Professional Documents
Culture Documents
Forty Years of Research in Character and Document Recognition-An Industrial Perspective
Forty Years of Research in Character and Document Recognition-An Industrial Perspective
Pattern Recognition
journal homepage: w w w . e l s e v i e r . c o m / l o c a t e / p r
A R T I C L E I N F O A B S T R A C T
Article history: This paper presents an overview on the last 40-years of technical advances in the field of character and
Received 15 February 2008 document recognition. Representative developments in each decade are described. Then, key technical
Received in revised form 10 March 2008 developments in the specific area of Kanji recognition in Japan are highlighted. The main part of the paper
Accepted 11 March 2008
discusses robustness design principles, which have proven to be effective to solve complex problems
in postal address recognition. Included are the hypothesis-driven principle, deferred decision/multiple-
Keywords: hypotheses principle, information integration principle, alternative solution principle, and perturbation
OCR principle. Finally, future prospects, the 'long-tail' phenomena, and promising new applications are dis-
Character recognition cussed.
Handwriting recognition © 2008 Elsevier Ltd. All rights reserved.
Kanji recognition
Postal address recognition
Robustness design
Information integration
Hypothesis-driven approaches
Digital pen
engines have become ubiquitous and are expanding their reach into
the areas of image documents, photographs, and videos. People are
re-evaluating the importance of handwriting and trying to integrate
it into the digital world. It seems that paper is still not going to dis-
appear. Mobile devices with micro cameras now have CPUs capable
of real-time recognition. The future prospects of these developments
are discussed here.
2.1. Overview
The first practical OCR appeared in the United States in the 1950s,
in the same decade as the first commercial computer UNIVAC. Since
then, each decade has seen advances in OCR technology. In the early
1960s, IBM produced their first models of optical readers, the IBM
1418 (1960) and IBM 1428 (1962), which were, respectively, capable
of reading printed numerals and handprinted numerals. One of the
models of those days could read 200 printed document fonts and
were used as input apparatus for IBM 1401 computers. Also in the Fig. 1. First desktop Hitachi OCR HT-560 (1982).
The International Association for Pattern Recognition began hold- the basic theory had been known, but computers of the 1970s did
ing conferences such as ICDAR, IWFHR, and DAS in the early 1990s. not have sufficient computational power to be applied to studies
Many intensively studied methods have been reported in these con- of such statistical approaches. Today, the four-directional feature
ferences. Examples are artificial neural networks, hidden Markov vector for Kanji patterns consists of 8 × 8 × 4 elements, and the
models (HMMs), polynomial function classifiers, modified quadratic subspace obtained by statistical covariant analysis is of from 100 to
discriminant function (MQDF) classifiers [12], support vector ma- 140 dimensions. However, the size of the 8 × 8 array is surprisingly
chines (SVMs), classifier combination [13--15], information integra- (counter-intuitively) small in light of many complex Kanji characters.
tion, and lexicon-directed character string recognition [16--19], some Recognition accuracy for individual freely handwritten Kanji is not
of which are based on original ideas from the 1960s [20,21]. Most yet high enough, however. Therefore, linguistic context such as name
of these play key roles in today's systems. In contrast with previ- and address is used to enhance total recognition accuracy. To reduce
ous decades, in which industry mostly used proprietary in-house computational cost, cluster-based two-stage classification is used
technology, the 1990s witnessed important interactions between to reduce the number of templates that must be matched. One of
academia and industry. Academics studied real technical problems the recent advances in Kanji (and Chinese character) recognition
and developed sophisticated theory-based methods, enabling indus- is the reduced size of recognition engines designed especially for
try to benefit from their research. Readers may find the state of the mobile phone applications. A compact recognition engine reported
art of character recognition systems, including image preprocessing, in Refs. [35,36] requires only 613 kB of memory to store parameters
feature extraction, pattern classification, and word recognition, well to recognize 4344 classes of printed Chinese characters.
described in the literature [22].
In the following subsections, major pre-1990s technical achieve-
ments in the area of Kanji character classifiers, character segmenta- 2.3. Character segmentation algorithms
tion algorithms, and linguistic processing are described.
