Yuzhe Zhang, Feiran Liu, Yi Shan, Xinyi Huang, Xin Yang, Yueqi Zhu, Xuxin Cheng, Cao Liu, Ke Zeng, Terry Jingchen Zhang, Wenyuan Jiang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Adverse weather (e.g., rain, snow, and haze) degrades visual content and undermines downstream multimedia retrieval and object recognition. While recent restoration paradigms utilizing vision transformers and diffusion models have achieved remarkable perceptual quality, their prohibitive computational cost and latency render them unsuitable for low-latency large-scale multimedia stream processing. To bridge the gap between restoration quality and system efficiency, we propose the Hybrid State Space Network (HSS-Net), an efficient preprocessing solution tailored for multimedia understanding systems. In this work, we innovatively construct a novel Hybrid State Space Block (HSSB) by integrating Mamba-based State Space Models (SSMs) into a unified restoration framework. This block leverages the linear complexity of SSMs to model long-range degradation contexts, complemented by a parallel Texture-Aware Gated Module (TAGM) for precise local texture recovery. Furthermore, to address scale variations in multimedia objects, we design a Cross-Scale Feature Fusion Module (CFFM) to enhance semantic feature representation. Extensive experiments demonstrate that HSS-Net achieves state-of-the-art restoration performance. More importantly, with only 12.1M parameters, it significantly boosts downstream object detection accuracy while maintaining low-latency inference speeds, positioning it as an optimal solution for efficient, edge-deployed multimedia analysis. The code will be available at: https://github.com/Moonlitwine/HSS-Net.

Locating the start, apex and end keyframes of moving contrast agents for keyframe counting in X-ray coronary angiography (XCA) is very important for the diagnosis and treatment of cardiovascular diseases. To locate these keyframes from the class-imbalanced and boundary-agnostic foreground vessel actions that overlap complex backgrounds, we propose long short-term spatiotemporal attention by integrating a convolutional long short-term memory (CLSTM) network into a multiscale Transformer to learn the segment- and sequence-level dependencies in the consecutive-frame-based deep features. Image-to-patch contrastive learning is further embedded between the CLSTM-based long-term spatiotemporal attention and Transformer-based short-term attention modules. The imagewise contrastive module reuses the long-term attention to contrast image-level foreground/background of XCA sequence, while patchwise contrastive projection selects the random patches of backgrounds as convolution kernels to project foreground/background frames into different latent spaces. A new XCA video dataset is collected to evaluate the proposed method. The experimental results show that the proposed method achieves a mAP (mean average precision) of 72.45% and a F-score of 0.8296, considerably outperforming the state-of-the-art methods. The source code is available at https://github.com/Binjie-Qin/STA-IPCon.

Although robust PCA has been increasingly adopted to extract vessels from X-ray coronary angiography (XCA) images, challenging problems such as inefficient vessel-sparsity modelling, noisy and dynamic background artefacts, and high computational cost still remain unsolved. Therefore, we propose a novel robust PCA unrolling network with sparse feature selection for super-resolution XCA vessel imaging. Being embedded within a patch-wise spatiotemporal super-resolution framework that is built upon a pooling layer and a convolutional long short-term memory network, the proposed network can not only gradually prune complex vessel-like artefacts and noisy backgrounds in XCA during network training but also iteratively learn and select the high-level spatiotemporal semantic information of moving contrast agents flowing in the XCA-imaged vessels. The experimental results show that the proposed method significantly outperforms state-of-the-art methods, especially in the imaging of the vessel network and its distal vessels, by restoring the intensity and geometry profiles of heterogeneous vessels against complex and dynamic backgrounds. The source code is available at https://github.com/Binjie-Qin/RPCA-UNet.