仪器仪表学报
儀器儀錶學報
의기의표학보
CHINESE JOURNAL OF SCIENTIFIC INSTRUMENT
2004年
z2期
203-207
,共5页
反射型超声衍射CT%图像重构%迭代法%非均匀快速Fourier变换
反射型超聲衍射CT%圖像重構%迭代法%非均勻快速Fourier變換
반사형초성연사CT%도상중구%질대법%비균균쾌속Fourier변환
Reflection mode diffraction tomography (RMDT) Image reconstruction Iteration Non-uniform fast Fourier transform (NUFFT)
基于衍射的计算机层析成像术是建立在Fourier衍射投影定理基础上的.衍射CT图象重构可看作由非均匀频率样点重建信号的问题.提出一种用于反射型衍射CT的图像重构算法,此方法利用反向散射数据进行2D非均匀Fourier反变换.由于直接的非均匀Fourier反变换不易实现,所以采用基于min-max优化准则的非均匀快速Fourier正变换,通过迭代实现非均匀Fourier逆变换的快速有效计算.为了减少迭代次数加快收敛速度,首先用频域插值法得到重构图像的初值,然后根据min-max准则,每经过一次迭代得到重构图像的一个更新版本,重复多次迭代直至得到可接受的重构结果.给出了数值实验结果.与传统重构算法如Gridding方法相比,该算法计算复杂度相当而重构精度较高.
基于衍射的計算機層析成像術是建立在Fourier衍射投影定理基礎上的.衍射CT圖象重構可看作由非均勻頻率樣點重建信號的問題.提齣一種用于反射型衍射CT的圖像重構算法,此方法利用反嚮散射數據進行2D非均勻Fourier反變換.由于直接的非均勻Fourier反變換不易實現,所以採用基于min-max優化準則的非均勻快速Fourier正變換,通過迭代實現非均勻Fourier逆變換的快速有效計算.為瞭減少迭代次數加快收斂速度,首先用頻域插值法得到重構圖像的初值,然後根據min-max準則,每經過一次迭代得到重構圖像的一箇更新版本,重複多次迭代直至得到可接受的重構結果.給齣瞭數值實驗結果.與傳統重構算法如Gridding方法相比,該算法計算複雜度相噹而重構精度較高.
기우연사적계산궤층석성상술시건립재Fourier연사투영정리기출상적.연사CT도상중구가간작유비균균빈솔양점중건신호적문제.제출일충용우반사형연사CT적도상중구산법,차방법이용반향산사수거진행2D비균균Fourier반변환.유우직접적비균균Fourier반변환불역실현,소이채용기우min-max우화준칙적비균균쾌속Fourier정변환,통과질대실현비균균Fourier역변환적쾌속유효계산.위료감소질대차수가쾌수렴속도,수선용빈역삽치법득도중구도상적초치,연후근거min-max준칙,매경과일차질대득도중구도상적일개경신판본,중복다차질대직지득도가접수적중구결과.급출료수치실험결과.여전통중구산법여Gridding방법상비,해산법계산복잡도상당이중구정도교고.
It proposed a reconstruction algorithm for the reflection mode diffraction tomography (RMDT),which uses back-scattered data and treats image reconstruction as a two-dimensional non-uniform inverse Fourier transform problem. As it is difficult to implement a direct non-uniform inverse Fourier transform, a non-uniform forward Fourier transform is performed based upon a min-max optimization criterion, and fast computation of the inverse transform is obtained through iteration. In order to reduce the number of iterations and speed up convergence, an initial estimate of the image by frequency domain interpolation is used. Updated versions are generated with iteration until an acceptable result is obtained. Experimental results are presented. Compared to the traditional methods such as gridding, the proposed algorithm has the advantage of better reconstruction accuracy with similar computational complexity.