, so that gyro-platform would keep it in a constant orientation while the shutter is open. This would be similar to how some aerial photography is done. If I read it correctly, this is also the way how it is done in the Canon IS and Nikon VR lenses. 2. Electronic (software) stabilization. There are different types of shaking. If the sensor is moving in the image (focal) plane (let's call it XY) ONLY, the lens characteristic shouldn't matter. All what matters is the change in the x and y coordinate. If the camera knows the delta-X,delta-Y shift at each moment of the time (due to the gyro-sensors) - at each moment, it can "reassign" the pixels to those, slightly shifted by delta-X,delta-Y. >From what I read, this is the way how it is done in the digital video cameras. See it on the page 4 of this document from Pentax: http://www.pentax.de/downloads/cctv/de/Operating%20manual%20Profile%20image%20stabilizer%20(English).pdf http://tinyurl.com/cg25w If you have motion along the optical axes (Z-direction), then the lens information becomes important. If the sensor is shifted by delta-Z, the rays that were coming to one point in the focal plane, now are all at different points in the plane separated by delta-Z from the focal plane. So, instead of a dot you form a spot which size depends on the angle of those rays coming in, and therefore, on the FL of the lens. If we know the lens FL, mathematically it may be possible to to reconstruct the image (YMMV). I don't have time to think about it now, but my intuition tells me that in this case one needs not only the FL, but also the focusing distance. There is a variety of what I would call the image-analysis-based (feature tracking) algorithms. See, e.g. here: http://www.usc.edu/dept/LAS/CAMS/Stabilization.htm To summarize, 1a (and 1c) could be compatible with an SLR. 1b - will not be. None of this methods would require knowledge of the lens (at least in the first approximation). 2 - may or may not need knowledge of the lens, depending on the type of shaking and algorithm used. (Some algorithms are not applicable for still imaging.) I hope this lengthy overview explains different stabilization techniques, and I hope I didn't make too many mistakes in it. Unfortunately, upon a quick search in the database of US patents and and patent applications, I was not able to find one from Pentax, but I didn't have time for a thorough search. Besides, Pentax might be licensing the technology from somebody else. (I was able to find one of the Canon's IS patent: http://patft.uspto.gov/netacgi/nph-Parser?Sect1=PTO1&Sect2=HITOFF&d=PALL&p=1&u=/netahtml/srchnum.htm&r=1&f=G&l=50&s1=6,933,968.WKU.&OS=PN/6,933,968&RS=PN/6,933,968 http://tinyurl.com/9acfl ) Igor
