Extension of the virtual electric field model using bilateral-like filter for active contours

Abstract

The gradient vector flow (GVF) model has been proven as an effective external force for active-contour-based image segmentation. However, it suffers from high computation cost since there are two PDEs to be solved in an iterative manner. As a remedy, the virtual electric field (VEF) model is proposed, which can be implemented in real time using the fast Fourier transform. However, the VEF model cannot preserve weak edges since it employs linear kernels. In this work, we extend the VEF model by using bilateral-like filters, and a fast algorithm is also employed for the proposed model. The proposed model is referred to as bilateral-filter-based VEF (BVEF) model. Experimental results on synthetic and real images demonstrate that the BVEF snake possesses some desired properties of the GVF, CNGGVF and VEF snakes such as large capture range and concavity convergence, and the BVEF model can be implemented in near real time, and its computation cost is comparable to that of the VEF model and much shorter than that of the GVF and CNGGVF models; it also can preserve weak edges, thanks to the bilateral-like nonlinear kernels.