Convolution decomposition of 1-D and 2-D linear stationary signals

A linear stationary signal is represented as a convolution model, that is, a stationary driving noise convoluted with a system response sequence. The input noise is assumed to be non-Gaussian and independent and identically distributed. The author studies kurtosis deconvolution. The convergence theorems of kurtosis deconvolution in the mean square sense are proven in 1-D and 2-D cases. It shows that one can extract the driving noise and the system response only from the output signal by using kurtosis deconvolution