Cloud fusion, the process of combining optical and Synthetic Aperture Radar (SAR) images, is a challenging task due to the different imaging modalities and resolutions. Cloud cover in optical remote sensing images is a significant challenge, causing missing information and limiting the usage of these images for Earth observation. Specifically, cloud cover hinders flood extent delineation and damage assessment, making it essential to develop methods that can reconstruct missing information and enhance flood classification accuracy. This paper presents an innovative approach to cloud image fusion, leveraging Generative Adversarial Network (GAN) classification, termed CFGANC, to accurately identify flood boundaries in Sentinel-2 real-world image datasets. The proposed methodology of preprocessing and Fusion using Generative Adversarial Network classification (FGANC) addresses cloud cover challenges in optical remote sensing images, hindering flood extent delineation and damage assessment, by utilizing a cloud-aware GAN architecture, multi-temporal Sentinel-2 image fusion, and flood boundary detection. The proposed method demonstrated exceptional performance, outperforming existing methods (RF, GoogleNET, and WCDL) with impressive accuracy metrics of Overall Accuracy (OA) 97.25%, Producer’s Accuracy (PA) 98.69%, and User’s Accuracy (UA) 91.08%.