Previous deblurring methods mostly tackle global motion blur due to camera shake but struggle with local motion blur from object movement, facing challenges like the random motion blur locations, data imbalance, directional ambiguity, and positional uncertainty. To fill the vacancy of real-world local motion deblurring, we establish ReLoBlur, the first real-world local motion deblurring dataset. ReLoBlur is captured by a synchronized beam-splitting photographing system and annotated via our developed Local Blur Foreground Mask Generator (LBFMG). To bridge the gap between local and global motion deblurring, we propose a Local Blur-Aware Gated network (LBAG) with gate blocks to focus deblurring on blurred regions, and a Blur-Aware Patch Cropping Strategy (BAPC) to address the data imbalance problem. Acknowledging directional ambiguity and positional uncertainty from shooting errors and non-uniform object motion, we enhance LBAG with LBAGp, guided by center-related distance, and optimized by a symmetric minimization loss. Extensive experiments prove the reliability of the ReLoBlur dataset, and demonstrate that LBAG and LBAGp achieve better local motion deblurring performance compared to state-of-the-art (SOTA) CNN-based deblurring methods.

Auxiliary-free human video matting methods, which rely solely on input frames, often struggle with complex or ambiguous backgrounds. To address this, we propose MatAnyone, a robust framework tailored for target-assigned video matting. Specifically, building on a memory-based paradigm, we introduce a consistent memory propagation module via region-adaptive memory fusion, which adaptively integrates memory from the previous frame. This ensures semantic stability in core regions while preserving fine-grained details along object boundaries. For robust training, we present a larger, high-quality, and diverse dataset for video matting. Additionally, we incorporate a novel training strategy that efficiently leverages large-scale segmentation data, boosting matting stability. With this new network design, dataset, and training strategy, MatAnyone delivers robust and accurate video matting results in diverse real-world scenarios, outperforming existing methods.

Local motion blur commonly occurs in real-world photography due to the mixing between moving objects and stationary backgrounds during exposure. Existing image deblurring methods predominantly focus on global deblurring, inadvertently affecting the sharpness of backgrounds in locally blurred images and wasting unnecessary computation on sharp pixels, especially for high-resolution images. This paper aims to adaptively and efficiently restore high-resolution locally blurred images. We propose a local motion deblurring vision Transformer (LMD-ViT) built on adaptive window pruning Transformer blocks (AdaWPT). To focus deblurring on local regions and reduce computation, AdaWPT prunes unnecessary windows, only allowing the active windows to be involved in the deblurring processes. The pruning operation relies on the blurriness confidence predicted by a confidence predictor that is trained end-to-end using a reconstruction loss with Gumbel-Softmax re-parameterization and a pruning loss guided by annotated blur masks. Our method removes local motion blur effectively without distorting sharp regions, demonstrated by its exceptional perceptual and quantitative improvements (+0.28dB) compared to state-of-the-art methods. In addition, our approach substantially reduces FLOPs by 66% and achieves more than a twofold increase in inference speed compared to Transformer-based deblurring methods. We will make our code and annotated blur masks publicly available.

Mobile Edge Computing (MEC) has solved a sharp increase in data volume caused by various emerging network applications. The edge data center is an essential part of MEC, which connects the edge of the network and the backbone network. Faced with a complex network environment, edge data centers suffer low bandwidth resource utilization and high network latency. This paper proposes Twin Delayed Deep Deterministic policy gradient based Routing Optimization (TRO) algorithm to improve the performance of edge data centers. The TRO algorithm uses Deep Reinforcement Learning (DRL) and Software-Defined Networking (SDN) to achieve routing optimization from two aspects of bandwidth utilization and load balancing. Experiments demonstrate that compared with other algorithms, the TRO algorithm proposed in this paper significantly improves network throughput and reduces average packet latency and average packet latency error.