Kernel estimation from blurred edge profiles using Radon Transform for shaken images

Motion blur due to camera shake during exposure often leads to noticeable artifacts in images. In this paper, we address the problem of recovering the true image from its blurred version. The problem is challenging since both the blur kernel and the sharp image are unknown. The quality of a deblurred image is closely related to the correctness of the estimated blur kernel. In this work we focus on the use of Radon Transform for blur kernel estimation. It is done by analyzing edges in the blurred image and there by constructing the projections of the blur kernel. Estimation of the blur kernel from its projections is done by incorporating the sparse nature of the blur kernel. The problem is solved through l₁ minimization making use of the estimated projections. After building the kernel, we use a non-blind deconvolution algorithm for producing the sharp image. Results show that this approach is well suited for blurred images having significant edges.