Image segmentation is able to provides elements for enhancing a physical real-world environment. Although many existing segmentation methods have achieved impressive performances, they face problems where multiple similar objects are in close proximity to one another. We improve geodesic distance transform and define a symmetric morphology filter for segmentation. We embed shape prior knowledge into this geodesic distance transform filter. The proposed geodesic distance transform filter considers three factors simultaneously: the geometric distance, weighted gradients, and the distance to the boundary of the shape priors. As a result, it provides segmentation in line with the real shape of a particular kind of object. Positive results are demonstrated for several images and video sequences.
展开▼
