Weighted frame averaging for motion compensation of laser radar image data

Recent interest in the collection of laser radar imagery has motivated the development of automatic, accurate image registration techniques to reduce laser speckle, increase image signal to noise ratio, decrease the deleterious effects of atmospheric tip/tilt, and enhance image detail. This research seeks a new method to assign weights to each of the frames that have been previously registered using an arbitrary translational registration algorithm. Frames with poor registration are assigned low weights according to a maximum likelihood cost function, allowing such frames to be either discarded from the ensemble or re-registered using an alternate algorithm. The estimator is based on the underlying statistics of the intensity distribution of partially coherent illumination, which has found to be well modeled by the negative binomial probability mass function. Simulated and experimentally collected data is presented to support the development and performance of the misregistered frame detector