Scholarlink

Image rendering is crucial in various domains like the metaverse, creative computing, VR, and AR, enhancing user immersion and merging technology with art. However, compared to research on haze and rain conditions, dust weather rendering has received less attention. Existing methods for simulating dust weather have limitations and lack realism. This paper proposes a dust weather synthesis algorithm using GANs and an atmospheric scattering model. Combining GANs’ generative power with the accuracy of the atmospheric scattering model improves visual effects, overcoming issues like hollow artifacts and poor detail. Our method outperforms the comparison algorithm in IE by 0.25 and reduces FID by 11.81, validating DustGAN’s effectiveness in producing realistic dust images. The code is available at https://github.com/GuoYuxx/DustGAN .

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

DustGAN: Unpaired Learning for Dust Image Simulation Based on ASM with GANs

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Yuxin Guo,
  • Pengfei Cao,
  • Yang Chen,
  • Fuquan Zhang,
  • Zhuoli He,
  • Manying Shen,
  • Jiayan Huang,
  • Chuansheng Wang

摘要

Image rendering is crucial in various domains like the metaverse, creative computing, VR, and AR, enhancing user immersion and merging technology with art. However, compared to research on haze and rain conditions, dust weather rendering has received less attention. Existing methods for simulating dust weather have limitations and lack realism. This paper proposes a dust weather synthesis algorithm using GANs and an atmospheric scattering model. Combining GANs’ generative power with the accuracy of the atmospheric scattering model improves visual effects, overcoming issues like hollow artifacts and poor detail. Our method outperforms the comparison algorithm in IE by 0.25 and reduces FID by 11.81, validating DustGAN’s effectiveness in producing realistic dust images. The code is available at https://github.com/GuoYuxx/DustGAN .