Influence Factors And Optimization Algorithm Of Image Reconstruction Based On Coherent Diffraction Imaging

Coherent Diffraction Imaging(CDI)is a research focus in the computational imaging field,it is a new lensless imaging technology with coherent light source.CDI is the imaging for two-dimensional or three-dimensional objects by using phase information of wavefront obtained by object wave diffraction or intensity distribution of interfering light field.It is widely used in X-ray crystallography,synchrotron radiation imaging and biomedicine because of its advantages of simple experimental equipment,which no need of focusing components in the experimental process and free from the limitation for imaging resolution caused by focusing components.In coherent diffraction imaging,phase retrieval of the diffraction pattern(the experimental captured diffraction pattern is the power spectrum of the imaging object after Fourier lens)is the core of the technology.When the experimental captured power spectrum,the photo-detector can detect the intensity distribution of the light field,but can not directly obtain the phase distribution.Therefore,the robustness of the phase retrieval algorithm directly affects the reconstructed result of the object.Based on the principle of coherent diffraction imaging and hybrid input and output(HIO)phase retrieval algorithm,we found that the support constraint in the domain,the selection of the feedback coefficient ?value and the exposure state of the input power spectrum pattern affect the reconstruction directly.We study each influencing factor by controlling variables and propose corresponding solutions.At the same time,we use the experimental framework of coherent diffraction imaging to image objects hidden behind scattering media,and our method can reconstruct clear objects.The main research contents of this paper are as follows:We proposed the inverse Fourier transform of the input power spectrum pattern as the support constraint.Compared with other support constraints,the algorithm has the advantages for fast convergence and fewer iterations.In many references,the support constraints in hybrid input-output phase retrieval algorithms are various,and the asymmetric center support domain is generally chosen.We found that the position and size of support constraints can affect the reconstruction results,the support constraints are usually set according to the prior knowledge of researchers,and generally do not change with the input power spectrum,which leads to the good or bad reconstruction effect.Our method avoids searching for global optimal solution in a wider range,stabilizes the reconstructed result in the middle of the size of the object domain,and achieves fast convergence of the iterative algorithm.We proposed that the feedback parameter ? can be selected between 0.2 and l.3 in HIO algorithm.In the HIO,the value of feedback parameter ? determines whether the algorithm can reconstruct the clear object.The feedback coefficient ? value for the HIO reconstruction algorithm has to be carefully selected from a wide range of 0.5-1.0 which Fienup suggested,with the majority of ? 0.7 and 0.9.In this paper,the theoretical analysis for HIO algorithm and a lot of experiments show that the value of beta is not limited to 0.5-1.0 interval to reconstruct better objects,but also to recover clear objects in the extended interval.According to further research,it is found that the selection of ? value is closely related to the exposure condition of the input power spectrum pattern.We propose a method that optimizing single-shot coherent power-spectrum scattering imaging adaptively by tuning feedback coefficient for practical exposure conditions.The quality of the input power spectrum pattern in phase retrieval algorithm is the most direct influence in the whole algorithm,when the value of ? is between 0.7 and 1.0,we conclude that the algorithm can reconstruct clear objects with less iterations when the gray distortion of the power spectrum is less than 0.1.The gray level distortion of power spectrum image is mainly due to the high or low exposure time set by CCD photo-detector,which leads to the overexposure or underexposure of the captured power spectrum pattern.Combining the function of feedback parameter ? in the algorithm,we propose a simple formula to adjust the feedback parameter ? adaptively according to the actual exposure condition.According to the formula,when the captured power spectrum pattern is underexposed,appropriate exposed and overexposed,the range of ? is 0.3-0.8,0.6-1.1 and 1-1.25,respectively.Different ? values are set according to the exposure condition to avoid the object can not be reconstructed due to the limitation of ? values in the range of 0.7-1.0.At the same time,this conclusion can also be applied to coherent diffraction imaging of imaging objects hidden in scattering media.