The Local Binary Fitting (LBF) model based on local region information has its certain advantages in image segmentation of weak boundary or uneven greay. But, the segmentation results are very sensitive to the initial contours, and improper initial contour can directly lead to segmentation failure. Thus, a fast and stable LBF approach was proposed. First, after adding the area item with variable weights to the traditional LBF model, better initial contour could be obtained than manual one. Second, the traditional LBF model would be used for further segmentation. The experimental results show that, under the precondition of preferable results, this method can not only get promising segmentation results with flexible initial contour selection, but also faster than the traditional LBF model.