Joint Noise Suppression and Resolution Enhancement of ISAR Images Using Integrated Neural Networks

This paper proposes an integrated neural network for joint noise suppression and resolution enhancement of inverse synthetic aperture radar (ISAR) images. Unlike conventional methods that address both challenges separately, we present a unified framework that can address them simultaneously. To achieve this, we first generate a comprehensive dataset of ISAR images for various targets under different conditions using a simulation-based method. Subsequently, we develop separate generative models for noise suppression and resolution enhancement, which are then combined sequentially. This combined network uses a joint optimization strategy in training process, simultaneously updating the weights of the two networks. The proposed integrated network achieved an average peak signal-to-noise ratio and structural similarity index measure of 34.69 dB and 0.95, respectively. It demonstrates that the proposed network effectively achieves both noise suppression and resolution enhancement within a single network.