The existing remote sensing (RS) image dehazing methods based on deep learning have sought help from the convolutional frameworks. Nevertheless, the inherent limitations of convolution, i.e ., local receptive fields and independent input elements, curtail the network from learning the long-range dependencies and non-uniform distributions. To this end, we design an effective RS image dehazing Transformer architecture, denoted as RSDformer. Firstly, given the irregular shapes and non-uniform distributions of haze in RS images, capturing both local and non-local features is crucial for RS image dehazing models. Hence, we propose a detail-compensated transposed attention to extract the global and local dependencies across channels. Secondly, to enhance the ability to learn degraded features and better guide the restoration process, we develop a dual-frequency adaptive block with dynamic filters. Finally, a dynamic gated fusion block is designed to achieve fuse and exchange features across different scales effectively. In this way, the model exhibits robust capabilities to capture dependencies from both global and local areas, resulting in improving image content recovery. Extensive experiments prove that the proposed method obtains more appealing performances against other competitive methods.

Lightweight video representation techniques have advanced significantly for simple activity recognition, but they still encounter several issues when applied to complex activity recognition: (i) The presence of numerous individuals and varying spatial positions makes it difficult for traditional token pruning methods to maintain accuracy. (ii) Simply discarding entire frames may result in the loss of crucial clues. (iii) To maintain parallel computing, applying the same pruning rate to every frame leads to significant redundancy in frames with low information content. To this end, we propose a lightweight and novel Spatial-Temporal Token Pruning and Merging (STPM) framework, specifically designed for complex action videos where human actors occupy a small spatial resolution within video frames. Our framework considers two critical factors: semantic importance and spatial-temporal redundancy, to further reduce overhead. For semantic importance, STPM captures class-specific attention scores by learning multiple class tokens within the transformer to guide token pruning. For spatial-temporal redundancy, STPM employs an anchor graph and temporal attention to perform spatial and temporal token merging, preserving appearance and temporal cues while eliminating semantic duplication and redundancy. We conduct extensive experiments on JRDB-PAR primarily using recently introduced video transformer backbones, e.g., MViT and ViT. Our framework achieves similar results while requiring 40% less computation.

Computer vision technology is increasingly being used in areas such as intelligent security and autonomous driving. Users need accurate and reliable visual information, but the images obtained under severe weather conditions are often disturbed by rainy weather, causing image scenes to look blurry. Many current single image deraining algorithms achieve good performance but have limitations in retaining detailed image information. In this paper, we design a Scale-space Feature Recalibration Network (SFR-Net) for single image deraining. The proposed network improves the image feature extraction and characterization capability of a Multi-scale Extraction Recalibration Block (MERB) using dilated convolution with different convolution kernel sizes, which results in rich multi-scale rain streaks features. In addition, we develop a Subspace Coordinated Attention Mechanism (SCAM) and embed it into MERB, which combines coordinated attention recalibration and a subspace attention mechanism to recalibrate the rain streaks feature information learned from the feature extraction phase and eliminate redundant feature information to enhance the transfer of important feature information. Meanwhile, the overall SFR-Net structure uses dense connection and cross-layer feature fusion to repeatedly utilize the feature maps, thus enhancing the understanding of the network and avoiding gradient disappearance. Through extensive experiments on synthetic and real datasets, the proposed method outperforms the recent state-of-the-art deraining algorithms in terms of both the rain removal effect and the preservation of image detail information.

Nighttime image deraining presents unique challenges compared to daytime scenarios, primarily due to the complex characteristics of nighttime scenes and the limited availability of datasets that effectively capture the interplay between lighting and rain streaks. In this paper, we contribute a high-quality nighttime image deraining dataset, BDD-NightRain, consisting of 5,600 synthetic nighttime rain images with higher realism and harmony for driving scenes. Furthermore, we develop a two-stage Nighttime Rain Removal Network (NRRNet) that leverages color space transformation. Specifically, the coarse processing stage focuses on rain streak removal in the Y channel, where luminance contrast is most pronounced under nighttime rain conditions, while the fine processing stage further refines the deraining results of the previous stage. In addition, we introduce coordinate-based implicit neural representations between these two stages to guide the model's focus toward areas with appropriate illumination. Extensive experiments show that our method outperforms others in removing rain streaks from nighttime scenes, with potential applications in night time object detection. The code and dataset will be available at https://github.com/guanqiyuan/NRRNet.