衛星圖像已被用於農業、資源管理、氣象等領域。然而，大約67%的衛星圖像被雲層覆蓋，因此難以提取地面信息。尤其是當厚厚的雲層完全遮擋地面信息時，雲層覆蓋的問題變得更加嚴重。
針對上述問題，本文提出了一種三階段厚雲去除方法，利用 U-Net 對厚雲進行分割，然後利用 CGAN 製備具有缺失區域的背景衛星圖像重建地面信息。最後，我們獲得了無雲衛星圖像。我們使用包含 22,500 張用於訓練的圖像和 752 張用於測試的圖像的合成數據集來訓練 U-Net 和 CGAN。我們通過實驗表明，我們提出的方法優於傳統方法，並且可以實現峰值信噪比（PSNR）和結構相似性（SSIM）的高分。
此外，我們構建了一個合成數據集來擴充數據，並且不需要人工標註。合成數據集由無雲衛星圖像、背景衛星圖像、多雲衛星圖像和掩膜雲圖像組成。缺失或云覆蓋區域小於圖像中總像素的 50%。

The satellite images have been used in many applications such as agriculture, resource management, meteorology etc. However, roughly 67% of satellite images are covered by clouds, thus makes it difficult to extract ground information. The problem of cloud cover becomes more serious especially when thick clouds totally block the ground information.
To solve the problem mentioned above, this thesis proposes a three-stage method for thick cloud removal to segment thick clouds by using U-Net, then preparing background satellite images with missing regions to reconstruct ground information by using CGAN. Finally, we obtain a cloud-free satellite image. We train both U-Net and CGAN with our synthetic dataset which contains 22,500 images for training and 752 images for testing. We show experimentally that our proposed method outperforms the traditional method and can achieve a high score of peak signal-to-noise ratio (PSNR) and Structural Similarity (SSIM).
In addition, we build a synthetic dataset to augment data and no need to require hu-man annotation. The synthetic dataset consists of cloud-free satellite image, background satellite image, cloudy satellite image, and mask cloud image. The missing or cloud cov-er regions are less than 50% of the total pixels in the image.

[1] M. D. King, S. Platnick, W. P. Menzel, S. A. Ackerman and P. A. Hubanks, "Spatial and Temporal Distribution of Clouds Observed by MODIS Onboard the Terra and Aqua Satellites," IEEE Transactions on Geoscience and Remote Sensing, vol. 51, no. 7, pp. 3826-3852, July 2013.
[2] I. Goodfellow et al., "Generative Adversarial Nets," in Proc. Adv. Neural Inf. Process. Syst. , pp. 2672–2680, 2014.
[3] H. Zhang, V. Sindagi, and V. M. Patel, "Image Deraining Using a Conditional Generative Adversarial Network," arXiv preprint arXiv:1701.05957, 2017.
[4] Praveer Singh and Nikos Komodakis, "Cloud-GAN: Cloud Removal for Sentinel-2 Imagery using a Cyclic Consistent Generative Adversarial Networks," in IGARSS 2018-2018 IEEE International Geoscience and Remote Sensing Symposium, pages 1772–1775. IEEE, 2018.
[5] Jun-Yan Zhu, Taesung Park, Phillip Isola, and Alexei A Efros, "Unpaired Image-to-Image Translation using Cycle Consistent Adversarial Networks," in Proceedings of the IEEE International Conference on Computer Vision, pages 2223– 2232, 2017.
[6] R. Furuta, "Synthetic Aperture Radar (SAR) Utilization for Disaster Management," in Technological Seminar on Environmental Monitoring, 2014.
[7] Kenji Enomoto, Ken Sakurada, Weimin Wang, Hiroshi Fukui, Masashi Matsuoka, Ryosuke Nakamura, and Nobuo Kawaguchi, "Filmy cloud removal on satellite imagery with multispectral conditional generative adversarial nets," arXiv preprint arXiv:1710.04835, 2017.
