On frequency-domain implementations of filtered-gradient blind deconvolution algorithms
M. Joho, Philip Schniter · 2003
This paper describes an efficient realization of an adaptive single-channel blind deconvolution algorithm. The algorithm uses fast convolution and correlation techniques, operates mainly in the frequency domain, and the adaptation of the deconvolution filter is based on the natural gradient learning algorithm. The proposed algorithm is compared to other methods via computational complexity analysis and simulations.