next up previous
Next: Edges Up: Image Segmentation and Object Tracking Previous: Segmentation Based on Regularization

Segmentation Based on Globally Organized Clustering

In [117] Verri et. al. look for regions of uniform flow or uniform expansion, rotation or shear. Optic flow is found using a method derived from the BCCE. This is used to segment the image into different objects. The work is extended in [86] where global segmentation is used. Here each small patch in the image has its centre of expansion/contraction, centre of rotation and direction of constant flow calculated. (Zero shear is assumed.) These three positions are drawn onto three global maps, and strong clusters are looked for in each map. Next each image patch is assigned to a cluster found in one of the three graphs. The results are smoothed using deterministic relaxation, providing as output the segmented image. The results presented do not give very accurate motion boundaries.



next up previous
Next: Edges Up: Image Segmentation and Object Tracking Previous: Segmentation Based on Regularization



© 1997 Stephen M Smith. LaTeX2HTML conversion by Steve Smith (steve@fmrib.ox.ac.uk)