Joint Transmission and Deblurring: A Semantic Communication Approach Using Events

TL;DR

The paper tackles motion blur in bandwidth-limited wireless image transmission by introducing EV-JSCC, a semantic communication framework that jointly transmits blurred images and event streams. It leverages a shared encoder to capture common information and separate image/event encoders for domain-specific content, paired with a U-Net–based deblurring decoder that uses cross-attention to fuse modalities. A unified event representation and a three-stage training strategy are proposed to stabilize learning and maximize task-specific performance. Experiments on the GoPro dataset with ESIM-generated events show that EV-JSCC outperforms baselines such as DeepJSCC, DeepJSCC-Deblur, and BPG+LDPC in PSNR, SSIM, and DISTS across multiple channel bandwidth ratios, demonstrating robust, high-quality deblurring under AWGN conditions. The work offers a practical path toward efficient, high-fidelity semantic transmission in real-world surveillance and mobile imaging applications, where motion blur and limited bandwidth are prevalent.

Abstract

Deep learning-based joint source-channel coding (JSCC) is emerging as a promising technology for effective image transmission. However, most existing approaches focus on transmitting clear images, overlooking real-world challenges such as motion blur caused by camera shaking or fast-moving objects. Motion blur often degrades image quality, making transmission and reconstruction more challenging. Event cameras, which asynchronously record pixel intensity changes with extremely low latency, have shown great potential for motion deblurring tasks. However, the efficient transmission of the abundant data generated by event cameras remains a significant challenge. In this work, we propose a novel JSCC framework for the joint transmission of blurry images and events, aimed at achieving high-quality reconstructions under limited channel bandwidth. This approach is designed as a deblurring task-oriented JSCC system. Since RGB cameras and event cameras capture the same scene through different modalities, their outputs contain both shared and domain-specific information. To avoid repeatedly transmitting the shared information, we extract and transmit their shared information and domain-specific information, respectively. At the receiver, the received signals are processed by a deblurring decoder to generate clear images. Additionally, we introduce a multi-stage training strategy to train the proposed model. Simulation results demonstrate that our method significantly outperforms existing JSCC-based image transmission schemes, addressing motion blur effectively.

Figures (6)

• Figure 1: (a) A blurry image $\boldsymbol{B}$ with exposure period [$t_f-T/2$, $t_f+T/2$] and events $\mathcal{E}$ triggered during this period, where red and blue dots represent positive and negative events, respectively; (b) A larger version of the blurry image $\boldsymbol{B}$; (c) The produces of events for a specific pixel ($x,y$) in (b). When the logarithm intensity change exceeds a threshold $c$, the event camera sends an event (i.e., intensity increase results in a positive event (plotted in red) while intensity decrease results in a negative event (plotted in blue)). The generated event set is $\mathcal{E}_{xy}=\{ e_i = (x,y,t_i,p_i):0 \leq i \leq 7 \}$.
• Figure 2: The framework of the proposed EV-JSCC.
• Figure 3: The architecture of the proposed EV-JSCC.
• Figure 4: The performance of the proposed model on the GoPro dataset at a CBR of $1/3$ versus SNR.
• Figure 5: The performance of the proposed model on the GoPro dataset at a CBR of $1/6$ versus SNR.