Title: A FAST ALGORITHM FOR SIMULTANEOUS SEGMENTATION AND REGULARIZATION OF IMAGES
Gunay Dogan*, Ricardo H. Nochetto, Pedro Morin
*Affiliation: Mathematical and Computational Sciences Division,
NIST Gaithersburg, MD 20899.
We introduce a novel computational method for a Mumford-Shah functional, which decomposes a given image into smooth regions separated by closed curves. Our goal is to distinguish the foreground objects on a background and regularize the image in a way that respects the object boundaries. We cast this problem as a shape optimization problem, and develop a gradient descent approach at the continuous level that yields non-linear PDE flows. We propose time discretizations that linearize the problem, and space discretization by continuous piecewise linear finite elements. Our method incorporates topological changes, such as splitting and merging of curves for detection of multiple objects, space-time adaptivity and a coarse-to-fine approach to process large images efficiently. We present several simulations that illustrate the performance of the method, and investigate the model sensitivity to various parameters.
Reference: G. Dogan, R.H. Nochetto, P. Morin, SIAM Journal on Scientific Computing, 30, 6, 3028-3049 (2008)
Mentors Name:Stephen A. Langer
Mathematical and Computational Sciences Division
Information Technology Laboratory
Building 225, Mail stop 891
Tel: (301) 975-8883
Fax: (301) 975-3800
Is your mentor a Sigma Xi Member? No