DCT Feature Based Text Capturing and Tracking
Shi Jianyong, Xiling Luo, Jun Zhang · 2009
An autonomous text capturing and tracking method for digital video is proposed in this paper. Frames are divided into blocks. Text blocks are extracted first using a discrete cosine transformation (DCT) based method, then the optimal text region is extracted from duplicate regions within the consecutive frames using the method of text tracking, and is used as an input for optical character recognition (OCR) system. With the reduction of duplicate regions and optimization of region extraction, OCR's recognition efficiency and accuracy are both improved.