Perception-Oriented Video Frame Interpolation with Asymmetric Blending
The core message of this paper is to introduce a novel perception-oriented video frame interpolation paradigm called PerVFI, which tackles the challenges of blur and ghosting artifacts by incorporating an asymmetric synergistic blending module and a conditional normalizing flow-based generator.
ColorMNet: A Memory-based Deep Spatial-Temporal Feature Propagation Network for Efficient Video Colorization
The proposed ColorMNet effectively explores spatial-temporal features for video colorization by: 1) using a large-pretrained visual model to guide the estimation of robust spatial features for each frame, 2) developing a memory-based feature propagation module to adaptively propagate useful features from far-apart frames, and 3) exploiting the similar contents of adjacent frames through a local attention module.
Space-Time Video Super-Resolution with Efficient Motion Estimation and Compensation
The proposed Space-Time Neural Operator (STNO) effectively extracts fine-grained spatiotemporal representations from coarse-grained intra-frame features by modeling the task as a mapping between two continuous function spaces. The Galerkin-type attention mechanism in STNO enables precise and efficient motion estimation and compensation, particularly for large motions.
Temporal-Aware Cluster-based Video Summarization: An Efficient and Interpretable Approach
A novel training-free approach called Temporal-Aware Cluster-based SUMmarization (TAC-SUM) that leverages temporal relations between video frames to generate concise and coherent video summaries.
Guided Slot Attention Network for Robust Unsupervised Video Object Segmentation
The proposed guided slot attention network leverages guided slots, feature aggregation transformer, and K-nearest neighbors filtering to effectively separate foreground and background spatial structural information, achieving state-of-the-art performance on challenging video object segmentation datasets.
FMA-Net: A Novel Flow-Guided Dynamic Filtering and Iterative Feature Refinement Framework for Joint Video Super-Resolution and Deblurring
The proposed FMA-Net framework effectively handles spatio-temporally-variant degradations in blurry low-resolution videos through flow-guided dynamic filtering and iterative feature refinement with multi-attention.
BlazeBVD: A Comprehensive Approach to Blind Video Deflickering
BlazeBVD introduces a novel approach to blind video deflickering, leveraging histogram-assisted solutions to enhance temporal consistency and eliminate flickering artifacts.
Learning-based Axial Video Motion Magnification: Enhancing Legibility of Small Motions
Improving legibility of small motions through axial motion magnification.
Text-Conditioned Resampler For Long Form Video Understanding
Efficiently process long video sequences using a text-conditioned resampler for improved performance in various tasks.
Enhancing Video Transformers for Action Understanding with VLM-aided Training
Integrating Visual Language Models with Vision Transformers enhances video action understanding by aligning spatio-temporal representations.
