Summary: Traditional edge detection methods can not perform well with noise resisting and edge continuity maintaining simultaneously. An improved color edge detection method is developed based on the Hermite technique and multipass processing. Edge discontinuities are reduced by a multi-scale Hermite technique based on color gradients in the image. Thus the algorithm provides better noise resisting and edge continuity maintaining simultaneously. Simulations show that compared with existing methods, the edges detected by this algorithm are smoother with less noise.