Abstract:
This thesis proposes a novel theoretical framework, namely the Geodesic Active Region model, for dealing with the Image Segmentation problems in computer vision. This framework is based on the variational calculus and involves a curve-based objective function which integrates in a generic form both boundary and region based information. Moreover, the optimization process of this objective function is directly related with the curve propagation theory. Curve propagation is achieved using the level set theory, resulting in a powerful paradigm that is independent of the initial curve configuration.
The proposed Geodesic Active Region model has been employed for providing solutions to important practical problems in Ultrasound Imaging, namely the tasks of Unsupervised Image segmentation of ultrasound images.
More specifically, in the image segmentation case, the proposed approach determines automatically the number and the intensity properties of the image regions and then obtains the edges of the desired region using the Canny edge detector and lastly performs the segmentation without being constrained by the initial conditions
Several experimental results for each of the above tasks are supplied, demonstrating the effectiveness of the proposed solutions and indicating the potentials of the Geodesic Active Region model.
