Best Practices for Convolutional Neural Networks Applied to Visual Document Analysis
ICDAR '03 Proceedings of the Seventh International Conference on Document Analysis and Recognition - Volume 2
ICDAR 2003 Robust Reading Competitions
ICDAR '03 Proceedings of the Seventh International Conference on Document Analysis and Recognition - Volume 2
Convolutional Face Finder: A Neural Architecture for Fast and Robust Face Detection
IEEE Transactions on Pattern Analysis and Machine Intelligence
ICDAR '05 Proceedings of the Eighth International Conference on Document Analysis and Recognition
Camera based Degraded Text Recognition Using Grayscale Feature
ICDAR '05 Proceedings of the Eighth International Conference on Document Analysis and Recognition
Text Recognition of Low-resolution Document Images
ICDAR '05 Proceedings of the Eighth International Conference on Document Analysis and Recognition
ICPR '06 Proceedings of the 18th International Conference on Pattern Recognition - Volume 02
Scene Text Recognition Using Similarity and a Lexicon with Sparse Belief Propagation
IEEE Transactions on Pattern Analysis and Machine Intelligence
ICDAR '09 Proceedings of the 2009 10th International Conference on Document Analysis and Recognition
A text reading algorithm for natural images
Image and Vision Computing
Text extraction from natural scene image: A survey
Neurocomputing
| Hi-index | 0.00 |
In this paper, a natural scene character recognition method using convolutional neural network(CNN) and bimodal image enhancement is proposed. CNN based grayscale character recognizer has strong tolerance to degradations in natural scene images. Since character image is bimodal pattern image in essence, bimodal image enhancement is adopted to improve the performance of CNN classifier. Firstly, a maximum separability based color-to-gray method is used to strengthen the discriminative power in grayscale image space. Secondly, grayscale distribution normalization based on histogram alignment is performed. Through increasing the data consistency among grayscale training and test samples, it leads to a better CNN classifier. Thirdly, a shape holding grayscale character image normalization is adopted. Based on these measures, a high performance natural scene character recognizer is constructed. The recognition rate of 85.96% on ICDAR 2003 robust OCR dataset is higher than existing works, which verified the effectiveness of the proposed method.