Published online by Cambridge University Press: 08 August 2003
In this article we present a method to identify and extract spherical particle images from noisy spot-scan and flood beam electron microscopy images. We use a template matching algorithm with additional image preprocessing operations to allow consistent particle selection in spot-scan and other highly spatially varying images. In addition, this algorithm incorporates an automated image-cutting and edge-sewing mechanism that allows efficient particle selection despite the large size of electron microscopy images. We have tested this template matching algorithm on various spherical virus particle images with a large range of defocus values and have found that the particles are consistently selected in an accurately centered manner. In addition, this method is able to extract spherical virus particles from 400-kV electron microscopy images with defocus values of less than 1.0 μm underfocus where the particles are not readily visible to the human eye.