Related Experiment Video
Updated: May 7, 2026
08:25
Combining Eye-tracking Data with an Analysis of Video Content from Free-viewing a Video of a Walk in an Urban Park Environment
Published on: May 7, 2019
Automatic inpainting scheme for video text detection and removal
Summary
This study introduces an automated framework for removing text from videos. It accurately detects text using novel edge detection and inpainting techniques, restoring video content seamlessly.
Area of Science:
- Computer Vision
- Image Processing
- Digital Signal Processing
Background:
- Video text removal is challenging due to text motion and visual integration.
- Existing methods often struggle with accuracy and visual consistency.
- Automated solutions are needed for efficient video post-processing.
Purpose of the Study:
- To develop a robust two-stage framework for automatic video text removal.
- To enhance video text detection using novel edge detection and clustering.
- To improve video inpainting for seamless restoration of removed text regions.
Main Methods:
- Unsupervised clustering on connected components from Stroke Width Transform (SWT) for text detection.
- A novel edge detector utilizing bandlet transform for accurate edge maps.
- Spatio-temporal geometric flows and 3D volume regularization for video inpainting.
Main Results:
- Effective localization and removal of embedded video texts.
- Accurate reconstruction of removed regions using advanced inpainting.
- Demonstrated visual consistency without additional post-processing.
Conclusions:
- The proposed framework significantly improves automatic video text removal.
- The novel edge detection and inpainting methods are effective.
- The system offers a complete solution for video text removal and restoration.