Boundary surface detection in 3D biomedical images based on local gradient thresholds

This paper proposes a method to detect the boundary surface with continuously changed gradient magnitudes from 3D biomedical images. Instead of selecting a constant gradient threshold, which sometimes cannot correctly separate such boundary surfaces from 3D images, the proposed method adaptively computes different local gradient thresholds in different local regions and uses them to detect different boundary surface patches. Using such scheme, the proposed method not only can well detect those boundary surfaces with uniformly high gradient magnitudes, but also can effectively detect the boundary surfaces with continuously varied gradient magnitudes from low to high. The performance of the method is illustrated by many examples from different 3D biomedical images.