DriftDecode delivers a 30 ms one-step SNR-conditioned U-Net decoder for wireless images by reformulating drifting-field mechanisms into an instance-level texture loss, yielding up to 1.13 dB PSNR gain and 4.8x speedup over 10-step flow-matching baselines on MNIST and DIV2K under AWGN and Rayleigh-f
The perception-distortion tradeoff,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.IV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
DriftDecode: One-Step Wireless Image Decoding via Drifting-Inspired Detail Recovery
DriftDecode delivers a 30 ms one-step SNR-conditioned U-Net decoder for wireless images by reformulating drifting-field mechanisms into an instance-level texture loss, yielding up to 1.13 dB PSNR gain and 4.8x speedup over 10-step flow-matching baselines on MNIST and DIV2K under AWGN and Rayleigh-f