Variable Norm Deconvolution is a generalization of Wiggin's Minimum Entropy Deconvolution. It is an iterative multichannel method which tries to achieve a deconvolution with a small number of events and least residual error. It is shown to be related to minimizing the entropy of the inputs. An efficient implementation in the frequency domain is presented. It is applied to synthetics and a real data set. Results are subjectively equivalent to those achieved with other methods of deconvolution.