A Novel SAR Sidelobe Suppression Method Based on CNN
Sen Yuan, Ze Yu, Chunsheng Li, Shusen Wang
Abstract
Sidelobe suppression is a basic but important task for synthetic aperture radar (SAR) images, since the existing sidelobes can reduce the image quality and complicate image interpretation. The main task of sidelobe suppression is to suppress the sidelobes effectively while maintaining high image resolution and the mainlobe's energy. However, the most effective method, robust spatially variant apodization (RSVA), still faces the problem of energy loss on cluttered targets caused by phase factors. This letter proposes an SAR sidelobe suppression method based on the convolutional neural network (CNN) to compensate for the energy loss in RSVA. The experiments on the SAR images show better performance on both sidelobe suppression evaluation metrics and signal-to-noise ratio (SNR) preservation compared with conventional methods.