A fast based on mathematical morphology smoothing approach in level set methods

This paper presents a novel narrow band smoothing framework for level set methods. This framework is based on mathematical morphology operators. Previous methods such as straightforward ways and partial differential equation methods are available to smooth narrow band, but they are accompanied with extensive computational cost or a great deal of numerical iterations. The proposed scheme in this paper considers both smoothing accuracy and real-time application. Through a binary mirror image, complicated narrow band smoothing procedure is reduced to usual noise filtering, where all the unreasoned points in the grids correspond to pepper-salt noises. The filtering method also performs a global analysis on the front, and makes results more desirable. Furthermore, this paper also presents a new reconstruction approach. This approach can be naturally embedded into the proposed smoothing framework. Experimental results on several images show that this method has an excellent performance in terms of accuracy and velocity. The computation time also make it suitable for real-time application in active contour evolution.