In the 1960s and 1970s, a flying-spot scanner or a laser scanner
with a rotated mirror was used together with a photo-multiplier to
2.2. Kanji character classifiers convert optical signals into electrical signals. Character segmenta-
tion was usually carried out with the help of these kinds of scan-
In the 1970s, there were two competing approaches to character ning mechanisms. For example, forms for handprint reading used
recognition, structural analysis and template matching (or the statis- marks on an edge that signaled the presence of a character line to
tical approach). Contemporary commercial OCRs were using struc- be scanned. In addition, the locations of writing boxes on the forms
tural methods to read handprinted alphanumerics and Katakana, and were registered beforehand, and the colors of the boxes were trans-
template matching methods to read printed alphanumerics. Tem- parent to the scanner sensor. Therefore, OCRs could easily extract
plate matching methods had been experimentally proven to be appli- images that contained exactly one single handprinted character.
cable to printed Kanji recognition by the late 1970s [23--26], but their Then, in the 1980s, semiconductor sensors and memories ap-
applicability to handwritten (or handprinted) Kanji was in question. peared, enabling OCRs to scan and store images of whole pages. This
The problem of recognizing handwritten Kanji seemed like a steep, was an epoch making change that was significant to users because
unexplored mountain. It was clear that neither the structural nor it relaxed strict conditions on OCR form specifications, for exam-
the simple template matching approaches could conquer it alone. ple, by enabling them to use smaller non-separated writing boxes.
The former had difficulty with the huge number of topological vari- However, it required a solution of the problem of touching numer-
ations due to complex stroke structures, while the latter had diffi- als and change in how images are represented in memory [37]. Be-
culty with nonlinear shape variations. However, in light of previous fore this change, scanned images had been arrays of binary pixels,
work on handwritten numeral recognition using a template match- and segmentation was pixel-based, but from this time on, the bi-
ing approach, the latter approach seemed to have a greater chance nary image in the memory was represented by run-length codes. The
of success [27]. run-length representation was suited to conducting connected com-
The key was the concept of blurring as feature extraction, which ponent analysis and contour following. The connected components
was applied to directional features and found to be effective in rec- were processed as black objects rather than as pixels. In 1983, Hi-
ognizing handwritten Kanji [27,28]. The introduction of continuous tachi produced one of the first OCRs that could segment and recog-
spatial feature extraction made the optimum amount of blurring sur- nize touching handwritten numerals based on a multiple-hypothesis
prisingly large. The first Hitachi OCR for reading handprinted Kanji segmentation--recognition method (Fig. 3). Contour shape analysis
used simple template matching based on blurred directional fea- was able to identify candidates of touching points, and multiple pairs
tures where the feature templates were four sets of 16 × 16 arrays of of forcedly separated patterns were fed into the classifier. By con-
gray values. The directional feature, which was patented in Japan in sulting the confidence values from the classifier, the recognizer was
1979, was computed using a two-dimensional gradient to determine able to choose the right hypothesis. This direction of changes has
stroke direction (Fig. 2) and was even applicable to grayscale images led us to forms processing whose ultimate goal is to read unknown
[29]. Although it was only indirectly relevant, Hubel and Wiesel's forms, or at least those forms that are not specifically designed for
work encouraged our view that the directional feature was promis- OCRs. However, this means that users might become less careful in
ing [30]. Nonlinear shape normalization [31--33] and statistical clas- their writing, so OCRs have to be more accurate for freely handwrit-
sifier methods [12,34] boosted recognition accuracy. We learned that ten characters as well.
blurring should be considered as a means of obtaining latent dimen- The segmentation problem was far tougher in postal address
sions (subspace) rather than as a means of reducing computational recognition. Fig. 4 shows horizontally handwritten addresses. The
cost, though the effects might seem similar. For example, the mesh width of a character varies by as much as a factor of two, and some
size of 8 × 8 used in statistical approaches was determined by the of the radicals and components are also valid characters. As shown
optimum blurring parameter in light of the Shannon sampling the- in Fig. 4, it is difficult to group the right components to form the
orem, and bigger mesh sizes with the same blurring parameter did right character patterns, where some characters are quite wide and
not give better recognition performances. others narrow. To resolve the grouping problem, linguistic informa-
The thorough studies of the research group led by Prof. Kimura tion (or address knowledge) is required in addition to geometric and
contributed to advancing statistical quadratic classifiers [12], which similarity information. This issue will be discussed in more detail in
were successfully applied to handwritten Kanji recognition. Actually, Section 3.
