On-orbit Modulation Transfer Function (MTF) is an important indicator to evaluate the performance of the optical
remote sensors in a satellite. There are many methods to estimate MTF, such as pinhole method, slit method and so on.
Among them, knife-edge method is quite efficient, easy-to-use and recommended in ISO12233 standard for the wholefrequency
MTF curve acquisition. However, the accuracy of the algorithm is affected by Edge Spread Function (ESF)
fitting accuracy significantly, which limits the range of application. So in this paper, an optimized knife-edge method
using Powell algorithm is proposed to improve the ESF fitting precision. Fermi function model is the most popular ESF
fitting model, yet it is vulnerable to the initial values of the parameters. Considering the characteristics of simple and
fast convergence, Powell algorithm is applied to fit the accurate parameters adaptively with the insensitivity to the initial
parameters. Numerical simulation results reveal the accuracy and robustness of the optimized algorithm under different
SNR, edge direction and leaning angles conditions. Experimental results using images of the camera in ZY-3 satellite
show that this method is more accurate than the standard knife-edge method of ISO12233 in MTF estimation.