关键词: 深度学习, 高质量图像生成, 生成对抗网络, 声呐图像处理, 特征融合

Abstract: Due to the inherent characteristics of sonar imaging principles and the interference of complex underwater environments, underwater sonar images generally suffer from insufficient resolution and missing target details. To address this issue,a high-quality sonar image generation method based on multi-scale feature fusion was proposed . Firstly, the residual dense connection module was used to extract image features at a shallow level, capturing basic texture and contour information, and established the spatial layout of the image; Secondly, a Multi-Scale Attention feature extraction module (MSA) was designed, which can adaptively focus on key features at different scales and further enhance the expression of key features while suppressing redundant information expression through attention mechanisms; Finally, a discriminator network was designed using a pixel by pixel discrimination strategy based on spectral normalization, which improved the reconstruction ability of complex object edges and details. Experiments on underwater sonar image datasets show that the proposed method achieves relative improvements of 6.7% and 5.4% in Peak Signal-to-Noise Ratio (PSNR) and Structural SIMilarity (SSIM) metrics compared to the existing representative method ESRGAN(Enhanced Super-Resolution Generative Adversarial Networks). The above experimental results indicate that the proposed method effectively improves the generation performance on underwater sonar image datasets.

Key words: deep learning, high-quality image generation, generative adversarial networks, sonar image processing, feature fusion