[8] Alqahtani Hamed, Kavakli-Thorne Manolya, Kumar Ahuja Gulshan, " Applications of Generative Adversarial Networks (GANs): An Updated Review. Archives of Computational Methods in Engineering," 28. 10.1007/s11831-019-09388-y, 2019.
[9] Andrea Meraner, Patrick Ebel, Xiao Xiang Zhu, Michael Schmitt, "Cloud removal in Sentinel-2 imagery using a deep residual neural network and SAR-optical data fusion," ISPRS Journal of Photogrammetry and Remote Sensing, Volume 166, 2020.
[10] P. Isola, J.-Y. Zhu, T. Zhou, and A. A. Efros, "Image-to-Image Translation with Conditional Adversarial Networks," arXiv preprint arXiv:1611.07004, 2016.
[11] O. Ronneberger, P. Fischer, and T. Brox, "U-Net: Convolutional Networks for Biomedical Image Segmentation," in Medical Image Computing and Computer-Assisted Intervention. Cham, Switzerland: Springer, pp. 234–241, 2015.
[12] S. Mohajerani, T. A. Krammer and P. Saeedi, "A Cloud Detection Algorithm for Remote Sensing Images Using Fully Convolutional Neural Networks," IEEE 20th International Workshop on Multimedia Signal Processing (MMSP), Vancouver, BC, 2018.
[13] S. Mohajerani, and P. Saeedi, "Cloud-Net: An End-To-End Cloud Detection Algorithm for Landsat 8 Imagery," IGARSS 2019 - 2019 IEEE International Geoscience and Remote Sensing Symposium, Yokohama, Japan, 2019.
[14] Heng Pan, "Cloud Removal for Remote Sensing Imagery via Spatial Attention Generative Adversarial Network," arXiv preprint arXiv:2009.13015, 2020.
[15] Han Feilong, Zhang Hongbing, Chatterjee Snehamoy, Guo Qiang, and Wan Shulin, "A Modified Generative Adversarial Nets Integrated With Stochastic Approach for Realizing Super-Resolution Reservoir Simulation," IEEE Transactions on Geoscience and Remote Sensing, 2020.
[16] "Discriminator," Machine Learning, [Online], Available: https://developers.google.com/machine-learning/gan/discriminator
[17] "Generator," Machine Learning, [Online], Available: https://developers.google.com/machine-learning/gan/generator
[18] Sarukkai Vishnu, Jain Anirudh, Uzkent Burak, and Ermon Stefano, "Cloud Removal in Satellite Images Using Spatiotemporal Generative Networks," IEEE Winter Conference on Applications of Computer Vision (WACV), Snowmass Village, CO, USA, 2020.
[19] C. Grohnfeldt, M. Schmitt and X. Zhu, "A Conditional Generative Adversarial Network to Fuse Sar And Multispectral Optical Data For Cloud Removal From Sentinel-2 Images," IGARSS 2018 - 2018 IEEE International Geoscience and Remote Sensing Symposium, 2018.
[20] "IRIS-B," IRIS Project, [Online], Available: http://satellite.ncku.edu.tw/iris_b.html
[21] Cun jun L I, Liu L Y, Wang J H, "Automatic Detection and Removal of Thin Haze based on Own Features of Landsat Image" Journal of Zhejiang University, 40(1):10-13+37, 2006.
[22] He X Y , Hu J B , Chen W. Haze, "Removal based on Advanced Haze Optimized Transformation (AHOT) for Multispectral Imagery" International Journal of Remote Sensing, 31(20):5331-5348, 2010.
[23] Wang X W X, " A Modified Homomorphism Filtering Algorithm for Cloud Removal," International Conference on Computational Intelligence & Software Engineering, IEEE, 2010.
[24] T. Eelbode et al., "Optimization for Medical Image Segmentation: Theory and Practice When Evaluating With Dice Score or Jaccard Index," in IEEE Transactions on Medical Imaging, vol. 39, no. 11, pp. 3679-3690, Nov. 2020.