Fourier transforms of image-derived data: application to Albian coccoliths

摘要

The reduction of image-derived morphological data to a suite of characters that can be regarded as homologous is problematic. On the basis of analyses of hypothetical artificial morphologies, it is argued that the principal components of Fourier transforms of radially symmetrical structures are correlated with morphological attributes identified a priori as homologous. Application to grey-level images of Albian coccoliths yields results demonstrating that successful discrimination of genera and species is possible using an automatic retrieval and analysis system.