Linear and Nonlinear Multiscale Filtering
In this chapter, we analyse the main theories of image multiscale filtering (and subsequent multiscale edge detection). The main point of the doctrine is that “edge detection” is a kind of differentiation; now, the image must be filtered (and smoothed) before differentiation. Thus “edge detection” becomes multiscale, like the filtering, and any edge detection theory is directly deduced from a multiscale filtering theory. We show how variational formulations are naturally associated with the main multiscale filtering theories.
KeywordsHeat Equation Anisotropic Diffusion Multiscale Analysis Curvature Motion Scale Space Theory
Unable to display preview. Download preview PDF.