A Gradient-Based Probabilistic Method for Image Feature Extraction
Thanh Ha Le, Norbert W. Schuff · 2014
Image shape feature extraction by locating the exact shape boundaries has been applied in numerous research areas such as object tracking, content based image and video retrieval, robotics and biomedical imaging. Deformable active contour (snake) methods have been widely used. However, snake methods have limitations in requirement of manually initialized contour, slow convergence, random curve movement in case of missing energy forces and noise sensitivity. We develop a probabilistic model using gradient vector flow field for identifying contour curves and applications in brain MRI feature extraction. Our algorithm method performed better than popular snake-based algorithms on the simulated images and brain MR images.