Authors: Y. C. Wu, K. Chen, X. He, Z. Chen, F. Yin, C. L. Liu

Description: For text localization, the image is segmented into smooth and non-smooth regions based on local contrast. After removing non-text smooth regions, the remaining smooth regions are merged with non-smooth regions to form a candidate text image, which is binarized into high-value and low-value connected components (CCs). The CCs undergo CC filtering, line grouping and line classification to give the localized text lines, which are further segmented into words according to the between-component gap.
In text word recognition, the word image is first over-segmented into primitive segments using an MLP with 968-D features for candidate cut
classification. Based on over-segmentation, the word image first undergoes lexicon-driven recognition [1]. When it is rejected by lexicon-driven recognition (judged as beyond the lexicon), it then undergoes lexicon-free recognition with a statistical language model [2]. After word recognition, we analyze the result to correct the case of letters.

References:
[1] C.-L. Liu, M. Koga, H. Fujisawa, Lexicon-driven segmentation and recognition of handwritten character strings for Japanese address reading, IEEE Trans. Pattern Analysis and Machine Intelligence, 24(11): 1425-1437, 2002.
[2] Q.-F. Wang, F. Yin, C.-L. Liu, Handwritten Chinese text recognition by integrating multiple contexts, IEEE Trans. Pattern Analysis and Machine Intelligence, 34(8): 1469-1481, 2012.