Refereed Publications

1. Toru Wakahara, “Global/Local Affine Transformation Correlation for Handwritten Character Recognition as Distortion-Tolerant Matching,” in Proceedings of the 11^th International Conference on Frontiers in Handwriting Recognition, Montreal, August 2008, pp. 141-146.
   Abstract - This paper addresses the problem of distortion-tolerant matching for handwritten character recognition assuming that there is a limited quantity of data. As compared with statistical/probabilistic techniques based on a large sample size, every distortion-tolerant matching technique requires a kind of deterministic models for handwriting deformation. Those models might be parametric or non-parametric. We propose a hierarchical use of global/local affine transformation correlation featuring parametric deformation models that determine optimal affine parameters in global or local areas between input and template images to yield the maximum correlation value. In experiments made on the handwritten numeral database IPTP CDROM1B we prepare only a single template for each digit against a variety of handwriting deformation. Matching experiments have shown that the proposed method greatly increases the correlation value between each input image and its correct category’s template. Also, recognition experiments have achieved a much higher recognition rate of 94.6% than that of 85.8% obtained by the rigid template matching based on a simple correlation.
2. Toru Wakahara, “Figure-Ground Discrimination and Distortion-Tolerant Recognition of Color Characters in Scene Images,” in Proceedings of the 19^th International Conference on Pattern Recognition, Tampa, December 2008.
   Abstract - This paper proposes a new technique of figure-ground discrimination of color characters in scene images following two steps. The first step is temporary binarization by selecting one optimal projection axis in the RGB color space and a threshold value along the axis using Otsu’s criterion as a two-class classification problem. The second step is figure-ground determination based on the figure-to-ground ratio on the image periphery and common characteristics that a character pattern should have. Next, regarding distortion-tolerant character recognition under the condition of a small sample size we compare our global affine transformation (GAT) correlation method against the well-known tangent distance, where both methods use only a single template for each of 62 alphanumeric characters. Experiments are made on a total of 698 character images extracted from the ICDAR 2003 robust OCR dataset. The proposed figure-ground discrimination method achieves a correct binarization rate of 75.3%. Next, in recognition of correctly binarized characters the GAT correlation method and the tangent distance realize correct recognition rates of 94.1% and 91.6%, respectively. Moreover, the GAT correlation method is found to outperform the tangent distance in robustness against rotation at an angle of more than 20 degrees.

Other Publications

1. [Special Talk] Toru Wakahara, “Reconsideration of Deterministic Character Deformation Model,” IEICE Technical Report, PRMU2007-223, pp. 55-60, February 2008.

>PAGE TOP