Image segmentation using the level set and improved-variation smoothing

Traditional active contour models perform poorly on real images with inhomogeneous sub-regions. In order to overcome this limitation, this paper has proposed a novel segmentation algorithm. Firstly, analyzing the smoothing conditions for image segmentation, we construct a smoothing function with improved total variation. This function can smooth the inhomogeneous sub-regions, preserve the strong edges and enhance the weak edges. Then, the level set is employed to segment the smoothing component using the smoothing function. Lastly, according to the confidence level of segmentation sub-regions, we add a convergence condition to the smoothing to prevent the segmentation curve from vanishing. Experimental results indicate that this model is insensitive to noise and can deal with inhomogeneous intensity.