Fast Blur Kernel Estimation with Texture Preserving DRD for Motion Deblur

The paper presents a novel technique to restore motion blurred images. Motion blur generally appears due to camera shake and exposure time. The blur parameters being the motion blur length and angle are estimated in a fast way using radon transform with less computational time. The image is restored using least squares method with the estimated kernel. The deconvolved image is further passed through an edge preserving deringing filter for removal of ringing artifacts. The method works well for even high blur giving better image restoration with high PSNR and SSIM values.