Scholarlink

<p>With the explosion of data on mobile devices and users, the conflict between the explosion of data traffic in mobile communications and the resource constraints of mobile devices is intensifying. The device-to-device (D2D) communication technology is the ideal technology for direct communication between resource-limited devices. Therefore, for image transmission tasks on edge mobile devices, we propose a depth unfolding network based image transmission scheme. The proposed scheme improves the efficiency of the D2D communication of the device, while ensuring image reconstruction performance. The proposed scheme consists of a deep unfolding network-based image reconstruction model called DAMP-Net+ and a pre-trained compression model. DAMP-Net+ combines DWT with AMP algorithm unfolded to DNNs to increase the Quality of transferred image. In addition, DAMP-NET+ converts the matrix multiplication in the sampling module to a semi-tensor product, significantly reducing the size of the sampling matrix and saving storage resources on edge mobile devices. The experimental data indicates that the proposed scheme achieves the expected results in terms of image reconstruction effect, operation speed and security performance. Compared to the other six methods, DAMP-Net+ obtained better PSNR and SSIM at sampling rates of 1%, 4% and 10%, while obtaining better visual results.</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

A Deep Unfolding Network-Based Image Transmission Scheme in D2D Mobile Edge Networks

  • 期刊论文

  • 发表日期：

    2025年8月1日

查看全文
  • Shuang Bao,
  • Lixiang Li,
  • Haipeng Peng,
  • Junying Liang,
  • Lanlan Wang

摘要

With the explosion of data on mobile devices and users, the conflict between the explosion of data traffic in mobile communications and the resource constraints of mobile devices is intensifying. The device-to-device (D2D) communication technology is the ideal technology for direct communication between resource-limited devices. Therefore, for image transmission tasks on edge mobile devices, we propose a depth unfolding network based image transmission scheme. The proposed scheme improves the efficiency of the D2D communication of the device, while ensuring image reconstruction performance. The proposed scheme consists of a deep unfolding network-based image reconstruction model called DAMP-Net+ and a pre-trained compression model. DAMP-Net+ combines DWT with AMP algorithm unfolded to DNNs to increase the Quality of transferred image. In addition, DAMP-NET+ converts the matrix multiplication in the sampling module to a semi-tensor product, significantly reducing the size of the sampling matrix and saving storage resources on edge mobile devices. The experimental data indicates that the proposed scheme achieves the expected results in terms of image reconstruction effect, operation speed and security performance. Compared to the other six methods, DAMP-Net+ obtained better PSNR and SSIM at sampling rates of 1%, 4% and 10%, while obtaining better visual results.