
Computer Engineering
1999ꡡ259ڡVol.25 No.9  1999



˻ַͼĹǼܷ̬ʶ𷽷
¬֡쿡л
ժҪģϵ˻ַͼйǼܷṹ̬ʶ𷽷ֱӦ˻ַģͼҶͼ˶УԱʻȱַȷȡǼܣַʱܸݹǼ̬ܵṹ׵ȷзλã޼㷨ԸзֵĴַǼֱܿͨһ
ؼʣ˻ַͼģϡǼܡ㷨ַʶ
Topographic Approach Recognizer for Skeletonization of Regraded Character Images Based on Fuzzy Set
Lu Da Pu Wei
(Changshu College, Changshu 215500)
Xie Mingpei
(Department of Computer Science, Fudan University, Shanghai 200433)
AbstractA topographic approach recognizer is proposed for skeletonization of regraded character images based on fuzzy set. The method works directly on fuzzy or gray-scale images of regradedcharacted, so that it is less sensitive to noise. It can correctly extract the whole skeleton although a character is brocken into pieces. For some characters merged togther, the breaking positions can be easily located by topographic approaching for skeletons. A splitting algorithm is then applied on each breaking position. The separated skeletons can be directly passed through a classifier wirthout and further process.
Key wordsDegraded character images, Fuzzy set, Skeleton, Splitting algorithm, Character recognition.
˻ַȡʶǹѧַʶ(OCR)͹ͼַԶʶеһҪ档ǻڶֵͼϸ˴ֱͶӰ淨[1]Ǽ[2][3]ֶϴ۷[4]ȶʶ𷽷ЩĳЩ淶ض˻ַȡýϺõʶЧЩҪص˻ַʶϣûгַֿͼ˻ԭʶ֮ǰ˻ַͼЧͼԭ⣬ЩӦڻҶͼģͼ󣬱ûҶȡзöֵͼָͨ̶ȵ½Ӱ˻ַʶʡ
1ģͼı任
Ϊ˻ַȡʶģıͼ任Ƕ˻ַͼѰһʵı任˵ѧ⣬һɷ˻ַͼ˻ľԭڴ˻ѧ"ȥ˻"Ըı˻ַͼĳЩַͼʵĿ
f(x,y)Ϊһģͼµɢڱ任
h(x,y)=C(x)C(y)
(1)

(2)
f(x,y)ı任g(x,y)ʾ
(3)

(4)
Ϊһʽ(1)(2)(4)(3)ɵã
(5)
ͣ
g(x,y)=0



ͼ1ص
ֵͼеһؿɿǾ"1"ֵ"0"ֵդѡȡ-3Ϊһͼ1ʾһդϽΪԭ㣬(飬)Ϊ()f(x,y)h(x,y)ľ£
(6)
Цi+m,j+nʾ(i+m,j+n)ֵ

ͬأgij(x,y)һסƫƵ£
(7)
(8)
(9)
(10)
2˻ַͼĹǼܴ
2.1ȱַ
ڱʻȱַһǽɷϳһͼ󣬸ͼں׶δһֱһ壻һǰѸɷֿԶĲ֣ͨһĹĽܵĴڵһ֡ͼ2(a)Ϊַһѱʻάͼμͼ2(b) ZʾҶȵȼΪԤľɺѲӣƽͼͼ2(c)кߴͨεȡĹǼܡ

ͼ2 (a)ַһѱʻ (b)Ӧάͼ
(c)ƽͼ͹Ǽ
㷨ɽϿھΪ2ҵѲͬʱҲַܴ䲻ӦеӡˣֵĴСֱӰȱַǼܵȡ˻ַʶȷʣֵѡȡӦֿǱַͼиַʻĿȺַ֮ļ롣bΪַϸʻĿȣdΪַСd<hȡ<h/2ӳɷֵӾϸʻȣd<hȡ<d/2ʱСַӡˢַȱַĲΪӡˢ豸ȱݺӡˢַıһdbҵĴɻýЧ
2.2 ַķ롢޼㷨
ַַʻнٵؼϵ͵ֵͼ任㴦ƽֵԽϵͣϸıʻҲͬнϵ͵ֵߵĲڽϸʻֵ仯ƽΧֵı仯ʮּˣĳ㸽ߵбʱ仯ҾСֵõΪʣ仰λΪиֵ߱İ㡣
Ľַܵķ롢޼㷨
(1)㡡һֱǰĹַ̡ıʻܰһϵеļε㣬˿˳ĳһʻȥ١ݰҸбʣĳߵбΪ㣬ҸõƽڼߵʸصıֵΪֵȷõΪ㡣㲢ȷλڼε㣬ǿܴĳһ߶ΡΪ˿ɼΪμ߶εͷβ˵бʣ˵߾ָ⣬аڸ߶ڡѧʾ
.㣺(1)g0(2)w0
򰰵㾫ȷλڼε㡣Vgwֱʾεбʼõƽڼߵʸصıֵ
.㣺
򰰵λڸü߶ϡVg1Vg2 W1W2ֱʾ߶˵бʼƽڼߵʸصıֵ(1)g1.g20(2)W10W20
(2)ȷзֵ㡡жϿзֵ㣺㱾ֵֵð㶨Ϊзֵ㣻ڶԸзֵıֵӦзֵ㴦صֵֵӳзֵ㴦صзִֵۣ͵ĵзֿԸߡֵȣ鱾ֵֵӳзֿзִֵۣзִ۵ͣۼзֿɿȡ
(3)޼Ǽߡжͣȷзֵ㴦ǼӦɾǲɾжзֵֶ߶εĳ̣ȷǼߵ޼
Kahan[2]Ϊ֣߽˫O㣬ͼ3(a)(b)ʾ㷨Ӧھ˫O ַʱ㴦һıʻͼ3(e)еĵߣöʻκһַ˱ɾ߽㴦ĹǼһַֿĳһʻͼ3(d)˽貿޼ͽ㴦ܴ
ڵĶǼĳһַһ֡ͼ3(f)ʾĶǼַ"y"һ֣һӣ󷽵ɾҷֻ貿޼

