作 者: (窦健泰); (高志山); (马骏); (袁操今); (杨忠明);
机构地区: 南京理工大学电子工程与光电技术学院,南京210094
出 处: 《物理学报》 2017年第16期107-115,共9页
摘 要: 在轴向距离参与运算的ptychography算法中,轴向距离误差会使重建图像变模糊并降低图像的分辨率.本文基于菲涅耳衍射理论建立了轴向距离误差模型,根据不同轴向距离误差对重构图像清晰度的影响,提出用图像信息熵确定图像最清晰时的轴向距离,并重建出清晰的ptychography图像.比较了图像能量变化、Tamura系数和图像信息熵这三种图像清晰度评价函数在轴向距离误差校正过程中的分布情况,发现它们均具有单峰性,且峰值确定的轴向距离相同.图像信息熵相比其他两种图像清晰度评价函数具有更高的灵敏性.仿真以及实验均证明了基于图像信息熵的ptychography轴向误差校正的可行性. Ptychography provides an extremely robust and highly convergent algorithm to reconstruct the specimen phase with a wide field of view. The resolution and accuracy of ptychography are severely restricted by the uncertainty of the position error that includes the scanning position and axial distance error. In fact, it is difficult to accurately measure the distance between the target plane and entrance pupil of charge-coupled device (CCD) or complementary metal oxide semiconductor, which results in the axial distance error. The axial distance error can blur the reconstructed image, degrade the reconstruction quality and reduce the resolution. In order to analyze the effect of the axial distance error, the model for axial distance error is derived based on the amplitude constraint in CCD and Fresnel diffraction integral. This model indicates that the axial distance error can cause a stretching deformation of the retrieved image, which is similar to the defocusing effect caused by different axial distances in holography. In this paper, we propose a method of correcting the axial distance error by using the image information entropy in an iterative way to obtain the accurate axial distance and retrieve the distinct image. The correction method based on the image information entropy is composed of four parts: the initial calculation, the determination of the direction search, the axial error correction and the reconstruction of the distinct image. The initial calculation part is to ensure that the intensity of the reconstructed object tends to be stable before entering into the other processing parts. The search direction portion is to indicate that the initial axial distance is greater than the actual axial distance, or less than the actual axial distance. The axial error correction section is to calculate the sharpness values of the image at different axial distance, and find the peak position of the sharpness distribution that corresponds to the position of the clearest image. The axial distance can be taken f