文献总结2021-08-18

无清洁图像去噪

1. Neighbor2Neighbor: Self-Supervised Denoising from Single Noisy Images

单幅图像自监督去噪

		输入噪声图
		随机近邻采样得到一组图像对
		计算网络输出与噪声目标之间的重构项和下采样图像与真实图像之间的差异
		最小化损失函数，更新网络参数

Noise2Noise: Learning Image Restoration without Clean Data

同一场景下，多幅含噪图像的自监督去噪

2. Nonlocal CNN SAR Image Despeckling

非局部CNN SAR图像去斑

		使用非局部加权替代卷积操作
		![非局部层结构图](https://img-blog.csdnimg.cn/46ce7006208547bfa9164402e887379c.png?x-oss-process=image/watermark,type_ZHJvaWRzYW5zZmFsbGJhY2s,shadow_50,text_Q1NETiBAaGV6aTMyMQ==,size_20,color_FFFFFF,t_70,g_se,x_16#pic_center)

Neural Nearest Neighbors Networks (N3层)

CNN Denoisers As Non-Local Filters: The Neural Tangent Denoiser

		CNN去噪与非局部滤波

3. Deep Multi-Scale Recurrent Network for Synthetic Aperture Radar Images Despeckling

深度学习多尺度卷积神经网络动态去模糊