MTF estimation via BP neural networks and Markov model for space optical camera

The modulation transfer function (MTF) is one of the essential criteria of space optical camera. However, the traditional measurement methods of MTF are limited by precise equipment and test site. In this paper, a novel method is proposed to estimate the MTF of space optical camera via BP neural networks and Markov model. Utilizing this method, the MTF of space optical camera can be estimated only from the images taken by the camera without additional measurement equipment. The principle is to use the information extracted from known MTF images to train a BP artificial neural networks (ANN), and then use the BP ANN to estimate the MTF of space optical camera from remote images. In the meanwhile, the Markov model is used to correct the results estimated by ANN. The experiment results show that the MTF estimation average relative error at Nyquist frequency can further narrow to 5% via BP neural networks and Markov model, compared with 9% using only BP ANN.