针对Chan-Vese水平集方法进行全域演化实现多相目标物体收敛时计算量大、演化速度极慢，而采用初始区域内进行轮廓曲线演化时又不能实现目标内部第三相区域的分割，提出二轮递进演化和演化背景灰度值转化的改进Chan-Vese方法实现目标内部第三相区域分割。改进方法能够对Chan-Vese方法演化多相拓扑不成功的情况实现目标内部第三相区域的检测。实验结果表明，改进的方法能快速稳定实现目标物体内部第三相区域分割。

To improve the use of the Chan-Vese method in the segmentation of the three-phase and to reduce its computational complexity, an improved Chan-Vese method is used to realize infrared image segmentation. The improved Chan-Vese method uses progressive processing and the reversion of gray value to achieve Chan-Vese multi-phase object segmentation. The improved Chan-Vese method can realize the three-phase infrared image segmentation. The experiment result demonstrated the efficiency and effectiveness of the improved method.