Curvature sensitive filter and its application in microfossil image characterisation

摘要

A new class of oriented, curvature sensitive filters is introduced. These filters provide a low-level detection facility for noisy curves without a prior edge extraction state, and are used here to characterise images of Carboniferous Foraminifers (a type of microfossil viewed in plane rock sections). A model for the performance of the filter in this application is described. This model uses a synthetic image with additive, spatially correlated, Gaussian noise to give predictions of sensitivity and specificity for the algorithm. These predictions are compared with results from actual images.