Good Effect of GVF Snake Segmentation By Good Parameters Settings

The basic theory and the key algorithm about GVF snake model are presented firstly. Then a lot of important experiments are done to achieve the effect of the GVF snake segmentation about stomach epidermis tumors after having set the values of some important parameters in GVF snake model. Firstly, the effect of two kinds image pretreatment processing such as the edge mapping and the Gauss blur are analyzed, and the conclusion is found that the edge mapping must be run before the GVF snake segmentation. Otherwise the snake can only converge to one point instead of the edge of the cells. Moreover, good image segmentation result will be attained if the Gauss blur which restrains the noise of an image has been run before the edge mapping. Flexible parameter alpha and rigid parameter beta are also analyzed by experiments, which prove that the snake model cannot get good convergence result if alpha is less than 1 and much smaller areas may be got if alpha is higher than 1. Furthermore, parameter beta doesn´t work in any experiments. At the same time, iteration times of GVF snake are analyzed too. The segmentation result becomes stable if the iteration times is bigger than 80 in these experiments. Finally, it is proved that good segmentation results can be got if mu is less than 0.3.