Single Channel Blind Source Separation Algorithm For Aliased Images

Blind source separation(BSS)is to recover or estimate the original signal only by analyzing the observed mixed signal under the condition that the original signal and mixed model are unknown.At present,multi-channel positive definite BSS algorithm has been widely used in remote sensing image processing,medical signal processing,communication signal processing and fault detection.But in reality,due to various limitations,it is impossible to deploy multiple detectors,and the number of observation signal channels obtained is often less than the source signal.In extreme cases,only one mixed signal can be obtained,which is called single channel BSS.At this time,the information obtained is further reduced,the positive BSS algorithm is invalid,and the separation difficulty is increased.At present,the BSS technology for single channel is still developing.Firstly,according to the principle of independent component analysis(ICA),the BSS algorithm for positive definite mixed images is studied.Firstly,the mixed images are preprocessed by de averaging and whitening,and then the BSS algorithm combining different objective functions and optimization algorithms is proposed to obtain the estimated image of the source image,Through the simulation experiment,the performance in different cases is analyzed,and the algorithm with the best separation performance is selected for further research.Then,based on the research idea of Virtual Multi-channel,the problem of single channel BSS is solved.Through empirical mode decomposition and variational mode decomposition,the single channel observation image is decomposed into multi-channel,and the ill condition problem is reconstructed into the adaptive problem.Combined with the positive definite BSS algorithm proposed above,the source image is estimated.Aiming at the mode aliasing phenomenon in empirical mode decomposition(EMD),an improvement is proposed.White noise is introduced to smooth the interference,and the PCA is used to select the appropriate component to replace the algorithm,which improves the separation effect and algorithm efficiency.For the influence of the number of modes in the variational mode decomposition,an improved scheme with feedback process is proposed to improve the performance of the algorithm.Through the simulation experiment,the separation effect of the algorithm is verified,and the influence of noise is analyzed.Finally,through the experiment of the reflection interference image in the actual situation,the interference can be better separated,and the effectiveness of the algorithm is verified.