Taking a digital image is often accompanied by its deterioration due to blurring (blurring, motion, etc.) and random noise. The deconvolutionhas the task of reconstructing the original, undamaged image from the damaged image. The image damage can be modeled by a specific mathematical operator - the so-called convolution:

Blind deconvolution, or blind deconvolution, occurs when the convolution mask (kernel) is not known in advance and must be estimated along with the reconstructed image. The reconstruction process is very challenging in this case, as there is usually an insufficient amount of data available, which is also burdened by random noise. From a mathematical point of view, a slightly simpler and relatively good quality image reconstruction occurs in the case of so-called multichannel blind deconvolution, where multiple corrupted frames (g_i) of the same scene of the original undamaged image are corrupted by different unknown convolution kernels (h_i).