Unsupervised Neural Network for Nonlinear Noisy Image Separation

X. Zhang, J. Lu, and T. Yahagi (Japan)


NLBSS, SOM, noisy image separation.


Nonlinear Blind Source Separation (NLBSS) has received much research attention recently due to the emergence of simple, powerful algorithms that show promise in prac tical applications. In this paper, we consider the prob lem of nonlinear noisy mixed images. We will propose EM(Expectation-Maximization) as a learning algorithm of Self-Organizing Maps (SOM) for NLBSS problem. It has the benefits of both EM and SOM algorithms, without con straints on source signals. Wefirst used the proposed ap proach to denoise the mixed images and then use it again to achieve separation. We show in the simulation that the SOM-based approach can provide a solution to the nonlin ear noisy image separation problem.

Important Links:

Go Back