Application of improved CNN in SAR image noise reduction. Issue 1 (February 2021)

• Record Type: Journal Article
• Authors: Lu, Li; Zhang, Ganchun; Nie, Ying; Liu, Jiayi; fang, Yibiao; Zhang, Guoling; Wu, Yahui
• Is Part Of: Journal of physics. Volume 1792:Issue 1(2021)
• Publication Date: 2021-02
• Subjects: Physics -- Congresses; 530.5
• DOI: 10.1088/1742-6596/1792/1/012053
• Languages: English
• ISSNs: 1742-6588

Abstract

Abstract: Aiming at the problem of coherent speckle noise in SAR image processing, this paper improves the network based on the convolutional neural network method, adjusts the hierarchical relationship in the network hierarchy, and introduces the principles of batch normalization and residual network to eliminate. The gradient disappears when the network level is too deep. The loss function is improved by using multiple convolution kernels of different sizes which are used to extract a variety of different levels of feature information. Finally, based on the peak signal-to-noise ratio (PSNR) and edge index two parameters, the experimental comparison with PPB, SAR-BM3D and NL-SAR three types of denoising methods, The experimental results show that the SAR image generated by the improved CNN network can better achieve despeckle and denoising, and the edges of the image are kept intact.