Particle image identification, classification and correction techniques in digital holographic microscopy

This thesis aimed to investigate digital methods for identification, classification, and correction of particle images using digital holographic microscopy. The work begins with a micro-scale flow experiment employing a digital off-axis holographic microscope. To overcome noise from out-of-focus particle images due to the increase in the particle concentration, a new particle image identification algorithm with threedimensional reconstruction was introduced. It shows that individual particle images could be identified at about 62% of the expected number of particles. However, a cylindrical micro-channel used in the experiment was found as the main source of astigmatism. Following this, an automatic image classification using neural network was proposed to classify reliable and astigmatic particle images. A feed-forward backpropagation neural network with two class classifier was trained, achieving overall accuracy of 99.8%. Next, an original method of aberration correction using a priori information and digital wavefront aberration processing was applied. Astigmatism introduced by the micro-channel was modelled according to quadratic phase function and optimized using peak detection algorithm. The results show that astigmatism in the detected particle images was effectively compensated. A variant digital off-axis holographic microscope was later developed for large-scale flow measurement, allowing for the first time both amplitude and phase of the holographically reconstructed images to be registered simultaneously. The aberration introduced by a tilt in the optical hologram during the reconstruction was effectively corrected using adaptive optics. In overall, the work discussed in the thesis has proven effective to overcome noise and aberration. The main implication is that the holographic techniques can be successfully employed in complex three-dimensional fluid flow measurements.