A Comparison of Edge-Based and Region-Based Segmentation Performed with PDEs

Abstract

The paper discusses the level set segmentation method, which employs the solution of the partial differential equation (PDE) describing the multidimensional function whose zero slice in plane xy determines the contour of the object. One of the possible approaches to the contour evolution is using the image edges; the edges define the boundaries of the segmented objects. A major disadvantage of this procedure consists in the often unclear or noisy edges, a condition that may lead to incomplete segmentation. Such a drawback related to the curve evolution is solvable through utilizing regions with similar brightness levels; consequently, the actual segmentation cannot be affected by unclear transitions between the objects and the background. The final comparison then shows that both methods find specific application within the segmentation process.