中文摘要：

相位重建是实现X射线相干衍射成像的关键,它利用远场采集的样品傅里叶相干衍射花样、结合过采样理论,再采用迭代算法复原样品的相位信息.文中采用数字模拟的方法,利用小尺寸二维非周期性图形作为物场,研究了过采样比对重构结果的影响,研究发现,迭代次数为1000次时最佳过采样比的范围是3—7之间.利用噪声模拟方法,研究了噪声对相位重建的影响,找到了完成相位重建的噪声限是信噪比不能低于10.分析了重构结果中孪生像以及随机平移的产生原因,并给出了相应的解决办法,结果表明,此方法可有效地提高重构图像的质量.

英文摘要：

Phase retrieval is one of the most important steps for coherent X-ray diffractive imaging （ CXDI）,Which uses the oversampled far-field diffraction pattern for phase retrieval iterative algorithm in order to achieve the lost phase information. Here we used a small nonperiodic 2D digital image as the object for studying the effect of the oversampling ratio and obtained the optimum oversampling ratio of 3—7 when the iteration times was 1000. We also added random noise to the diffraction pattern to examine the applicability of this approach to real data. We found the reconstruction failed when the signal to noise ratio is less than 10. As the reconstruction process may fail when the twin image or stochastic shift appears in the reconstructed image simultaneously,we explained why these phenomena come into being,and found methods to overcome the difficulty effectively.