Skip to Main Content
The Recovery of a motion-blurred image is an important ill-posed inverse problem. The goal of the present work is to provide a probabilistic method for the detection of motion parameters based on the geometrical characteristic of the Fourier spectrum. To achieve this goal, the method first detects the maximum meaningful parallel alignments in the frequency domain of the blurred image using the Helmholtz principle. Then the parameters of the blur are identified by an adapted K-means cluster algorithm. Feasibility of the algorithm is demonstrated by three experiments.