Post-nonlinear blind image separation algorithm with SOM initialization method

An algorithm for post-nonlinear blind image separation is proposed based on self-organizing maps (SOM). This method can improve the convergence rate and avoid lapsing into the local minima. Simulation results verify its validity.