A New Method to Estimate Direction from Noisy Motion-blurred Image

摘要

At present, it is difficult, to identify the blurred direction of a motion-blurred image degraded by noise. In this paper, a new approach is proposed to solve this problem by combining median filer, MB-LBP operator with directional derivation. First, the contaminated motion-blurred image is put into median filter to filter lots of noise. Second, to reduce the fuzzy performance in first step, MB-LBP operator is used to make detailed motion texture more prominent in blurred direction. Finally, MDD curve of MB-LBP feature image is introduced to identify direction. Several experiments are conducted on a series of motion-blurred images, and the results show the benefit of the proposed approach in terms of identification accuracy and computational cost in wide range of noise.