Title: Multi-hop Deep Joint Source-Channel Coding with Deep Hash Distillation for Semantically Aligned Image Retrieval Show Chinese title

Title: 多跳深度联合源信道编码与深度哈希蒸馏用于语义对齐的图像检索

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. Show Chinese abstract

Abstract: 我们考虑通过深度联合源信道编码（DeepJSCC）在多跳加性高斯白噪声（AWGN）信道上进行图像传输，通过使用预训练的深度哈希蒸馏（DHD）模块来训练一个DeepJSCC编码解码器对，以语义聚类图像，通过增强的语义一致性促进安全导向的应用，并提高感知重建质量。我们训练DeepJSCC模块以同时减少均方误差（MSE）并最小化源图像和重建图像的DHD哈希之间的余弦距离。由于语义对齐而显著改进的感知质量在不同的多跳设置中得到说明，对于这些设置，经典DeepJSCC可能会因噪声累积而受到影响，这是通过学习的感知图像块相似性（LPIPS）度量来衡量的。