Sub-Diffraction and Sub-Pixel Microscopic Deconvolution

描述(申请人提供)：该项目的目标是使用专门为恢复亚像素级图像而设计的去卷积算法来提高光学光学显微镜的有效分辨率。活细胞的光学显微镜(LM)通常被限制在最高分辨率约200纳米，分辨率的提高通常被认为需要硬件修改，如结构照明。亚像素去卷积软件旨在为生命科学研究人员提供一个机会，利用他们现有的仪器，以低成本提高他们3D样本中精细结构的分辨率。亚像素算法基于光子受限数据的最大似然去卷积和解析延拓。特别重要的是开发算法加速技术，以减少处理要求并使软件具有商业吸引力。将研究建立摄像机像素尺寸和噪声影响的模型的方法。噪声抑制方法对于提高算法在弱光成像情况下的稳健性，以及在防止不需要的伪影的同时保留精细特征具有重要意义。反卷积将使用相位恢复方法，直接从观测数据估计显微镜出瞳处的波前误差。这种盲解卷积的方法将提高算法的能力，使点扩散函数适应物镜规格中的细微像差和公差。此外，亚像素算法将使使用芯片上相机入库或大像素的欠采样图像以及间距超过奈奎斯特限制的光学部分能够得到正确处理。这些算法将被扩展，以兼容广域荧光、透射光亮场、旋转圆盘共焦和激光扫描光焦模式。性能将使用制造的测试目标和已知结构的生物样本进行验证。

DESCRIPTION (provided by applicant): The objective of this project is to increase the effective resolution of optical light microscopes using deconvolution algorithms specifically designed for restoring imagery at a sub-pixel level. Light microscopy (LM) of living cells is typically diffraction limited to a maximum resolution of approximately 200nm, and resolution improvements are often considered to require hardware modifications such as structured illumination. The sub-pixel deconvolution software is intended to give life-science researchers an opportunity to increase the resolvability of fine structures within their 3D specimens, using their existing instrumentation and at a low cost. The sub-pixel algorithm is based on maximum-likelihood deconvolution and analytic continuation of photon-limited data. Particular importance is placed on developing algorithm acceleration techniques to reduce processing requirements and make the software commercially attractive. Methods will be investigated that model the effect of the camera pixel dimension and noise. Noise suppression methods are important to improve algorithm robustness in low-light imaging situations, and to retain fine features while preventing unwanted artifacts. The deconvolution will employ phase retrieval methods to estimate the wavefront error at the exit pupil of the microscope directly from the observed data. This approach to blind deconvolution will improve the ability of the algorithm to adapt the point spread function to subtle aberrations and tolerances in the objective lens specifications. Additionally, the sub-pixel algorithm will enable under-sampled imagery using on-chip camera binning or large pixels, and optical sections that are spaced beyond the Nyquist limit, to be correctly processed. The algorithms will be extended to be compatible with widefield fluorescence, transmitted light brightfield, spinning disk confocal and laser scanning ponfocal modalities. Performance will be verified using manufactured test targets and biological specimens with known structures.