An improved restriction item is introduced into variational GAC vector-valued model on the basis of analysis on
color information and evolution characteristics of active contour curve. The proposed restriction item, which is a nonlinear
heat equation with normalized diffusion rate, is added to the level set function to maintain the signed distance function
properties fast and stably, and therefore the costly re-initialization procedure is completely eliminated. The algorithm reduces
the number of iterations and run time. In addition, more efficient discretization method with spatial rotation-invariance
gradient and divergence operator are proposed as numerical implementation scheme to improve strong robustness. The
experiment results show that the proposed algorithm is effective and has the ability to distinguish the fuzzy edges.