Scholarlink

<p>In view of the inherent characteristics of sonar images such as low resolution and blurry target edges, this paper proposes a single image super-resolution reconstruction method based on generative adversarial networks, and verifies the reconstructed images by subjective and objective evaluation methods. Compared with the existing results, our proposed method has the following advantages: (i) this method uses two discriminators to constrain pixel and edge information separately, which helps to highlight target edges while improving image resolution; (ii) the generator of this method uses an upsampling module to upsample the image features in stages, which is beneficial to alleviate the mutual interference between low- and high-frequency information in the reconstruction process; (iii) on the premise of ensuring the reconstruction effect, we decrease the number of residual blocks in the generator.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Sonar image super-resolution: stage-wise generative adversarial network with dual discriminators

  • 期刊论文

  • 发表日期：

    2025年4月12日

查看全文
  • Zeyu Dong,
  • Zhuo Wang,
  • Chunbo Lian,
  • Jihong Shen

摘要

In view of the inherent characteristics of sonar images such as low resolution and blurry target edges, this paper proposes a single image super-resolution reconstruction method based on generative adversarial networks, and verifies the reconstructed images by subjective and objective evaluation methods. Compared with the existing results, our proposed method has the following advantages: (i) this method uses two discriminators to constrain pixel and edge information separately, which helps to highlight target edges while improving image resolution; (ii) the generator of this method uses an upsampling module to upsample the image features in stages, which is beneficial to alleviate the mutual interference between low- and high-frequency information in the reconstruction process; (iii) on the premise of ensuring the reconstruction effect, we decrease the number of residual blocks in the generator.