针对传统时域高通滤波校正算法存在的鬼影问题，提出一种新的基于混合高斯模型的红外图像自适应校正算法。新算法利用混合高斯模型对场景进行建模，只有在像元输出值满足一定条件的时候，才将其更新到校正系数中，实现有选择性地更新校正系数。通过一组仿真和真实的红外图像序列评价算法的性能,仿真图像采用峰值信噪比指标进行定量评价，新算法比传统时域高通滤波校正算法的峰值信噪比提高了约9 dB。真实图像采用主观的定性评价，传统算法校正结果中存在着明显的鬼影，而新算法校正结果中不存在鬼影。

A new self-adaptive calibration algorithm for infrared images based on Gaussian mixed model (GMM) was put forward according to traditional temporal highpass filter (THPF). GMM is applied in the new algorithm for background modeling, and only the pixels whose output values satisfy certain conditions can be updated to the correction coefficient, and the background is updated selectively so as to avoid the influence of the foreground target on the update of the background. The performance of the proposed algorithm was evaluated through infrared image sequences with simulated and real fixed-pattern noise, and the simulated sequences quantitative evaluation was carried out by using the peak signal noise ratio (PSNR). Results show that the PSNR of the new algorithm increases by about 9 dB compared with the THPF. The subjective qualitative evaluation was adopted for real images, the correction results in the traditional method show obvious ghosting artifact while the results of the new algorithm are in the absence of ghosting artifact.