Hiromichi Fujisawa

Handwritten digit recognition: benchmarking of state-of-the-art techniques

This paper presents the results of handwritten digit recognition on well-known image databases using state-of-the-art feature extraction and classification techniques. The tested databases are CENPARMI, CEDAR, and MNIST. On the test data set of each database, 80 recognition accuracies are given by combining eight classifiers with ten feature vectors. The features include chaincode feature, gradient feature, profile structure feature, and peripheral direction contributivity. The gradient feature is extracted from either binary image or gray-scale image. The classifiers include the k-nearest neighbor classifier, three neural classifiers, a learning vector quantization classifier, a discriminative learning quadratic discriminant function (DLQDF) classifier, and two support vector classifiers (SVCs). All the classifiers and feature vectors give high recognition accuracies. Relatively, the chaincode feature and the gradient feature show advantage over other features, and the profile structure feature shows efficiency as a complementary feature. The SVC with RBF kernel (SVC-rbf) gives the highest accuracy in most cases but is extremely expensive in storage and computation. Among the non-SV classifiers, the polynomial classifier and DLQDF give the highest accuracies. The results of non-SV classifiers are competitive to the best ones previously reported on the same databases.

Handwritten digit recognition: investigation of normalization and feature extraction techniques

The performance evaluation of various techniques is important to select the correct options in developing character recognition systems. In our previous works, we have proposed aspect ratio adaptive normalization (ARAN) and have evaluated the performance of state-of-the-art feature extraction and classification techniques. For this time, we will propose some improved normalization functions and direction feature extraction strategies and will compare their performance with existing techniques. We compare ten normalization functions (seven based on dimensions and three based on moments) and eight feature vectors on three distinct data sources. The normalization functions and feature vectors are combined to produce eighty classification accuracies to each dataset. The comparison of normalization functions shows that moment-based functions outperform the dimension-based ones and the aspect ratio mapping is influential. The comparison of feature vectors shows that the improved feature extraction strategies outperform their baseline counterparts. The gradient feature from gray-scale image mostly yields the best performance and the improved NCFE (normalization-cooperated feature extraction) features also perform well. The combined effects of normalization, feature extraction, and classification have yielded very high accuracies on well-known datasets.

Segmentation methods for character recognition: from segmentation to document structure analysis

A pattern-oriented segmentation method for optical character recognition that leads to document structure analysis is presented. As a first example, segmentation of handwritten numerals that touch are treated. Connected pattern components are extracted, and spatial interrelations between components are measured and grouped into meaningful character patterns. Stroke shapes are analyzed and a method of finding the touching positions that separates about 95% of connected numerals correctly is described. Ambiguities are handled by multiple hypotheses and verification by recognition. An extended form of pattern-oriented segmentation, tabular form recognition, is considered. Images of tabular forms are analyzed, and frames in the tabular structure are extracted. By identifying semantic relationships between label frames and data frames, information on the form can be properly recognized. >

Lexicon-driven segmentation and recognition of handwritten character strings for Japanese address reading

This paper describes a handwritten character string recognition system for Japanese mail address reading on a very large vocabulary. The address phrases are recognized as a whole because there is no extra space between words. The lexicon contains 111,349 address phrases, which are stored in a trie structure. In recognition, the text line image is matched with the lexicon entries (phrases) to obtain reliable segmentation and retrieve valid address phrases. The paper first introduces some effective techniques for text line image preprocessing and presegmentation. In presegmentation, the text line image is separated into primitive segments by connected component analysis and touching pattern splitting based on contour shape analysis. In lexicon matching, consecutive segments are dynamically combined into candidate character patterns. An accurate character classifier is embedded in lexicon matching to select characters matched with a candidate pattern from a dynamic category set. A beam search strategy is used to control the lexicon matching so as to achieve real-time recognition. In experiments on 3,589 live mail images, the proposed method achieved correct rate of 83.68 percent while the error rate is less than 1 percent.

Discriminative learning quadratic discriminant function for handwriting recognition

In character string recognition integrating segmentation and classification, high classification accuracy and resistance to noncharacters are desired to the underlying classifier. In a previous evaluation study, the modified quadratic discriminant function (MQDF) proposed by Kimura et al. was shown to be superior in noncharacter resistance but inferior in classification accuracy to neural networks. This paper proposes a discriminative learning algorithm to optimize the parameters of MQDF with aim to improve the classification accuracy while preserving the superior noncharacter resistance. We refer to the resulting classifier as discriminative learning QDF (DLQDF). The parameters of DLQDF adhere to the structure of MQDF under the Gaussian density assumption and are optimized under the minimum classification error (MCE) criterion. The promise of DLQDF is justified in handwritten digit recognition and numeral string recognition, where the performance of DLQDF is comparable to or superior to that of neural classifiers. The results are also competitive to the best ones reported in the literature.

