作者: M. Sawaki , H. Murase , N. Hagita
关键词:
摘要: Presents a multiple-dictionary method for recognizing low-quality characters in scene images. First, the environmental conditions of an input image are estimated using initial dictionary. Then, relevant dictionary from multiple dictionaries reflecting different is automatically selected estimation and used recognition. Experiments made images bookshelves. The results show that proposed achieves higher recognition rate (89.8%) than obtained by single (76.4%). Furthermore, accuracy improves 89.8% to 95.2% contextual postprocessing.