ABSTRACT
This chapter presents the use of B-spline level-set segmentation model in medical image segmentation. Compared to the traditional implicit presentation, the use of B-splines as a basis for the level-set representation provides some specific additional benefits, such as fast convergence and the intrinsic smoothing contour to the segmentation solution. The chapter also presents a new shape prior-based implicit level-set model for image segmentation based on linear combination of B-spline basic functions. Splines are piecewise polynomials with pieces that are smoothly connected together. The joining points of the polynomials are called knots. Traditionally, the B-spline interpolation problem has been approached by setting up and solving a band-diagonal system of equations. In contrast with the classical variational approaches, the minimization of the model is respect to the B-spline coefficients. Different from the general partial differential equations based alignment, one can takes into account the affine transformation to compute the pose transformation utilizing the theory of moment invariants and shape normalization.
