Tilt Correction of Italic Character Images

Skewing of Chinese character images interferes with structure perception and feature extraction, and is a key factor affecting recognition and generation accuracy. Existing methods focus on text line level correction, which is difficult to cope with single character level images, and especially challenging when dealing with characters with italicized styles. Italicized styles tend to lead to systematic tilting, exacerbating structural deformation and further interfering with recognition. To this end, this paper proposes a single-character tilt correction method based on image alignment algorithm, which introduces multi-angle rotation and scale normalization in the feature search stage, and minimum outer rectangle analysis in the geometric transformation stage, to achieve correction of italicized tilted character images.