SAR decompressed image reconstruction algorithm based on generative adversarial network

doi:10.7523/j.ucas.2023.033

Abstract

Abstract: The high multiple compression processing of SAR image will cause damage to the target and texture information in the image, which makes the problems of blur and indistinguishable targets often appear in the decompressed SAR image, and is difficult to effectively reflect the real features of ground objects. To solve the above problems, a new SAR image reconstruction algorithm is proposed based on the generative adversarial network. Based on the codec structure, the algorithm takes the parallel fusion of convolutional neural network and self-attention mechanism as the generator, and designed a simple and efficient architecture called ConTransformer, which can get richer global features and effectively improve the effect of small-target reconstruction. For the Discriminator, spectral normalization is introduced into the U-Net feature extractor to reduce the sensitivity of the model to input disturbance, so as to suppress artifacts. Synchronously, the pre-training mask mechanism is used to enhance the extraction of high-level semantic features and improve the authenticity of reconstructed images. Experimental results show that the reconstructed images obtained by this method have clearer visual effects and better key performance indexes than those obtained by Real-ESRGAN and other typical methods based on generative adversarial networks, among which the peak signal-to-noise ratio is improved by 0.57dB~1.54dB.

Key words: SAR decompressed image, generative adversarial networks(GAN), contransformer encoder, masking mechanism

CLC Number:

TP751.1

ZHANG Bingyu, PAN Zhigang, YAO Kai, DONG Xubin. SAR decompressed image reconstruction algorithm based on generative adversarial network[J]. Journal of University of Chinese Academy of Sciences, DOI: 10.7523/j.ucas.2023.033.

