Motion blur identification based on phase change experienced after trial restorations

In this paper, we proposed and developed a practical algorithm to identify the motion blur either symmetric or nonsymmetric. The algorithm is novel and has sound theoretical foundation. The extent and type of motion blur modelling as five prototypes (square pulse, reversed ramp, forward ramp, reversed trapezoid and forward trapezoid), was found to be associated with the phase change experienced after two trial restorations, which can further be accurately estimated by using image registration algorithm. The characteristics of the algorithm is highly robust, for both measurement noise level and severity of blur. It can still provide satisfactory result when the noise level is up to 15 dB. The examples of blur identification for both simulation image and real image were conducted to demonstrate effective performance of the developed algorithm.