Extracting Dynamics from Blur

This paper considers dynamic state estimation using blurry measurements from image sensors such as CCD(charge coupled device) or CMOS(complementary metal oxide semiconductor) arrays. Typically, the information obtained from these sensors is the time-averaged output measurement during the exposure time. The additional information available in the intensity distribution, termed blur, is disregarded as noise. This manuscript models the image sensor as an integrative intensity sensor and exploits its unique properties to extract additional (non-linear) output information through spatial moments of the intensity distribution. An extended Kalman filter is then designed to exploit this information for better state reconstruction. We illustrate this modeling and algorithm development in the context of state estimation for adaptive optics systems. Simulation results verify that using the spatial moments can lead to more fidelous state estimation.