A short introduction to the multichannel (MC) degradation model and its discrete mathematical representation will be given. Then an overview of MC deconvolution methods will follow. The subspace-based technique that is intrinsically MC and does not have any single-channel counterpart will be discussed in detail. I will present a fast implementation of this technique using a Fourier transform, which can handle very large matrices that arise from images of moderate size. Other possible advantages of implementing the subspace-based technique in the frequency domain will be discussed.