Blind deconvolution for noise reduction in satellite imaging

Abstract

Due to launching ‎vibrations and space harsh environment, high resolution remote sensing satellite imaging systems require permanent assessment and control of image quality, which may vary between ground pre-launch measurements, ‎after launch and over satellite lifetime. In order to mitigate noise, remove artifacts and enhance image interpretability, the Point Spread Function PSF of the imaging system is estimated. Image deconvolution can be performed across the characterization of the actual Modulation Transfer Function MTF of the imaging system. In this work we focus on adapting and applying a no reference method to characterize in orbit high resolution satellite images in terms of geometrical performance. Moreover, we use natural details contained in images as edges transitions to estimate the impulse response via the assessment of the MTF image. The obtained results are encouraging and promising. This method allows us to exploit the information contained in the image without prior knowledge of the instrument or the satellite which recorded the image. We applied this method to estimate MTF of an optical high resolution satellite. In this method, the MTF profile results from the Fourier transform of LSF, which is itself derived from ESF. The quality of the ESF estimation impacts mainly the MTF rebuilding. We have implemented an algorithm through several stages of calculations to estimate MTF of an image in such a way to be reproducible and robust to noise.