DGFO-cGAN: A Dual-Generators Framework-Based Fusion Optimization cGAN for SAR-to-Optical Image Translation
TL;DR
Experimental results demonstrate that DGFO-cGAN outperforms 11 mainstream image translation methods, achieving notably higher structural similarity index measure and learned perceptual image patch similarity, and generating optical images with clearer structures and richer details.
Abstract
Existing generative adversarial network-based methods for SAR-to-optical image translation typically utilize a single-branch generator architecture, which makes it challenging to simultaneously ensure the authenticity of local details and the consistency of global information, often resulting in generated images suffering from the loss of land cover information and reduced edge clarity. To address this limitation, this article proposes a dual-generator framework-based fusion optimization cGAN (DGFO-cGAN) for SAR-to-optical image translation. The method consists of a detail-preserving subnetwork and a holistic-balanced subnetwork, whose outputs are fused through pixel-level weighted optimization to achieve high fidelity in both local texture and global consistency. Experimental results on three public datasets demonstrate that DGFO-cGAN outperforms 11 mainstream image translation methods, achieving notably higher structural similarity index measure and learned perceptual image patch similarity, and generating optical images with clearer structures and richer details.