为了解决基于 Chan-Vese（CV）模型的传统水平集方法难以分割灰度不均匀红外图像的问题，本文提出一种基于改进 CV模型的水平集分割方法。通过加入可处理局部区域信息的局部项，使得改进的 CV模型能够有效避免不均匀背景对水平集演化过程的干扰。此外，通过加入符号距离能量惩罚项，使得该模型无需重新初始化过程，从而提高了水平集函数的演化效率。实验结果表明，本文方法对于红外图像的分割具有较高的精度。

To solve the problem that the traditional Chan-Vese(CV) model-based level set method was difficult to segment infrared images with inhomogeneous intensity, a kind of level set method based on improved CV model was proposed in this paper. By adding the local term which can deal with the local area information, the improved CV model can effectively avoid the interference of the inhomogeneous background to the level set evolution process. In addition, by adding the signed distance penalizing energy term, this model does not need to re-initialize the process, thus improving the evolution efficiency of the level set function. Experimental results show that this method has high precision for infrared image segmentation.