Tissue boundary refinement in magnetic resonance images using contour-based scale space matching.

An algorithm for computationally focusing the tissue boundaries detected from magnetic resonance images is presented. The proposed approach is a novel, whole-contour-based technique for tracing edges selected at a coarse scale into successively finer scales to recover the needed precision. The tracing algorithm builds consensus through a fast pixel voting scheme. Also presented is a rigorous method for determining the appropriate itinerary when traversing scale space, beginning from the premise of a maximum pixel migration per unit change in scale parameter. This leads to an efficient method of processing images so as to maximize accuracy and minimize noise. Although the LoG (Laplacian of Gaussian) is used for many of the experiments, results using a novel edge detector which is mathematically superior to and faster to compute than the LoG and for which fewer steps are required to traverse the same effective span in scale space are presented. Experimental results on real data are presented, and other potential applications are discussed.