针对复杂背景下，红外目标与背景的亮度或对比度发生变化等因素会影响目标的分割结果的问题，构造了一种基于边缘相位信息的分割算法。该算法首先利用系数间积(ICP)变换获取图像边缘的相位信息，并进行边缘提取；再根据边缘像素点的相位与幅值，确定后续区域生长的种子与判决条件；最后采用区域生长思想对整幅图像实施分割。仿真结果表明，该算法不受亮度和对比度变化影响，准确有效地实现红外图像分割，避免了传统的分割算法所导致的图像边缘信息丢失。

In complex background scene, the variation of brightness or contrast may affect the performance of image segmentation.To deal with this problem, a new algorithm for image segmentation was proposed based on phase information of geometric edges.The algorithm first extracted phase information of edge using Inter-Coefficient Product (ICP) transform, and obtained the image edge using the phase information.Then, according to the phase information and amplitude, a pixel was chosen as the seed of growing process, and the decision criterion was established.Finally, the whole image was segmented by using region growing.The simulation results show that:the proposed method is not influenced by the variation of brightness or contrast, and can implement infrared image segmentation accurately and effectively.The algorithm also helps to avoid missing the edge information of image.