Advanced Optical Imaging Group, School of Physics, University College Dublin, Dublin, Ireland.
Ophthalmic Physiol Opt. 2013 Jul;33(4):434-43. doi: 10.1111/opo.12076.
The presence of photon noise and readout noise can lead to centroiding errors in a Hartmann Shack wavefront sensor (HS) and hence limit the accuracy of wavefront reconstruction. The aim of this paper is to compare, via Monte Carlo simulations, the accuracy of various centroiding methods in detecting noisy focal spot patterns of the HS while sensing ocular aberrations of myopic eyes.
Myopic ocular aberrations were randomly simulated by using the modal statistics obtained from the measurements of 41 myopic subjects. HS spot patterns were simulated using a fast Fourier method where photon noise and readout noise were added using appropriate statistics. Adopting five different centroiding techniques: (1) centre of gravity, (2) weighted centre of gravity, (3) intensity weighted centroiding, (4) iteratively weighted centre of gravity and (5) matched filter based centroiding along with a zonal based wavefront sensing approach; the wavefronts were estimated and compared, by calculating the root mean square (RMS) wavefront error, with the initially simulated wavefront. The magnitude of readout noise was varied in terms of the maximum number of photons and electrons per subaperture per frame. The RMS error was calculated for each of the centroiding algorithms.
For higher magnitude of readout noise and lesser number of photons per subaperture per frame (n), matched filter, iteratively weighted centre of gravity and intensity weighted centroiding outperform centre of gravity and weighted centre of gravity methods, for an appropriately chosen focal length and subaperture pitch. The plots of RMS error as a function of 'n' show that for lower amplitude of readout noise, computationally efficient centre of gravity and intensity weighted centroiding methods can be safely adopted to obtain high enough reconstruction accuracy. Also, even at greater readout noise levels, for a large enough 'n', intensity weighted centroiding is enough to sense the aberrations with high accuracy. It is shown that the wavefront sensing accuracy depends on the size of the spots and bit resolution of the camera.
Five different centroid detection methods used in a HS in the presence of photon noise and readout noise were analysed in the context of sensing ocular aberrations of myopic subjects and identify cases under which each of these methods is appropriate.
光子噪声和读出噪声的存在会导致哈特曼夏克波前传感器（HS）的质心定位误差，从而限制波前重建的准确性。本文的目的是通过蒙特卡罗模拟比较各种质心定位方法在检测 HS 噪声焦点模式时的准确性，同时感知近视眼睛的像差。
使用从 41 名近视受试者的测量中获得的模态统计数据，随机模拟近视眼像差。使用快速傅里叶方法模拟 HS 光斑模式，其中使用适当的统计数据添加光子噪声和读出噪声。采用五种不同的质心定位技术：（1）质心，（2）加权质心，（3）强度加权质心，（4）迭代加权质心和（5）基于匹配滤波器的质心以及基于区域的波前感测方法；通过计算均方根（RMS）波前误差，将波前与初始模拟波前进行比较，从而估计和比较波前。以每个子孔径每帧的最大光子数和电子数的形式来改变读出噪声的幅度。为每个质心定位算法计算 RMS 误差。
对于较大的读出噪声幅度和每帧每个子孔径的较少光子数（n），匹配滤波器、迭代加权质心和强度加权质心优于质心和加权质心方法，对于适当选择的焦距和子孔径间距。RMS 误差作为“n”函数的图表明，对于较低的读出噪声幅度，可以安全地采用计算效率高的质心和强度加权质心方法来获得足够高的重建精度。此外，即使在较大的读出噪声水平下，对于足够大的“n”，强度加权质心足以高精度地感知像差。结果表明，波前感测精度取决于光斑的大小和相机的位分辨率。
在存在光子噪声和读出噪声的情况下，在 HS 中使用了五种不同的质心检测方法，在感知近视受试者的像差方面进行了分析，并确定了每种方法适用的情况。
