The body infrared radiation characteristics are related to physiological activities such as respiration. We developed a novel approach to remotely monitor respiration by using the time-series infrared thermography of face. To improve the accuracy and robustness， the algorithm is researched based on Harris corner detection and optical flow field tracking method， which enhance the accuracy of the nostril region tracking. We demonstrate the approach by monitoring the respiratory of human in different environment and evaluating its time-frequency characteristics and statistical parameters. The results show that the approach can reliably monitor the real-time respiration and its pattern. That would be helpful for the clinical application of the infrared thermography on monitoring physiological status.