Synthetic Aperature Diffraction Tomography and Its Interpolation-Free Computer Implementation
Synthetic Aperature Diffraction Tomography and Its Interpolation-Free Computer Implementation
复制标题
合成孔径衍射层析成像及其无插值计算机实现
DOI：
--
复制
发表时间：
1984
期刊：
影响因子：
--
通讯作者：
A. Kak
中科院分区：
文献类型：
--
作者：
D. Nahamoo;S. Pan;A. Kak
提出了一种新的层析成像技术，它只需要物体的两个旋转位置。虽然理想情况下两个旋转位置之间的角度应该是90°，但理论预测，即使在不满足该条件的情况下，也应该可以获得有效的结果，尽管空间分辨率降低。对于物体的每个旋转位置，通过使用发射侧和接收侧上的阵列来最有效地收集数据;发射阵列的元件被顺序地发射，并且对于每个这样的发射，在接收阵列的所有元件上测量接收场。结果表明，这种测量方法是在傅里叶空间上展开的，从傅里叶空间中可以通过简单的傅里叶逆变换来恢复物体。这种成像策略是从Born和Rytov近似的非均匀介质中的传播方程推导出来的。提出的算法不需要在频域或空域进行插值，只需2N个FFT就可以重建一幅N × N图像。由于不进行任何插值，因此算法本身不会引入计算误差。与滤波反向传播算法的O（Ar 4）和基于频域内插的过程的O（NZ log N）相比，该过程的总计算复杂度为Ow 3）的量级。一些计算机模拟结果已被包括来证明该算法的数值精度。
A new tomographic imaging technique is presented that requires only two rotational positions of an object. Although ideally the angle between the two rotational positions should be 90°, theory predicts that valid results should be obtainable, albeit with reduced spatial resolution, even when this condition is not satisfied. For each rotational position of the object, the data is collected most efficiently by using arrays on both the transmit and the receive sides; the elements of the transmit array are fired sequentially, and for each such firing the received field is measured over all the elements of the receive array. It is shown thot this measurement strotegy firrs up the Fourier space, from which the object can be recovered by simple Fourier inversion. This imaging strategy was derived from the equations of propagation in an inhomogeneous medium with Born and Rytov approximations. A digital implementation is also presented of the proposed algorithm that requires no interpolations in either the frequency or the space domain, and can be carried out with only 2N FFT's for reconstructing an N X Nimage. Since no interpolations are carried out whatsoever, no computational errors are introduced by the algorithm itself. The total computational complexity of the procedure is of the order of Ow3) as compared to O(Ar4) for a filtered-backpropagation algorithm and O(NZ log N) for procedures based on interpolation in the frequency domain. Some computer simulation results have been included to demonstrate the numerical accuracy of the algorithm.