2438 H. Fujisawa / Pattern Recognition 41 (2008) 2435 -- 2446
1 0 -1
d1
1 0 -1
1 0 -1
1 1 1
d2
0 0 0
-1 -1 -1
Fig. 3. Separation and recognition of touching numerals: (a) only geometric information used; (b) similarity information used in addition to geometric information; and (c)
address information used in addition to geometric and similarity information.
Variability means that no one solution can fit all situations. There-
fore, problems must often be divided into a certain number of cases
with a different solution (problem-solver) to each case. However,
the case to which an input in question belongs is unknown. The
hypothesis-driven principle can be applied in such cases, and the
problem of Japanese address block identification is one such case.
There are six layout types basically, but in real life, there are actually
twelve types because envelopes are sometimes used upside-down.
The approach we take is to choose salient features to distinguish be-
tween such cases and to evaluate the likelihood of each case based
on the observed value of such salient features.
As a general framework of the hypothesis-driven approach, we
Fig. 5. Finite state automaton. call the case a hypothesis and the observed salient features evidence,
and a statistical hypothesis test method may be used to evaluate
likelihood. The a posteriori probability of the k-th hypothesis after
accuracy of 99.1%. Here, we can note that error occurrences are not observing evidence for this hypothesis can be computed as in Eq. (1),
statistically independent. Linguistic processing that solves difficult where Hk represents the k-th hypothesis, and ek the feature vector
segmentation problems (cf. Fig. 4) is discussed in Section 3. for the k-th hypothesis. In Eq. (1), L is a likelihood ratio of hypothesis
Hk to null hypothesis H̄k and is computed as in Eq. (2) assuming
the statistical independence of the features. Functions, P(eki |Hk ) and
3. Robustness design to deal with uncertainty and variability P(eki |H̄k ), can be learned from the training samples.
Fig. 6. Japanese address recognition: rectangles show identified postal code and address blocks.
Table 1
Design principles for robustness
P1 Hypothesis-driven principle When type of a problem is uncertain, set up hypotheses and test them
P2 Deferred decision/multiple hypotheses principle Do not decide; leave decision to next experts carrying over multiple hypotheses
P3a Process integration Solve a problem by multiple different-field experts as a team
P3b Information integration principle Combination-based integration Decide as a team of multiple same-field experts
P3c Corroboration-based integration Utilize other input information; seek more evidence
P4 Alternative solutions principle Solve a problem by multiple alternative approaches
P5 Perturbation principle Modify problem slightly and try again
These functional modules generate multiple hypotheses each of known search methods, we basically use the Hill Climbing Search
which is then forwarded to the next module, which again gener- with backtracking, by which we can reach the optimum solution
ates multiple hypotheses. This process therefore creates the kind in the shortest time. When an optimum branch is rejected at a
of hierarchical tree of hypotheses shown in Fig. 8. The question later stage because it has a confidence value smaller than a preset
here is how to find which optimum branches to follow to reach the threshold, other branches are processed. The use of the Beam Search
best possible answer in the shortest possible time. Among the well- at the later stages effectively boosts the recognition accuracy, while
H. Fujisawa / Pattern Recognition 41 (2008) 2435 -- 2446 2441
Character
Optimum search for
classification Result
interpretation
engines
Fig. 11. Contents of long-term memory: linguistic knowledge in terms of RTN (recursive transition network) generated from a context-free grammar.
application of the multiple-hypotheses principle was not sufficient. sults. Another example of the alternative solutions approach is used
An approach known as the lexicon-directed or lexicon-driven ap- to solve the window noise problem. When existence of window
proach has been developed and can be considered a hypothesis- noise is suspected, two problem-solvers are needed. One attempts to
driven approach, as explained below. The approach is illustrated in eliminate such noise by erosion (thinning) operation, assuming the
Fig. 9, where an input pattern is interpreted by searching for the path shadow is thin or faint. The other attempts to extract line segments
in the presegmentation network (Fig. 10) that best matches the path that form a frame, assuming the shadow is rather solid. These two
in the network that represents linguistic knowledge (Fig. 11). We can problem-solvers are used hoping one will succeed.
