Graph‐based approach to scene text localisation and tracking in videos. Issue 13 (2nd June 2016)

• Record Type: Journal Article
• Authors: Kim, Y.G.; Koo, H.I.
• Is Part Of: Electronics letters. Volume 52:Issue 13(2016)
• Page Start: 1110
• Page End: 1112
• Publication Date: 2016-06-02
• Subjects: text detection -- text analysis -- video signal processing -- graph theory -- natural scenes -- feature extraction -- object tracking -- image filtering -- interpolation; scene text localisation -- scene text tracking -- unique ID assignment -- graph‐based framework -- image‐level text detection -- graph vertices -- correspondence scores -- text‐region trajectories -- graph‐cut algorithm -- patch‐based tracking -- false positive filtering -- interpolation -- International Conference on Document Analysis and Recognition 2015 -- video text localisation -- CLEAR MOT metrics -- VACE metrics; Electronics -- Periodicals; 621.381
• DOI: 10.1049/el.2016.1008
• Languages: English
• ISSNs: 0013-5194

Abstract

Abstract : A text localisation and tracking method is presented that finds text‐regions in videos and assigns unique IDs to their trajectories. For the goal, a graph‐based framework that can work with existing text detection methods is developed. To be precise, graphs are built where vertices are image‐level text detection results and edges represent the correspondence scores of the vertices. From these graphs, text‐region trajectories by using the graph‐cut algorithm are extracted. This approach allows considering false positives and misses, as well as their patch‐based tracking results at the same time, and text trajectories are reliably extracted. Finally, the results are refined by interpolating misses and filtering out false positives. The proposed method is submitted to the International Conference on Document Analysis and Recognition 2015 robust reading competition (video text localisation) and the method showed the best performance in terms of CLEAR MOT metrics and was ranked third place according to VACE metrics among the seven participating methods.