Resolution Enhancement from Document Images for Text Extraction

Text extraction from low resolution document image sequences suffers from high error rate for most character recognition systems. To address this problem, an effective and efficient interpolation-based resolution enhancement algorithm is proposed and applied in this paper. Image registration and interpolation algorithms are discussed and specified. Further, based on a reliability measurement for pixels, a new iterative weighed average strategy for filling in holes after interpolation is introduced. Finally, comparisons with several other Resolution Enhancement (RE) algorithms are shown in experiments. The results show that definition of images is improved and recognition error rate is decreased.