References

[1] Wiley C A. Synthetic aperture radars[J]. IEEE Transactions on Aerospace and Electronic Systems, 1985, AES-21(3): 440-443. DOI: 10.1109/TAES.1985.310578.
[2] 潘志刚, 王小龙, 李志勇. SAR原始数据压缩的自适应比特分配BAQ算法[J]. 中国科学院大学学报, 2017, 34(1): 106-111. DOI: 10.7523/j.issn.2095-6134.2017.01.014.
[3] 潘志刚, 高鑫. 针对纹理图像压缩的改进SPIHT算法[J]. 中国科学院研究生院学报, 2010, 27(2): 222-227. DOI: 10.7523/j.issn.2095-6134.2010.2.012.
[4] 李萌, 刘畅. 基于特征复用的膨胀-残差网络的SAR图像超分辨重建[J]. 雷达学报, 2020, 9(2): 363-372. DOI: 10.12000/JR19110.
[5] Gu F, Zhang H, Wang C, et al.SAR image super-resolution based on noise-free generative adversarial network[C]//IGARSS 2019-2019 IEEE International Geoscience and Remote Sensing Symposium. July 28-August 2, 2019, Yokohama, Japan. IEEE, 2019: 2575-2578. DOI: 10.1109/IGARSS.2019.8899202.
[6] Cen X, Song X, Li Y C, et al.A deep learning-based super-resolution model for bistatic SAR image[C]//2021 International Conference on Electronics, Circuits and Information Engineering (ECIE). January 22-24, 2021, Zhengzhou, China. IEEE, 2021: 228-233. DOI: 10.1109/ECIE52353.2021.00056.
[7] Dosovitskiy A, Beyer L, Kolesnikov A, et al. An image is worth 16x16 words: Transformers for image recognition at scale[EB/OL]. (2020-10-22)[2023-03-22]. https://arxiv.org/abs/2010.11929.
[8] Schönfeld E, Schiele B, Khoreva A.A U-net based discriminator for generative adversarial networks[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). June 13-19, 2020, Seattle, WA, USA. IEEE, 2020: 8204-8213. DOI: 10.1109/CVPR42600.2020.00823.
[9] Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[EB/OL]. (2014-09-04)[2023-03-22]. https://arxiv.org/abs/1409.1556.
[10] He K M, Chen X L, Xie S N, et al.Masked autoencoders are scalable vision learners[C]//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). June 18-24, 2022, New Orleans, LA, USA. IEEE, 2022: 15979-15988. DOI: 10.1109/CVPR52688.2022.01553.
[11] Carion N, Massa F, Synnaeve G, et al.End-to-end object detection with transformers[C]//European Conference on Computer Vision. Cham: Springer, 2020: 213-229. DOI: 10.1007/978-3-030-58452-8_13.
[12] Ramachandran P, Parmar N, Vaswani A, et al. Stand-alone self-attention in vision models[EB/OL]. (2019-06-13)[2023-03-22]. https://arxiv.org/abs/1906.05909.
[13] Si C Y, Yu W H, Zhou P, et al. Inception transformer[EB/OL]. (2022-05-25)[2023-03-22]. https://arxiv.org/abs/2205.12956.
[14] Xu W J, Xu Y F, Chang T, et al.Co-scale conv-attentional image transformers[C]//2021 IEEE/CVF International Conference on Computer Vision (ICCV). October 10-17, 2021, Montreal, QC, Canada. IEEE, 2022: 9961-9970. DOI: 10.1109/ICCV48922.2021.00983.
[15] Xu Y F, Zhang Q M, Zhang J, et al. ViTAE: Vision transformer advanced by exploring intrinsic inductive bias[EB/OL]. (2021-06-07)[2023-03-22]. https://arxiv.org/abs/2106.03348.
[16] Chen B Y, Li P X, Li C M, et al.GLiT: neural architecture search for global and local image transformer[C]//2021 IEEE/CVF International Conference on Computer Vision (ICCV). October 10-17, 2021, Montreal, QC, Canada. IEEE, 2022: 12-21. DOI: 10.1109/ICCV48922.2021.00008.
[17] Yang R, Ma H L, Wu J, et al.ScalableViT: rethinking the context-oriented generalization of vision transformer[C]//European Conference on Computer Vision. Cham: Springer, 2022: 480-496. DOI: 10.1007/978-3-031-20053-3_28.
[18] Gu J J, Dong C.Interpreting super-resolution networks with local attribution maps[C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). June 20-25, 2021, Nashville, TN, USA. IEEE, 2021: 9195-9204. DOI: 10.1109/CVPR46437.2021.00908.
[19] Park N, Kim S. How do vision transformers work? [EB/OL]. (2022-02-14)[2023-03-22]. https://arxiv.org/abs/2202.06709.
[20] Ronneberger O, Fischer P, Brox T.U-net: Convolutional networks for biomedical image segmentation[C]//International Conference on Medical Image Computing and Computer-Assisted Intervention. Cham: Springer, 2015: 234-241. DOI: 10.1007/978-3-319-24574-4_28.
[21] Lata K, Dave M, Nishanth K N.Image-to-image translation using generative adversarial network[C]//2019 3rd International conference on Electronics, Communication and Aerospace Technology (ICECA). June 12-14, 2019, Coimbatore, India. IEEE, 2019: 186-189. DOI: 10.1109/ICECA.2019.8822195.
[22] Miyato T, Kataoka T, Koyama M, et al. Spectral normalization for generative adversarial networks[EB/OL]. (2018-02-16)[2023-03-22]. https://arxiv.org/abs/1802.05957.
[23] Ledig C, Theis L, Huszár F, et al.Photo-realistic single image super-resolution using a generative adversarial network[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). July 21-26, 2017, Honolulu, HI, USA. IEEE, 2017: 105-114. DOI: 10.1109/CVPR.2017.19.
[24] Wang X T, Yu K, Wu S X, et al.ESRGAN: enhanced super-resolution generative adversarial networks[C]//European Conference on Computer Vision. Cham: Springer, 2019: 63-79. DOI: 10.1007/978-3-030-11021-5_5.
[25] Wang X T, Xie L B, Dong C, et al.Real-ESRGAN: Training real-world blind super-resolution with pure synthetic data[C]//2021 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW). October 11-17, 2021, Montreal, BC, Canada. IEEE, 2021: 1905-1914. DOI: 10.1109/ICCVW54120.2021.00217.