Preprint
Aug 2026
Every Packet Counts: Dispersing Information for Loss-Resilient Learned Image Compression
This work proposes an end-to-end loss-resilient image compression scheme that generalizes to bursty loss modeled by the Gilbert-Elliott channel, outperforming methods explicitly trained for such conditions.
Yuhang Wei, Chu-Qin Zhou, Yibo Shi et al.
· 0 citations