Digital image acquisition is often accompanied with its degradation by blur (out-of-focus, motion etc.) and noise. In many cases, the degradation process can be modeled by convolution g=u*h+n where g denotes the acquired image, u the original image, h the convolution mask (blur), and n random noise. The goal of deconvolution is to recover the original image based on the observed image. If the blur is unknown, the task is called blind deconvolution because it is neccessary to estimate the blur as well. It is an extremely difficult and ill-posed problem, because the available data is insufficient and corrupted by noise. State-of-the-art methods are only partially successful, there is no unifying theory of optimal blind deconvolution. There is a variant called multichannel deconvolution which works with multiple input images of the same scene degraded by unknown and generally different blurs. Such a case is mathematically much less ill-posed and the recostruction is often very satisfactory. Our developed method lies on the treshold of these two approaches, we want to effectively gather information from a single input image in such a way that we can transform the problem to that of multichannel deconvolution and benefit from its excellent reconstruction results.