A statistical approach to edge extraction in sampled images

The problem of edge extraction in sampled images is in formulated as a statistical estimation and detection problem. Known methods of statistical inference are applied to the problem, providing insight into the nature of edge extraction. Practical edge extraction algorithms are designed and applied to several images. Employing a formal approach to edge extraction, the fundamental requirements and objectives of edge extraction are defined, and a two-dimensional image model is formulated. Edges are characterized as boundaries between image regions having homogeneous statistical properties, and the locations of edges become parameters in the model.