ALGORITHM FOR FINDING OPTIMAL VALUES OF PARAMETERS OF SINGULAR POINT DETECTORS FROM LOCAL CHARACTERISTICS OF THE IMAGE
A. V. Likhachev
Institute of Automation and Electrometry, Siberian Branch, Russian Academy of Sciences, Novosibirsk, Russia
Keywords: detectors of singular points, local image moments, optimal values of configurable parameters
Abstract
A method is proposed that allows determining the optimal values of the parameters p of the detectors of singular points from the local characteristics of the image µ. The method is based on the procedure for finding the extremum of the analytical dependence δ = P(p, µ), where δ is the criterion of detector efficiency, defined as the weighted sum of errors of the first and second kind. The dependence P(p, µ) interpolates a set δ(pi, µj) obtained from a sample chosen from a set of similar images. An experiment is performed with 40 radiographs of the lower part of the human spine, which are processed by the Harris-Stephens and SUSAN detectors. It is found that, when choosing the configurable detector parameters by the proposed method, the criterion δ turns out to be significantly lower than that for the values of these parameters at the boundaries of the recommended intervals.
|
