![[2510.06868] Multi-hop Deep Joint Source-Channel Coding with Deep Hash Distillation for Semantically Aligned Image Recovery](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2510.06868] Multi-hop Deep Joint Source-Channel Coding with Deep Hash Distillation for Semantically Aligned Image Recovery
Summary
This paper presents a novel approach to image transmission using multi-hop deep joint source-channel coding (DeepJSCC) combined with deep hash distillation for improved semantic alignment and perceptual quality in image recovery.
Why It Matters
The research addresses challenges in image transmission over noisy channels, particularly in security-sensitive applications. By enhancing semantic consistency and perceptual quality, this work has implications for fields such as telecommunications, computer vision, and artificial intelligence, where reliable image recovery is crucial.
Key Takeaways
- Introduces a multi-hop DeepJSCC framework for image recovery.
- Utilizes deep hash distillation to improve semantic clustering of images.
- Demonstrates significant improvements in perceptual quality over classical methods.
- Addresses noise accumulation issues in multi-hop transmission settings.
- Highlights the importance of semantic alignment in security-oriented applications.
Computer Science > Information Theory arXiv:2510.06868 (cs) [Submitted on 8 Oct 2025 (v1), last revised 24 Feb 2026 (this version, v2)] Title:Multi-hop Deep Joint Source-Channel Coding with Deep Hash Distillation for Semantically Aligned Image Recovery Authors:Didrik Bergström, Deniz Gündüz, Onur Günlü View a PDF of the paper titled Multi-hop Deep Joint Source-Channel Coding with Deep Hash Distillation for Semantically Aligned Image Recovery, by Didrik Bergstr\"om and Deniz G\"und\"uz and Onur G\"unl\"u View PDF HTML (experimental) Abstract:We consider image transmission via deep joint source-channel coding (DeepJSCC) over multi-hop additive white Gaussian noise (AWGN) channels by training a DeepJSCC encoder-decoder pair with a pre-trained deep hash distillation (DHD) module to semantically cluster images, facilitating security-oriented applications through enhanced semantic consistency and improving the perceptual reconstruction quality. We train the DeepJSCC module to both reduce mean square error (MSE) and minimize cosine distance between DHD hashes of source and reconstructed images. Significantly improved perceptual quality as a result of semantic alignment is illustrated for different multi-hop settings, for which classical DeepJSCC may suffer from noise accumulation, measured by the learned perceptual image patch similarity (LPIPS) metric. Comments: Subjects: Information Theory (cs.IT); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR); Machine Lear...
