Perimeter in material images: Comparison of continuous representation and global stereological estimation methods

A set of objects in a material structure is typically represented by a binary digital image. Two methods of estimation of the perimeter of objects are discussed: 1. “Searching for reality”: a continuous regression representation of the borders between objects and background is estimated, from which the desired characteristic is computed. 2. Stereological solution: easily measurable qualities of the digital set are measured, from which the desired characteristics are estimated on the basis of a known global formula. Both approaches are compared in an application on sets of digital images representing original patterns with successively lowered resolution.