Improved Descriptor for Chinese Character Recognition
Jianbo Chen · Information Security and Communications Privacy · 2010
The paper proposes a new local descriptor for Chinese character recognition,which adds contrast gray histogram to SIFT descriptor.The proposed descriptor could be used in feature description of Chinese characters.The method of test system in this paper treats the character recognition as the image matching problem.To detect the property of the new descriptor,Chinese character recognition is tested under the four situations of affine,noise,background and complicated circumstance,and the right rate of Chinese character recognition is compared between the proposed descriptor and SIFT.Experiments indicate that the proposed descriptor could describe more local information and perform better than SIFT in Chinese character recognition.