In the past decade there has been significant interest in image processing for brightfield cell microscopy. Much of the previous research on image processing for microscopy has focused on fluorescence microscopy, including cell counting, cell tracking, cell segmentation and autofocusing. Fluorescence microscopy provides functional image information that involves the use of labels in the form of chemical stains or dyes. For some applications, where the biochemical integrity of the cell is required to remain unchanged so that sensitive chemical testing can later be applied, it is necessary to avoid staining. For this reason the challenge of processing images of unstained cells has become a topic of increasing attention. These cells are often effectively transparent and appear to have a homogenous intensity profile when they are in focus. Bright field microscopy is the most universally available and most widely used form of optical microscopy and for this reason we are interested in investigating image processing of unstained cells recorded using a standard bright field microscope. In this paper we investigate the application of a range of different autofocus metrics applied to unstained bladder cancer cell lines using a standard inverted bright field microscope with microscope objectives that have high magnification and numerical aperture. We present a number of conclusions on the optimum metrics and the manner in which they should be applied for this application.
Digital holographic microscopy is suitable for the detection of microbial particles in a rapidly flowing fluid since in this technique the focusing can be carried out as post-processing of a single captured image. This image, known as a digital hologram, contains the full complex wave front information emanating from the object which forms an interference pattern with a known reference beam. Post-processing is computationally intense and it constitutes a bottleneck for real time inspection of fast moving scenes. In the current work, GPU computation is used to accelerate the post-processing of the holographic images captured by digital holographic microscopy. Efficiency and reliability of a pre-processing step in order to eliminate low information content holographic images is also investigated.