Image sequence stabilisation using fuzzy adaptive Kalman filtering

Image sequence stabilisation based on fuzzy adaptive Kalman filtering of absolute frame displacements is proposed. Two Kalman filters are operated in parallel, one of which is used as a reference filter and is run with a constant high process noise variance to enable smooth and close tracking of global camera displacements. The process noise variance of the stabilisation filter is adaptively varied according to the residual between the stabilisation filter and the reference filter output.