End-to-End Scene Text Recognition in Videos Based on Multi Frame Tracking.
Xiaobing WangYingying JiangShuli YangXiangyu ZhuWei LiPei FuHua WangZhenbo LuoPublished in: ICDAR (2017)
Keyphrases
- end to end
- multi frame
- multiple frames
- temporal coherence
- scene text recognition
- video frames
- single frame
- motion analysis
- image sequences
- motion estimation
- point correspondences
- moving camera
- video sequences
- computationally efficient
- motion model
- motion segmentation
- optical flow
- object recognition
- super resolution
- consecutive frames
- optic flow
- motion detection
- real time
- admission control
- filtering algorithm
- video surveillance
- video data
- dynamic scenes
- congestion control
- particle filter
- kalman filter
- object tracking
- image processing
- video content
- background subtraction
- video streams
- scale space
- surveillance videos
- multiscale
- feature extraction
- image enhancement
- computer vision