Sparse Discriminative Information Preservation for Chinese character font categorization

• We propose a new dimension reduction method termed Sparse Discriminative Information Preservation (SDIP) for Chinese character font recognition.
• The scheme applies LBP descriptor based Chinese character interesting points for representing font information.
• Experimental results demonstrate that SDIP is more effective.

With the rapid development of optical character recognition (OCR), font categorization becomes more and more important. This is because font information has very wide usage and researchers came to know this point recently. In this paper, we propose a new scheme for Chinese character font categorization (CCFC), which applies LBP descriptor based Chinese character interesting points for representing font information. Specifically, it classifies Chinese character font through the cooperation between a new Sparse Discriminative Information Preservation (SDIP) for feature selection and NN classifier. SDIP focus three aspects as follows: (1) it preserves the local geometric structure of the intra-class samples and maximizes the margin between the inter-class samples on the local patch simultaneously; (2) it models the reconstruction error to preserve the prior information of the data distribution; and (3) it introduces the L1-norm penalty to achieve the sparsity of the projection matrix. We conduct experiments on our new collect text block images which include 25 popular Chinese fonts. The average recognition demonstrates the robustness and effectiveness of SDIP for CCFC.