Block-based Static Caption Detection Using Intensity Range

초록

This paper proposes a new method to detect static caption regions using intensity range maps and a statistical model of motion vectors. The method relies on the probability of block based intensity ranges and magnitudes of motion vectors. Intensity range is used to find candidate locations of static captions and the statistical model of motion vector is used to refine accurate caption regions. The proposed method was tested with dynamic scenes and static scenes. The proposed method detects static caption region with almost 90% accuracy.

키워드