An Improved Model for Image Selective Segmentation Based on Local and Global Statistics
Shurong Li, Yu Meng, Zhang Minyi · 2019
In this paper, an improved selective model is proposed for image segmentation, which is built based on local and global statistics of the image. Firstly, by utilizing the advantage of CV model which is less sensitive to original location of initial contour and can segment noisy images, and taking advantage of LCV which can efficiently solve intensity inhomogeneity, the new model integrate the local image information from local Chan-Vese (LCV) model and global statistics from CV model into selective segmentation. In addition, a weight coefficient is added to balance the relationship between global energy and local energy. Then extending the improved model to the color images. Finally, simulation experiment results show that our model can efficiently segment noisy and intensity inhomogeneity images, and is less sensitive to the initialization. New model has stronger robustness and is more efficient.