In the process of aviation remote sensing image restoration, the measurement of Modulation Transfer Function (MTF) of a system is very important. Because of many advantages, a knife-edge method becomes a relatively common method for measuring MTF. When using the knife-edge method to measure MTF of an infrared camera, the extraction and fitting of Edge Spread Function (ESF) are critical, which directly determines the effectiveness of subsequent MTF calculation. A new method for extracting edge data points is designed in an experiment. Since the method directly observes the gray-scale variation of a single pixel by moving the knife-edge, it is simpler and more intuitive than the traditional method which obtains ESF from remote sensing images. At the same time, the effectiveness of three fitting methods including linear fitting, Guassian fitting and Fermi fitting is compared. The results show that the Fermi fitting smoothes the noise at each end of ESF nicely and obtains the best result.