Abstract

We are studying the application of digital image processing to confocal-scanning-microscope images to achieve axial resolution comparable with the lateral diffraction limit. Previously, we have studied the trade-off between signal-to-noise ratio (SNR) and axial resolution in unprocessed images as the size of the confocal aperture (pinhole) is varied. In this partially confocal regime, a small pinhole yields good axial resolution at the expense of reduced SNR, and a large pinhole improves the SNR but degrades axial resolution. We reconstructed three-dimensional, partially confocal images by means of Jansson-van Cittert (JvC) and maximum-likelihood (ML) algorithms. Because, in the limit, both methods converge to inverse filtering, they are somewhat sensitive to noise. As the iterations progress the axial resolution improves, but at the same time the SNR decreases. We present a quantitative study of the microscope-algorithm combination, which considers the posterior axial resolution and image SNR versus the pinhole size and the number of iterations of the reconstruction algorithm. By treating the microscope and processing as a system, optimal combinations may be determined.

© 1990 Optical Society of America