Dual-receiver blind identification for image blurs

Degradation of images by blurs and noise frequently occurs in data transmission. Identification of the blur parameters is crucial to accurate restoration of the degraded image. This paper proposes a deterministic blind blur identification technique that uses two different blurred versions of a single image in the identification process. Through row-ordering, the blurred images are put into a matrix form and then combined using a portion of the matrix having the same size as the blur. It can be shown that through singular value decomposition, the blur coefficients may be uniquely determined, given a matrix null space of unity dimension. Experiments show that identification is also possible with noisy data that has no null space. Application of the process in image restoration confirms the usefulness of the method in actual practice.