Image Underdetermined Blind Source Separation And Its Anti-noise Performance Analysis

In the rapid development of communication and information technology today,image as a major way of information transmission,is bringing us more and more convenience.While enjoying the convenience brought by the development of science and technology,we are also faced with a series of urgent problems in the field of image transmission.It has become a difficult problem for researchers to extract and separate received images effectively in the absence of transmission channel information.As an indispensable and important technology in the field of modern signal processing,blind source separation of images is of great practical value and research significance.At the same time,noise interference is inevitable in the complex communication environment,which makes blind source separation face some challenges in practical application.Considering the information security and encryption transmission,the analysis of the influence of source noise and channel noise is the key to the practicability of the algorithm.On the other hand,in practical application,the number of array elements of the receiving sensor is usually smaller than the number of source signals,which also deepens the difficulty of blind source separation and extraction of source signals and brings some difficulties to the estimation of source signals.In this dissertation,the performance of image blind source separation is analyzed according to different noise conditions and different number of receiving array.When the number of receiving array is small,considering the source noise,we propose the "Component Complement Method" to construct the virtual array,and use the variational mode decomposition(VMD)algorithm to extract the optimal mode as the supplementary array element,convert the underdetermined model into the positive definite model,and realize the blind source separation of the image signal.Based on the time-frequency domain characteristics of the original Image,the "two-step" hybrid matrix estimation is used to reconstruct the Image simultaneously,so as to realize the effective separation of the Image.Consider confidential transmission and information security,the use of the class of chaotic signal noise and wide frequency spectrum characteristic,to cover Image transmission.Moreover,the application effect of underdetermined blind source separation is analyzed,especially in the noisy environment of the channel.At the same time,the key experimental simulation is carried out in the single-channel receiving environment,which provides an effective support for the promotion of image blind source separation algorithm with low array elements.