Blindly Assess Quality of In-the-Wild Videos via Quality-aware Pre-training and Motion Perception.
Bowen LiWeixia ZhangMeng TianGuangtao ZhaiXianpei WangPublished in: CoRR (2021)
Keyphrases
- quality aware
- motion features
- visual perception
- visual motion
- video sequences
- space time
- motion analysis
- image sequences
- input video
- high quality
- periodic motion
- spatial and temporal
- motion cues
- moving camera
- dynamic scenes
- camera motion
- key frames
- motion model
- training set
- recognition of human actions
- surveillance videos
- neural network
- visual data
- video data
- video frames
- video signals
- moving objects
- video representation
- motion capture data
- successive frames
- video analysis
- static images
- motion segmentation
- motion capture
- visual cues
- video database
- motion trajectories
- global motion
- video content
- motion field
- temporal coherence
- temporal domain
- point correspondences
- activity recognition
- motion parameters
- optical flow
- motion patterns