Image Blind Source Separation Method Based On QPSO And ICA

Blind Source Separation (BSS) is to separate source signal from a set of mixed signals when the source signal system and the external environment can not be accurately informed. As it be Put forward in the 1980s, BBS is gradually become an important part of the modern signal. Independent component analysis technology in blind separation is concerned in recent years, which is a development direction in signal processing technique, it is widely used in image signal processing, seismic signal processing, speech signal processing, wireless communication, feature extraction and noise elimination and so on, the direction of this research has important theory significance and use value.Independent component analysis method estimates the independent source signals that is mixed by unknown factors from the observed signal by calculating the high order statistics of the data. The paper introduces the origin of the independent component analysis and the current research situation at home and abroad, focuses on the basic theoretical model of the independent component analysis, the selection of the objective function and the optimization problems, gives the mathematical model of image blind separation, and the mathematical description of objective function, such as non Gaussian distribution, minimum mutual information, maximum likelihood estimation and so on. As the question of numerical optimization method’s poor stability, easy to fall into local optimal solution, image separation effect is not good, the paper puts forward a kind of improved particle swarm optimization algorithm--QPSO algorithm, studies the potential field model, learning mode and the convergence of the algorithm, designs the process of QPSO algorithm and parameters. Then, the paper introduces QPSO algorithm to image blind source separation, instead of the gradient algorithm in original independent component analysis (ICA), in order to optimize the objective function of non Gaussian distribution. Through the simulation analysis, it shows that the QPSO-ICA algorithm has the advantages of fast separation speed, and good effect. Finally, the paper proposes two kinds of improved algorithms in order to further improve the performance of QPSO-ICA algorithm:CQPSO and AQPSO,CQPSO algorithm sacrifices time to improves the precision, and the AQPSO algorithm considers time and precision synchronously. Simulation results on the two kinds of improved algorithm that is applied in image blind separation show that the improved algorithm is superior to QPSO-ICA algorithm.