According to the characteristics of OCT images for macula edema, we studied a method for segmentation of the macula edema. Based on the Chan-Vese model, we proposed an improved level-set algorithm. With defining the integer-valued signed function directly, the curve could evolute outward or inward by changing the inside neighboring rid points and outside neighboring grid points into each other. We realized image segmentation which is much faster than the method of Chan-Vese model and smoothness regularization. We segmented 45 images and extracted the macula edema of each image. After achieving good segmentation results, we estimated the volume of the macular edema. The method provides quantitative analytic tools for clinical diagnosis and therapy.