Luminous Landscape is waxing poetic about Panasonic's video feature that uses just a portion on the image sensor. http://www.luminous-landscape.com/reviews/cameras/panasonic_gh2_11_mode_revealed.shtml
Why not use a larger portion of the sensor then crop the resulting video. This would allow you to remove image shake caused by camera movement. You could pick two frames and define the center of each frame the computer would then align the frames in between that would give the smoothest motion.
The simple answer is, that's not the way the technology works. Rather than my entering into a lengthy explanation and discussion here (likely beyond my technical knowledge in any event), I would suggest that you study up on how video is implemented, and the data rates and limitations of current ASICs and CODECs.
It's easy to say. "Why don't they
", but it's also safe to assume that if they