Multiresolution neural networks for omnifont character recognition

A multiresolution optical character recognition (OCR) using neural networks is proposed for omnifont character recognition. It is motivated by the human reading process in which a low resolution is used to effectively process the majority of clean and unambiguous text, while a more complicated recognition scheme is invoked only when a high resolution is needed. Compared with the method that utilizes single resolution, the multiresolution system not only speeds up recognition by up to 20 times, but also improves accuracy of isolated character recognition from 99.8% to 99.9%. The multiresolution approach captures the essence of better reading, and provides the building blocks for the next-generation OCR systems.<<ETX>>

[1]  Geoffrey E. Hinton,et al.  A time-delay neural network architecture for isolated word recognition , 1990, Neural Networks.

[2]  Mohamad T. Musavi,et al.  A neural network approach to character recognition , 1989, Neural Networks.

[3]  Theodosios Pavlidis,et al.  On the Recognition of Printed Characters of Any Font and Size , 1987, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[4]  Jin Wang,et al.  Resolving multifont character confusion with neural networks , 1993, Pattern Recognit..

[5]  Jin Wang,et al.  Weight smoothing to improve network generalization , 1994, IEEE Trans. Neural Networks.

[6]  Ching Y. Suen,et al.  Automatic recognition of characters by Fourier descriptors and boundary line encodings , 1981, Pattern Recognit..

[7]  Lawrence D. Jackel,et al.  Hardware requirements for neural network pattern classifiers: a case study and implementation , 1992, IEEE Micro.

[8]  Léon Bottou,et al.  Local Learning Algorithms , 1992, Neural Computation.

[9]  Sargur N. Srihari,et al.  High-performance reading machines , 1992 .

[10]  Jin Wang,et al.  Segmentation of merged characters by neural networks and shortest-path , 1993, SAC '93.

[11]  Yann LeCun,et al.  Handwritten zip code recognition with multilayer networks , 1990, [1990] Proceedings. 10th International Conference on Pattern Recognition.

[12]  Charles L. Wilson,et al.  Methods for enhancing neural network handwritten character recognition , 1991, IJCNN-91-Seattle International Joint Conference on Neural Networks.

[13]  Howard Eglowstein,et al.  Tame the paper tiger , 1991 .

[14]  K Fukushima,et al.  Handwritten alphanumeric character recognition by the neocognitron , 1991, IEEE Trans. Neural Networks.

[15]  Qi Tian,et al.  Survey: omnifont-printed character recognition , 1991, Other Conferences.

[16]  S. Ohhashi,et al.  Alphanumeric character recognition using a connectionist model with the pocket algorithm , 1989, International 1989 Joint Conference on Neural Networks.

[17]  R.M.K. Sinha,et al.  Integrating word level knowledge in text recognition , 1990, [1990] Proceedings. 10th International Conference on Pattern Recognition.