I am looking for an algorithm that can detect saddle points in a 2-D image represented as a gray-scale topology. I am familiar with the Hessian matrix approach to locating saddle points. However, this approach leads to missed detections -- looking for a different approach. Any ideas?
replied to: Vision424
Replied to: I am looking for an algorithm that can detect saddle points...
See V. Robins, P.J. Wood, and A.P. Sheppard
Theory and algorithms for constructing discrete Morse complexes from grayscale digital images IEEE Transactions on pattern analysis and machine intelligence (2010) Accepted for publication 21 January 2010.
This algorithm will extract the whole morse complex for the image, as well as any saddle points. It also works in three dimensions. The trick is to think of the image as a "cubical complex" made up of points, edges and squares (and cubes when the image is three dimensional). The most important saddle points can be identified with topological persistence.
