Curvature of lines in digitized images
Abstract
A shape descriptor has been developed for linear objects, which may be used to describe the deviation of a curved line from a straight one. The features required for this parameter are high invariance for orientation, translation, and scale, particularly in the discrete 2-dimensional space of digitized images. To that end, the impact of the single raster points of the image (pixels) on the computed shape parameter must be reduced. Only a robust but sensitive measurement of curvature can be successfully used to correlate this parameter with other properties belonging to linear objects. As a rule, commercially available image analyzers offer some formfactors, which do not accomplish this function well enough. A possible field of application is the study of cortical infolding (gyration) in relation to cytoarchitectonics in neuroanatomy.