say that this is the equivalent of searching for a path in the linguis-
tic network that best matches a path in the presegmented network
[18,19]. This interpretation of the knowledge-directed recognition 3.5. Perturbation principle
process is in line with an explanation given by Simon [43]:
The principle of perturbation is to modify the problem slightly
When it is solving problems in semantically rich domains, a large when it is difficult to solve and to try again to solve it. If pattern
part of the problem-solving search takes place in long-term mem- recognition were such a continuous process, the perturbation prin-
ory and is guided by information discovered in that memory. ciple would not work. In reality, however, it is often a discontinuous
In our case, the long-term memory refers to the linguistic process. Very small modifications may change the final recognition
knowledge, and the short-term memory refers to the presegmented results. It is hoped that the change is from rejection to correct recog-
network. nition or from error to correct recognition. This approach was used
We have developed several versions of such algorithms, one of in the 1980s to recognize handwritten numerals using a structural
which (Fig. 12) was presented by Liu et al. [19]. The recognition rate approach. Because slight topological variations caused rejection, per-
of the lexicon-driven handwritten address recognition algorithm was turbation of parameters or of input images improved the recognition
83.7% with 1.1% error in an experiment, which was done using 3589 rate. In recent years more systematic studies have again shown the
actual mail pieces and a lexicon containing 111 349 address phrases. effectiveness of the approach. Input images are perturbed by various
The linguistic model was represented in the TRIE structure, and the transformations such as morphological (dilation/erosion) and geo-
search was controlled by the Beam Search method. Recognition time metrical transformations (rotation, slanting, perspective, shrinking,
was about 100 ms using a Pentium III/600 MHz machine. and expanding). In Ha and Bunke's work [44], handwritten numerals
were transformed in twelve ways and recognized using the frame-
work of classifier combination. Their approach recognized difficult,
3.4. Alternative solutions principle eccentric handwriting better than classical classifiers such as k-NN
and neural network. By the way, blurring is one of image transfor-
There are many image level problems including touching charac- mations but has not been applied in the context of perturbation.
ters, touching underlines, window shadow noise, cancellation stamps Blurring used in character feature extraction is not the kind of 'slight
covering/touching address characters, and so on. The alternative so- transformation'.
lutions approach is to provide more than one solution to a problem. It The perturbation approach has also been successfully applied
effectively provides solutions that are complementary to each other. to Japanese postal address recognition. Our test of the approach
For example, the problem of touching characters may be solved using achieved about 10--15 percentage point improvements in recogni-
a holistic approach or a forced separation (dichotomizing) approach. tion rates on the average. When we did not set limits on recog-
Especially when dealing with numerals, a pair of touching numer- nition time and repeated more perturbation operations including
als can be treated as one character out of 100 classes. Training such rotational transformation, re-binarization, and some other paramet-
holistic classifiers enables the results of the holistic and dichotomiz- ric modifications in sequence, we found that 53% of rejected images
ing classifiers to be merged producing more reliable recognition re- were correctly recognized with a 12% error rate. Although the result
H. Fujisawa / Pattern Recognition 41 (2008) 2435 -- 2446 2443
0.4
Communication unit
Battery
Memory
Camera
HDMS / [1] H. Fujisawa, A view on the past and future of character and document
ODPS recognition, in: Proceedings of the Seventh ICDAR, Curitiba, Brazil, September
2007, pp. 3--7.
[2] The United States Postal Service: An American History 1775--2002, Government
Creation and reuse Relations, United States Postal Service, 2003.
Read and write [3] H. Genchi, K. Mori, S. Watanabe, S. Katsuragi, Recognition of handwritten
Original documents numeral characters for automatic letter sorting, Proc. IEEE 56 (1968)
Edited documents 1292--1301.
PDFs [4] W. Schaaf, G. Ohling, et al., Recognizing the Essentials, Siemens ElectroCom,
Printing conditions Konstanz, 1997.
