A new scheme for femur shape recovery from volumetric images using deformable models was proposed. First, prior 3 D deformable femur models are created as templates using point distribution models technology. Second, active contour models are employed to segment the magnetic resonance imaging (MRI) volumetric images of the tibial and femoral joints and the deformable models are initialized based on the segmentation results. Finally, the objective function is minimized to give the optimal results constraining the surface of shapes.
A genetic learning algorithm based fuzzy neural network was proposed for noisy image restoration, which can adaptively find and extract the fuzzy rules contained in noise. It can efficiently remove image noise and preserve the detail image information as much as possible. The experimental results show that the proposed approach is able to performa far better than conventional noise removing techniques.
