PARAMETRIC BLIND DECONVOLUTION OF MICROSCOPIC IMAGES

共焦扫描显微镜(CSM)现在被广泛应用于三个方面- 固定样本的三维(3D)可视化。然而，活着 样品经常被高强度激光损坏和/或漂白 CSM所需的激励光。一种常见的解决方案是打开 微微共焦光圈，从而允许更多的光到达 探测器。然而，更大的共焦光圈也可以接受更多的离焦 聚焦光线并降低CSM的深度分辨率。迷失的人 分辨率可以使用计算反卷积算法来恢复。 即使是严格的共焦显微镜也受益于 去卷积算法提供的分辨率。大多数调制解调器CSM 配备一台可以数字化3D共焦图像堆叠的主机。 尽管有计算机，共焦图像的去卷积 还没有被广泛使用。其中一个关键原因是，大多数 反卷积算法需要精确的点扩展知识。 描述显微镜的功能(PSF)。在现代的CSM中， PSF的测量或计算充其量也是困难的。在这些 情况下，有必要估计PSF和标本荧光 从图像同时分布，一种称为盲的方法 反褶积(BD)。目前用于BD的方法速度很慢，因为它们 逐点估算PSF。 提出的研究的长期目标是提供更快和更多的 稳健的计算反卷积算法，不需要 测量或计算显微镜的PSF。 为了实现这一目标，使用了PSF的数学-物理模型 这取决于少量的未知参数，然后进行估计 将派生方法以同时获取这些参数的值 与标本的荧光分布相吻合。虽然长期目标是 是BD的部分共焦图像，非共焦显微镜用户会 也受益于这项研究的结果。

Confocal scanning microscopes (CSM) are now widely used for three- dimensional (3D) visualization of fixed specimens. However, living specimens are often damaged and/or bleached by the high intensity laser excitation light required by the CSM. A common solution is to open the confocal aperture slightly and thus allow more light to reach the detector. A larger confocal aperture, however, accepts also more out-of- focus light and degrades the depth resolution of the CSM. The lost resolution can be recovered using computational deconvolution algorithms. Even the strictly confocal microscope benefits from the increased resolution that deconvolution algorithms provide. Most of the modem CSM's come with a host computer that digitizes 3D stacks of confocal images. Despite the availability of a computer, deconvolution of confocal images has not been widely used. One of the key reasons is that most deconvolution algorithms require accurate knowledge of the point-spread- function (PSF) that describes the microscope. In modern CSM's, the measurement or computation of the PSF is difficult at best. In these cases, it is necessary to estimate the PSF and the specimen fluorescence distribution simultaneously from the image, an approach called blind deconvolution (BD). Current methods for the BD are slow because they estimate the PSF point by point. The long term goal of the research proposed is to provide faster and more robust computational deconvolution algorithms that do not require measuring or computing the microscope's PSF. To achieve this goal, a mathematical-physical model for the PSF is used that depends on a small number of unknown parameters, then estimation methods will be derived to obtain the values of these parameters together with the specimen fluorescent distribution. Although the long term goal is BD of partially confocal images, nonconfocal microscope users will also benefit from the results of this research.