Snakes Active Contour Models MICHAEL KASS International journal

Snakes : Active Contour Models MICHAEL KASS International journal of computer vision

Abstract • A snake: an energy-minimizing spline guided by external constraint forces and influenced by image forces • A unified account of a number of visual problems: detection of edges, lines and subjective contours; motion tracking; and stereo matching • Interactive interpretation

Introduction • image information ->The surface evolution ->lock onto nearby edges • segmentation problems ->Optimization problems • energy minimization->energy functions • Design principles: advantageous properties ->Energy reduction • Internal forces&Image forces&external constraint forces

Methodology • The traditional method of contour extraction: gradient->edge points->contour • Snake: continuous and smooth closed contour>near the features of interest->gradient>ultimate location • uniqueness

mathematical model V(s): represent the position of a snake

• V’(s), v’’(s) • Eint: the internal energy of the spline due to bending • α(s) : makes the snake act like a membrane “ 应力 ” • β(s): makes it act like a thin plate “刚度” • noise • Contour convergence performance • Proper location

ext • Eext: external energy function ->moving direction • Eimg: image forces ->edge->minimization>accuration • Econstraint: external constraint forces, specific problems

principle of operation • Snake: parameterized curve • Internal forces: Shape characteristics of contour curve • External forces: behavior • Energy function ->minimize