[5] http://www.industry.siemens.com/postal-automation/usa/.
Annotations
[6] K. Yamamoto, H. Yamada, T. Saito, I. Sakaga, Recognition of handprinted
Metadata characters in the first level of JIS Chinese characters, in: Proceedings of the
Retrieval Annotations Eighth ICPR, 1986, pp. 570--572.
[7] J.R. Ullmann, Pattern Recognition Techniques, Butterworths, London, 1973.
[8] C.Y. Suen, M. Berthod, S. Mori, Automatic recognition of handprinted
characters---The state of art, Proc. IEEE 68 (4) (1980) 469--487.
[9] S. Mori, C.Y. Suen, K. Yamamoto, Historical review of OCR research and
development, Proc. IEEE 80 (7) (1992) 1029--1058.
[10] G. Nagy, At the frontiers of OCR, Proc. IEEE 80 (7) (1992) 1093--1100.
Who/when printed? [11] J.C. Simon, Off-line cursive word recognition, Proc. IEEE 80 (7) (1992)
1150--1161.
Who/when annotated?
[12] F. Kimura, K. Takashina, S. Tsuruoka, Y. Miyake, Modified quadratic discriminant
functions and the application to Chinese character recognition, IEEE Trans.
Fig. 16. iJIT system that manages hybrid documents containing handwritten anno- PAMI 9 (1) (1987) 149--153.
tations. [13] C.Y. Suen, C. Nadal, T.A. Mai, R. Legault, L. Lam, Recognition of totally
unconstrained handwritten numerals based on the concept of multiple experts,
in: Proceedings of the First IWFHR, Montreal, Canada, 1990, pp. 131--143.
[14] L. Xu, A. Krzyzak, C.Y. Suen, Methods of combining multiple classifiers and
their applications to handwriting recognition, IEEE Trans. SMC 22 (3) (1992)
practice, it will be a natural demand to ask for capabilities of hand-
418--435.
written character recognition, handwritten query processing, and [15] T.K. Ho, J.J. Hull, S.N. Srihari, Decision combination in multiple classifier systems,
more intelligent knowledge management. Effort to create informa- IEEE Trans. PAMI 16 (1) (1994) 66--75.
[16] F. Kimura, M. Sridhar, Z. Chen, Improvements of lexicon-directed algorithm
tion systems that would require recognition technology is a way
for recognition of unconstrained hand-written words, in: Proceedings of the
that we may pursue. We hope more advanced information systems Second ICDAR, Tsukuba, Japan, October 1993, pp. 18--22.
require more advanced recognition technology. [17] C.H. Chen, Lexicon-driven word recognition, in: Proceedings of the Third ICDAR,
Montreal, Canada, August 1995, pp. 919--922.
[18] M. Koga, R. Mine, H. Sako, H. Fujisawa, Lexical search approach for character-
5. Conclusion string recognition, in: Proceedings of the Third DAS, Nagano, Japan, November
1998, pp. 237--251.
[19] C.-L. Liu, M. Koga, H. Fujisawa, Lexicon-driven segmentation and recognition of
Vision and fundamental technologies are both key to the future handwritten character strings for Japanese address reading, IEEE Trans. PAMI
of our technical community. Vision takes the form of forecasted ap- 24 (11) (2002) 425--1437.
[20] M. Aizermann, E. Braverman, L. Rozonoer, Theoretical foundations of the
plications with new value propositions. For investment to be made potential function method in pattern recognition learning, Automat. Remote
in new technology, such new propositions need to be attractive Control 25 (1964) 821--837.
to many people or at least to some innovative people. This is a [21] U. Miletzki, Schürmann-polynomials---roots and offsprings, in: Proceedings of
the Eighth IWFHR, 2002, pp. 3--10.
top--down approach to innovation. Fundamental technologies may [22] M. Cheriet, N. Kharma, C.-L. Liu, C.Y. Suen, Character Recognition Systems---A
start innovation from the bottom as well. Here, the technologies we Guide for Students and Practitioners, John Wiley & Sons, Inc., Hoboken, NJ,
are discussing have two parts: one is the technology that supports 2007.
