Research On Phase Retrieval Algorithms In Speckle Imaging

Speckle imaging is a technique for imaging objects hidden behind scattering media.Recently,one of the most popular research direction is speckle autocorrelation imaging technology based on optical memory effect.This method has the advantage of non-intrusive which is not possessed by traditional methods.However,the speckle autocorrelation imaging technology still has many shortcomings.The phase retrieval algorithm which is extremely sensitive to initial values is applied in speckle autocorrelation imaging technology.This defect leads to the technology requires multiple selections of different initial values to obtain a satisfactory reconstruction image.In addition,the current technology lacks an objective judgment standard to calculate the error of the restored image.When the imaging device link is disturbed by noise,the reconstruction quality of the speckle image is quite poor.This thesis analyzed and studied the advantages and defects of the phase retrieval algorithm in this technology according to the basic principle of speckle autocorrelation imaging basic on the optical memory effect.The research focusing on the optimization of the traditional phase retrieval algorithm.Two innovative technologies were proposed to improve the stability of the algorithm.The resilience and noise resistance of the innovative algorithm is improved and enhanced.The main work of this thesis is as follows:(1)The phase retrieval algorithm guided by the sum of gradient amplitudes was proposed in this thesis.This scheme can obtain a high-quality restored image iteratively under most initial values.According to the requirements of error calculation in the algorithm,a parameter called the sum of gradient amplitude was proposed in this thesis.This parameter can not only serve as an objective error standard for the recovery image,but also as a basis for controlling the number of iterations of the algorithm.The simulation and real image experiments show that the improved phase retrieval algorithm is not sensitive to initial values.A superior rebuild image can be iterated under most initial values.In addition,the experiment also proved that the sum of gradient amplitude can be used as an error judgment standard.(2)The phase retrieval algorithm based on Huber penalty optimization was proposed in this thesis.A parameter is introduced into the Huber penalty as the regularization correction parameter.The intermediate solution obtained by each iteration constraint in HIO algorithm is modified according to the Huber penalty.The optimized algorithm can develop the quality of reconstructed images.Furthermore,the algorithm can restore superior rebuild images when the speckle pattern is contaminated by noise.The simulation and real image experiments show that the optimized method can effectively enhance the reconstruction ability of the phase retrieval algorithm.Confront Gaussian noise and Poisson noise interference under different noise levels,the algorithm exhibits better stability and noise resistance.