68 / 2024-08-14 18:22:37

Spatially Variant Convolutional Neural Network for Sidelobe Suppression on SAR Images

sidelobe suppression,spatially variant filtering,synthetic aperture radar (SAR) interferometry,deep learning

The superior signal-to-noise ratio (SNR) and high resolution of synthetic aperture radar (SAR) images are originated from the pulse compression technology [1]. The corresponding impulse response presents the shape of sinc function along azimuth and range. The mainlobe is 13.26 dB higher than the nearest sidelobe in a uniform sinc function and the amplitude of the sidelobe decreases with octave [2]. Since the dynamic range of SAR images can reach 50 dB or even higher [3], the sidelobes originated from strong targets will overshadow weak targets around. This problem degrades the image quality and interferes the image interpretation seriously, so sidelobe suppression is a meaningful mission.