[23] R. Casey, G. Nagy, Recognition of printed Chinese characters, IEEE Trans.
our community from the bottom; the other is the technology of Electron. Comput. EC-15 (1) (1966) 91--101.
our own, i.e. character and document recognition. For the first part, [24] S. Yamamoto, A. Nakajima, K. Nakata, Chinese character recognition by
we have seen impacts of advanced semiconductor devices, high- hierarchical pattern matching, in: Proceedings of the First IJCPR, Washington,
DC, 1973, pp. 183--194.
performance computers, and more advanced software development [25] H. Fujisawa, Y. Nakano, Y. Kitazume, M. Yasuda, Development of a Kanji OCR:
tools, which have supported the advances in recognition technology. an optical Chinese character reader, in: Proceedings of the Fourth IJCPR, Kyoto,
They not only enabled more advanced OCR systems on the surface, November 1978, pp. 815--820.
[26] G. Nagy, Chinese character recognition: a twenty-five-year retrospective, in:
but also invited and promoted more academia into this community, Proceedings of the Ninth ICPR, 1988, pp. 163--167.
which have also contributed to the advances in recognition tech- [27] M. Yasuda, H. Fujisawa, An Improvement of Correlation Method for Character
nology. We would like to see this kind of virtuous cycles happen Recognition, vol. 10 (2), Systems, Computers, Controls, Scripta Publishing Co.,
1979, pp. 29--38.
forever. [28] H. Fujisawa, C.-L. Liu, Directional pattern matching for character recognition
revisited, in: Proceedings of the Seventh ICDAR, Edinburgh, August 2003, pp.
794--798.
Acknowledgments [29] H. Fujisawa, O. Kunisaki, Method of pattern recognition, Japanese Patent
1,520,768 granted in 1989, filed in 1979.
[30] D.H. Hubel, T.N. Wiesel, Functional architecture of macaque monkey visual
The author is grateful to the members of his research team in cortex, Proc. R. Soc. London Ser. B 198 (1977) 1--59.
Hitachi who worked on development of the postal address recogni- [31] J. Tsukumo, H. Tanaka, Classification of handprinted Chinese characters using
tion system: H. Sako, K. Marukawa, M. Koga, H. Ogata, H. Shinjo, K. non-linear normalization and correlation methods, in: Proceedings of the Ninth
ICPR, Rome, Italy, 1988, pp. 168--171.
Nakashima, H. Ikeda, T. Kagehiro, R. Mine, N. Furukawa, and T. Taka- [32] C.-L. Liu, Normalization-cooperated gradient feature extraction for handwritten
hashi. He is also grateful to Dr. C.-L. Liu at the Institute of Automa- character recognition, IEEE Trans. PAMI 29 (6) (2007) 1465--1469.
tion of the Chinese Academy of Sciences, Beijing, and Prof. Y. Shima [33] C.-L. Liu, Handwritten Chinese character recognition: effects of shape
normalization and feature extraction, in: Proceedings of the Summit on Arabic
at Meisei University, Tokyo, for the work they did at our laboratory. and Chinese Handwriting, College Park, September 2006, pp. 23--27.
The author also thanks Prof. G. Nagy of Rensselaer Polytechnic Insti- [34] K. Jain, R.P.W. Duin, J. Mao, Statistical pattern recognition: a review, IEEE Trans.
tute for his valuable discussions and comments on this manuscript. PAMI 22 (1) (2000) 4--37.
[35] C.-L. Liu, R. Mine, M. Koga, Building compact classifier for large character
Thanks also go to Dr. U. Miletzki of Siemens ElectroCom for provid- set recognition using discriminative feature extraction, in: Proceedings of the
ing information regarding their historical work. Eighth ICDAR, Seoul, Korea, 2005, pp. 846--850.
