Blind image separation through kurtosis maximization

Blind source separation has been an extremely active area of research for the last few years. Most of the research has been focused on separation of sources from one-dimensional mixture signals such as speech. More recently, separation of two-dimensional sources (images) has been also examined to a limited extent using second-order statistics, information theoretic models and neural networks. We extend a simple kurtosis maximization algorithm, successfully used in separation of instantaneous speech signals, to images. The higher-order statistics-based algorithm is simple and performs relatively well.