Normalized Spare Matric Funciton and Shock Filter Jointed Blur Trajectory Estimation for Image Deblurring

Blur trajectory estimation is the core problem in motion blurring restoring from a single image. Normalized blind deconvolution and Shock filter were employed to estimate the blur trajectory more accurately by image details enhancement and steady state solution of non-monolithic local minimum. To reduce disturbance of noises and patch textures, the bilateral grid filter was utilized. During the iteration of the kernel estimation, degenerated edge was enhanced by shock filter, the gradient map was achieved by a setting threshold to suppress the noise amplification in previous step. In each step of the iteration of the blur trajectory estimation, the initial value of the next step was obtained by non-blind deconvolution for a more accurate result. Experimental results of synthetic and real images validated the efficient of our method.