Detection and Tracking of Rolling Caption in Video
Qinru Li · Journal of Chinese Computer Systems · 2015
The detection and tracking of rolling caption,which usually deliver complete messages in video,are the premise of caption recognition and application. Analyzing the features of rolling caption and Sobel edge images,this paper proposes a caption detection method based on the integration of rules and statistics and a caption tracking method using the vertical edge features. From the overall perspective of video,this paper differences the adjacent frame,extracts horizontal edge features to detect the candidate edge of the caption's area according to certain rules,and counts the frequency of the edge's occurrence to locate the rolling caption's area. Then this paper calculates the sum of square of adjacent caption's vertical edge features under different offsets,and takes the minimum value as the offset of adjacent captions to realize the caption tracking. Experimental result shows that our method can achieve 96. 61% precision in detecting rolling caption from video containing various kinds of caption,99. 43% precision in tracking rolling caption.