Performance evaluation of pattern classifiers for handwritten character recognition

Abstract. This paper describes a performance evaluation study in which some efficient classifiers are tested in handwritten digit recognition. The evaluated classifiers include a statistical classifier (modified quadratic discriminant function, MQDF), three neural classifiers, and an LVQ (learning vector quantization) classifier. They are efficient in that high accuracies can be achieved at moderate memory space and computation cost. The performance is measured in terms of classification accuracy, sensitivity to training sample size, ambiguity rejection, and outlier resistance. The outlier resistance of neural classifiers is enhanced by training with synthesized outlier data. The classifiers are tested on a large data set extracted from NIST SD19. As results, the test accuracies of the evaluated classifiers are comparable to or higher than those of the nearest neighbor (1-NN) rule and regularized discriminant analysis (RDA). It is shown that neural classifiers are more susceptible to small sample size than MQDF, although they yield higher accuracies on large sample size. As a neural classifier, the polynomial classifier (PC) gives the highest accuracy and performs best in ambiguity rejection. On the other hand, MQDF is superior in outlier rejection even though it is not trained with outlier data. The results indicate that pattern classifiers have complementary advantages and they should be appropriately combined to achieve higher performance.

Effects of classifier structures and training regimes on integrated segmentation and recognition of handwritten numeral strings

In integrated segmentation and recognition of character strings, the underlying classifier is trained to be resistant to noncharacters. We evaluate the performance of state-of-the-art pattern classifiers of this kind. First, we build a baseline numeral string recognition system with simple but effective presegmentation. The classification scores of the candidate patterns generated by presegmentation are combined to evaluate the segmentation paths and the optimal path is found using the beam search strategy. Three neural classifiers, two discriminative density models, and two support vector classifiers are evaluated. Each classifier has some variations depending on the training strategy: maximum likelihood, discriminative learning both with and without noncharacter samples. The string recognition performances are evaluated on the numeral string images of the NIST special database 19 and the zipcode images of the CEDAR CDROM-1. The results show that noncharacter training is crucial for neural classifiers and support vector classifiers, whereas, for the discriminative density models, the regularization of parameters is important. The string recognition results compare favorably to the best ones reported in the literature though we totally ignored the geometric context. The best results were obtained using a support vector classifier, but the neural classifiers and discriminative density models show better trade-off between accuracy and computational overhead.

An algorithm for the skew normalization of document image

An algorithm to normalize the skew of document images is proposed. The skew angle is detected in two stages. In the first stage, connected regions in an image are extracted and some feature parameters are extracted for each region. In the second stage, the Hough transform is calculated for the parameters, and the angle which gives the minimum of the transform is estimated as the skew angle. In experiments using CCITT standard documents, a detection accuracy of less than 0.1 degrees is obtained for printed documents. When graphical elements are included in the documents in addition to printed characters, the accuracy deteriorates to 0.2 degrees .<<ETX>>

Gabor feature extraction for character recognition: comparison with gradient feature

Gabor filter feature has been applied to character recognition but was not compared with the best direction feature: gradient feature. In this paper, we propose a principled method for implementing Gabor filters for character feature extraction and compare the recognition performances of Gabor feature and gradient feature on three databases. The results show that Gabor filters with low orientation sensitivity and broad frequency band favor recognition accuracy. The Gabor feature performs comparably or better than the gradient feature on two of the three databases, but is inferior on the rest one.

Handwritten Chinese character recognition: alternatives to nonlinear normalization

Nonlinear normalization (NLN) by line density equalization has been popularly used in handwritten Chinese character recognition (HCCR). To overcome the intensive computation of local line density and the excessive shape distortion of NLN, we tested some alternative methods based on global transformation, including a moment-based linear transformation and two nonlinear methods based on quadratic curve fitting. The alternative methods are simpler in computation and the transformed images have more natural shapes. In experiments of HCCR on large databases, the alternative methods have yielded comparable or higher accuracies to the traditional NLN.