2446 H. Fujisawa / Pattern Recognition 41 (2008) 2435 -- 2446
[36] M. Koga, R. Mine, T. Kameyama, T. Takahashi, M. Yamazaki, T. Yamaguchi, [44] T.M. Ha, H. Bunke, Off-line, handwritten numeral recognition by perturbation
Camera-based Kanji OCR for mobile-phones: practical issues, in: Proceedings method, IEEE Trans. PAMI 19 (5) (1997) 535--539.
of the Eighth ICDAR, Seoul, Korea, 2005, pp. 635--639. [45] H. Fujisawa, H. Sako, Y. Okada, S-W. Lee, Information capturing camera
[37] H. Fujisawa, Y. Nakano, K. Kurino, Segmentation methods for character and developmental issues, in: Proceedings of the Fifth ICDAR'99, Bangalore,
recognition: from segmentation to document structure analysis, Proc. IEEE 80 September 1999, pp. 205--208.
(7) (1992) 1079--1092. [46] F. Guimbretière, Paper augmented digital documents, in: Proceedings of the
[38] K. Marukawa, M. Koga, Y. Shima, H. Fujisawa, An error correction algorithm ACM Symposium on User Interface Software and Technology, UIST2003,
for handwritten Chinese character address recognition, in: Proceedings of the Vancouver, Canada, 2003, pp. 51--60.
First ICDAR, Saint-Malo, September 1991, pp. 916--924. [47] C. Liao, F. Guimbretière, PapierCraft: a command system for interactive paper,
[39] H. Fujisawa, How to deal with uncertainty and variability: experience and in: Proceedings of the ACM Symposium on User Interface Software and
solutions, in: Proceedings of the Summit on Arabic and Chinese Handwriting, Technology, UIST2005, Seattle, USA, 2005, pp. 241--244.
College Park, September 2006, pp. 29--39. [48] R. Yeh, C. Liao, S. Klemmer, F. Guimbretière, B. Lee, B. Kakaradov, J. Stamberger,
[40] H. Fujisawa, Robustness design of industrial strength recognition systems, in: A. Paepcke, ButterflyNet: a mobile capture and access system for field biology
B.B. Chaudhuri (Ed.), Digital Document Processing: Major Directions and Recent research, in: Proceedings of the International Conference on Computer--Human
Advances, Springer, London, 2007, pp. 185--212. Interaction, CHI2006, Montreal, Canada, 2006, pp. 571--580.
[41] T. Kagehiro, H. Fujisawa, Multiple hypotheses document analysis, in: S. Marinai, [49] H. Ikeda, K. Konishi, N. Furukawa, iJITinOffice: desktop environment enabling
H. Fujisawa (Eds.), Studies in Computational Intelligence, vol. 90, Springer, integration of paper and electronic documents, in: Proceedings of the ACM
Berlin, Heidelberg, 2008, pp. 277--303. Symposium on User Interface Software and Technology, UIST2006, Montreux,
[42] T. Kagehiro, M. Koga, H. Sako, H. Fujisawa, Segmentation of handwritten Switzerland, October 2006.
Kanji numerals integrating peripheral information by Bayesian rule, [50] A.J. Sellen, R.H. Harper, The Myth of the Paperless Office, The MIT Press,
in: Proceedings of the IAPR MVA'98, Chiba, Japan, November 1998, Cambridge, MA, 2001.
pp. 439--442.
[43] H.A. Simon, The Sciences of the Artificial, third ed., The MIT Press, Cambridge,
MA, 1998, pp. 87--88.
About the Author---HIROMICHI FUJISAWA is a Corporate Chief Scientist at the Research and Development Group of Hitachi, Ltd., and is an
advisor to the Global Standardization Office of Hitachi. He joined its Central Research Laboratory in 1974. Since then, he has engaged in
research and development for handwritten character recognition, document understanding, and document retrieval, as applied to business
OCR systems, mail sorting machines, e-Government systems, etc. He was a visiting scholar at Carnegie Mellon University's Computer Science
Department in 1980--1981 and at Stanford University's Computer Science Department in 2005--2006. He is a Fellow of the IEEE, Fellow of the
International Association for Pattern Recognition (IAPR), and Fellow of the Institute for Electronics, Information and Communication Engineers,
Japan (IEICE). He is also a recipient of the IAPR/ICDAR's Outstanding Achievements Award in 2007 for contributions to industrial document
analysis and to the ICDAR community. Fujisawa received the Dr. Eng. degree in Electrical Engineering from Waseda University in 1975.