A Blind Image Separation Method Based on Analytical Sparse Representation

Abstract

The invention discloses a blind separation method of an image on the basis of analysis sparse representation. The blind separation method comprises the following steps: firstly, utilizing the analysis sparse prior of an image signal, and adopting a subset tracing algorithm to obtain an analysis dictionary of a source image through learning; then, utilizing a Bregman distance as a target function, and adopting a splitting Bregman algorithm to recover a source signal; and thirdly, utilizing a least square method to estimate a hybrid system, and circularly carrying out the above three steps to obtain a final source image. The invention provides a quick and effective method for a learning dictionary, applies the splitting Bregman algorithm to the recovery of the source signal to quicken the iteration speed of the algorithm, and obtains an optimal separation result through an iteration estimation process of the analysis dictionary, the source signal and a hybrid matrix, wherein the iteration process improves separation effectiveness. The blind separation method can be widely applied to the fields including radio communication, radar and sonar signal processing, medical image analysis, image signal processing, voice recognition and the like.

Description

technical field [0001] The invention belongs to the technical field of image processing. It relates to an image blind separation method. Background technique [0002] In real life, the signal we get is often a mixture of multiple signals. In order to effectively extract the information in the signal, it is necessary to separate these mixtures. Blind separation refers to the process of recovering the source signal by only using the observed signal when the source signal and its mixing method are unknown. The blind separation problem can be described by the following model: [0003] Y=AX+V (1) [0004] The meaning of the above formula is that the n-dimensional source signal X=[x 1 ,x 2 ,...,x n ]∈R n×N Mix through m×n dimensional mixing matrix A, and then superimpose noise V∈R m×N Get the m-dimensional mixed signal Y=[y 1 ,y 2 ,...,y m ]∈R m×N . In recent years, many research results have been achieved in blind signal separation, and many algorithms have been prop...

AI Technical Summary

Problems solved by technology

However, most of the current methods use a greedy tracking algorithm with a large amount of calculation to estimate the source signal. Obviously, these methods are not optimal.

Images