How to measure the perimeter from digital images

Abstract

Image analysis aims at extracting quantitative information from digital images. As for any measurement method, the measures obtained from image analysis present some measurement errors that should be better considered during interpretation of results. This work investigates the quality of measures obtained from image analysis based on measurements on synthetic shapes with various geometries, sizes, positions and orientations, by focusing on the methods available in common image analysis software. The perimeter is used as illustration feature. The results provide some guidelines for choosing the most appropriate methods, and insights about the relative errors that can be expected.