Hierarchical motion deblurring method for single image

A hierarchical motion deblurring method for a single image, comprising:

• a. providing a target image;
• b. calculating a blur kernel of the target image;
• c. constructing a multi-scale representation for representing the target image and the blur kernel;
• d. from a coarse scale to a fine scale iteratively calculating a residual image between the target image and the blur kernel represented by the multi-scale representation in each scale, and restoring the residual image to obtain a first restored image by using a residual Richardson-Lucy (RL) algorithm, wherein in the iteration, the residual image is multiplied by a first weight matrix defined by a first gradient attenuation function;
• e. from the coarse scale to the fine scale iteratively calculating the residual image between the target image and the blur kernel represented by the multi-scale representation in each scale, and restoring the residual image to obtain a second restored image by using the residual RL algorithm, wherein in the iteration, the residual image is multiplied by a second weight matrix defined by a second gradient attenuation function and a strong edge suppressing function;
• and
• f. comparing the first and the second restored images so as to obtain a motion deblurring restored image.