ͼ3 (a)߽ (b)˫O (c)
(d)(e)(f)ֱ(a)(b)(c)ĹǼ
ԶǼ޼׼Ϊһзֵıʻڰ㴦зΪΣֶϲֱַ顣һν϶̣һĩ˴"T"ͽ㣬ý"O"㣬"O"ͽ˫"O"һʻӴ͹״ߣͼ3(f)˸òȫɾϳֶϲһ˵㲻Ƕ"T"ͽ㣬ⲿֲ"O"ͽ㣬ʱ÷ֶϲʵַĹǼߺͲֶǼߡ޼ĹǼߣֻӰ㿪ʼֱֱֵ㽻㣬Ӱ㿪ʼֵ㽻߶μΪӦɾĶǼߡ

ͼ4 (a)жϵȱʡַ (b)ģͼ任 (c)ȱʡַĹǼ
(4)ַʶȷϡ޼ַ̽ĹǼܱзֳɼַĹǼܡĿǰз㷨ÿзֵļͨзֵַͼȡԵһ֤ʵ⽫Ӵʱ䣬޼̽󣬲κδÿַǼַܾͨ⣬ͨȷзּĿɿԡȷǻڰַʵȷһԣɿֵͨĴʳ԰ַʵá
3
ڱĽܵ˻ַȡʶ𷽷ʵUNIXϵͳ CԱʵֵġʵ28600ַӡˢȾַΪ212ʻжϿڵȱַΪ103ʵȲúʵ˲ַͼƽ˲ͼ(4)ͼ(5)ֱΪʵַеļȱַַĴģͼ任99ȱַȱڴѱԽСֵӣȱַͼĻָʴ96.16%ȻַͼñĽܵĹǼ̬ṹͷֶϴ۷ַʶ𡢱Ƚϣʵ1ʾ

ͼ5(a)ַ (b)ģͼ任 (c)ַĹǼ
1ʵ

ʶ𷽷ٻ()ʶ()
Ǽ̬ṹ92.8379.24
ֶϴ۷88.7668.22

еٻʶ𷽷ַȡΪ
ٻ=(ȡַ/ַͼַܵ).100%
еʶʶ𷽷ַʶΪ
ʶ=(ȷʶַ/ַͼַܵ).100%
4ۺͽ
Ľܵʶ𷽷¼ŵ㣺(1)ֱӦַģͼҶͼ󣬶У(2)ͨѡúʵƽ˲ģıͼԤ˻ַʶʣ(3)ַĽܵķ㷨зֺ󣬶ԼзֵַͼǼܲȡֱַͨӶַʶ̣ʶٶȡ
Ľܵͨƽ˲ģıͼȡַεٴķṩһ˻ַʶµĿܣ÷оͿдһƣз㷨ܻػӦԽһַзֵ׼ȷԡ
߼飺¬(1949)УڣҪоΪģʽʶͼ
ߵλ¬֡쿡ߵרѧУ졡215500
лࡡѧѧϵϺ200433
ο
1 Fujisawa H,Nakano Y,Kurino K.Segmentation Methods for Character Recognition:from Segmentation to Document Structure Analysis. Proc. of the IEEE, 1992, 80(7): 1079-1092
2 Kahan S,Pavlidis T,Baird H S.On the Recognition of Printed Characters of Any Font and Size. IEEE Trans.Pattern Anal. Machine Intell, 1987, 9(2): 274-288
3 Mitchell B T,Gillies A M. A Model-based Computer Vision System for Recognizing Handwritten ZIP Codes. Machine Vision and Applications, 1989:231-243
4 Tsujimoto S,Asada H.Major Components of a Complete Text Reading System.Proc. of the IEEE, 1992,80(7):1133-1149
ոڣ1998-12-21
