Removal of aliasing effect on MTF measurement using bi-resolution images

摘要

One method to assess in-flight Modulation Transfer Function (MTF) relies on couples of images of the same landscape acquired with two spatial resolutions. The higher resolution image stands for the landscape and the ratio of the spectra gives the lower resolution instrument MTF. The main drawback of this method is the sensitivity to aliasing. This papers begins with a brief recall of the method putting the stress on the theoretical basis of the aliasing. The next step presents a way of aliasing correction when an MTF model is available. If it is not the case, the correction will be limited to the surrounding of the Nyquist frequency. This correction is applied to a test couple made with an airborne image of Toulouse and a simple simulation. The lower resolution image is obtained by the convolution of the airborne image with a large triangular function so that the ratio of the pixel sizes is equal to 10. The choice of 10 comes from the available couples of real images planned to be processed. For this test couple the MTF model is a simple sinc~2. The first results are quite noisy. The analysis of the magnitude and phase of the high resolution image spectrum shows that the landscape spectrum behaves like an 1/f function in the interesting low resolution frequency domain and the phase behaves like a Gaussian random variable. This is used to improve the correction and obtain good results. Correction is then applied on the available couples of images leading to SPOT4 HRVIR and VEGETATION MTF measurements.