同步辐射光成像中的最大似然迭代重建法

本研究主要目的在于探讨与评估在同步辐射光源显微成像时使用不同的图像重建算法时的图像质量与重建速度.此次实验我们采用鸟类的羽毛做为受测物体,并且使用位于新竹科学工业园区内的同步辐射研究中心的同步辐射光源及高分辨X光摄影机来取得X光的影像.本研究使用两种不同的算法来重建图像,分别为滤波反投影法与梯度最大似然迭代重建法,作为这次评估的算法.由重建图像的结果显示,我们可以发现梯度最大概似迭代式重建法在边缘的地方有模糊化的现象,另外在图像质量方面,滤波反投影算法的影像上的环状伪影较梯度最大似然迭代算法明显.因此本研究就针对目前同步辐射的图像重建法做一探讨,并对两种不同的重建算法做一适当的评估,以期望提供不同的图像重建法给予不同需求的使用者.

Development of Image Reconstruction Algorithms for Synchrotron Radiation

Our purpose was to evaluate two different reconstruction algorithms applied in micro-radiological imaging using synchrotron radiation.In our experiment,a bird feather was imaged by the synchrotron radiation and projections were acquired by high resolution X-ray camera.All equipments are supported by the Synchrotron Radiation Research Center located in scientific park in Hsinchu,Taiwan.In this work,filtered back-projection (FBP) and gradient-based maximum-likelihood algorithms (ML algorithm) were applied to reconstruct synchrotron tomographic images for comparison.An edge blurring is shown in results of the ML reconstructed image.Ring artifacts are more obvious in the image reconstructed by the FBP as compared to that of the ML algorithm.The FBP can do fast reconstruction,but the ML algorithm can reduce noise and artifacts.Thus,the two algorithms show very different characteristics.