JPH0668304A - Handwritten character normalization system - Google Patents
Handwritten character normalization systemInfo
- Publication number
- JPH0668304A JPH0668304A JP4217714A JP21771492A JPH0668304A JP H0668304 A JPH0668304 A JP H0668304A JP 4217714 A JP4217714 A JP 4217714A JP 21771492 A JP21771492 A JP 21771492A JP H0668304 A JPH0668304 A JP H0668304A
- Authority
- JP
- Japan
- Prior art keywords
- character
- reference point
- point
- center
- feature
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Character Input (AREA)
- Character Discrimination (AREA)
Abstract
Description
【０００１】[0001]
【産業上の利用分野】本発明は手書き文字正規化方式に
係り、特に手書き文字を認識するときの前処理の１部で
ある入力文字の傾きの補正を行う正規化方式に関する。BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a handwritten character normalization method, and more particularly to a normalization method for correcting the inclination of an input character, which is a part of preprocessing when recognizing a handwritten character.
【０００２】[0002]
【従来の技術】近年コンピュータ等による情報処理装置
の普及に伴いこれらの装置への入力手段として、例えば
紙等に書いた手書き文字を読み取ることが要求されてい
る。このため、様々な手書きの文字パターンをできるだ
け正確に読み取るため、文字認識の前処理として、文字
の大きさや傾きを一様に揃える文字正規化処理が求めら
れている。2. Description of the Related Art In recent years, with the spread of information processing apparatuses such as computers, it has been required to read handwritten characters written on, for example, paper as an input means for these apparatuses. For this reason, in order to read various handwritten character patterns as accurately as possible, there is a demand for a character normalization process as a preprocess for character recognition so that the sizes and inclinations of the characters are made uniform.
【０００３】手書き文字を認識する場合、図９（Ａ）に
示す如く、文字を複数の領域に分割（この例では４分
割）して、それぞれの領域毎に特徴量を計算し、特徴量
によるパターンマッチングを行う手法がある。When recognizing a handwritten character, as shown in FIG. 9A, the character is divided into a plurality of areas (in this example, four divisions), and the characteristic amount is calculated for each area. There is a method of performing pattern matching.
【０００４】このような手法においては、入力文字と辞
書の文字を比較する時に、同一の文字に対して同一の部
分が分割領域内に含まれているようなときに正確な認識
が可能となる。しかし、図９（Ｂ）に示す如く、文字が
傾いているような場合は正確な認識ができなくなるの
で、何等かの手段により文字の傾きを補正することが必
要となる。In such a method, when the input character and the character in the dictionary are compared, it is possible to perform accurate recognition when the same portion is included in the divided area for the same character. . However, as shown in FIG. 9B, when the character is tilted, accurate recognition cannot be performed, so it is necessary to correct the tilt of the character by some means.
【０００５】このため、従来の文字の傾きを補正する正
規化手段には、文字の方向性を方向コードヒストグラ
ムにより調べ、最も頻度の大きい方向をその文字方向と
して判定し、これにより傾きを補正する、ＯＣＲシー
ト全体を対象として、文字の切り出し時点でスキュー量
を補正する等のものがあった。For this reason, in the conventional normalizing means for correcting the inclination of a character, the directionality of the character is examined by the direction code histogram, the direction with the highest frequency is determined as the character direction, and the inclination is corrected by this. , The OCR sheet as a whole was subject to correction of the skew amount at the time of cutting out characters.
【０００６】[0006]
【発明が解決しようとする課題】ところが前記は、そ
れ自体正確な検出が難しく、特に漢字のように画数の多
い文字に至っては、どの方向がその文字の方向であるか
ということを判断することが不可能に近い。また前記
は、この手法はＯＣＲシートが観測部に入るときの搬送
部での機械的な誤差により生ずる用紙の傾きを補正する
ものであって、記入された文字そのものの傾きを補正す
るものではない。However, the above is difficult to detect accurately by itself, and especially for a character with a large number of strokes such as a Chinese character, it is necessary to judge which direction is the direction of the character. Is almost impossible. Further, as described above, this method corrects the inclination of the sheet caused by a mechanical error in the transport unit when the OCR sheet enters the observation unit, and does not correct the inclination of the written character itself. .
【０００７】したがって本発明では、このような手書き
文字に関して、漢字のような画劃数の多い文字であって
も、文字の傾きを検出してこれを補正する手書き文字正
規化方式を提供することを目的とする。Therefore, the present invention provides a handwritten character normalization method for detecting the inclination of a handwritten character and correcting the handwritten character even if the character has a large number of strokes such as Chinese characters. With the goal.
【０００８】[0008]
【課題を解決するための手段】前記目的を達成するた
め、本発明では、図１（Ａ）に示す如く、基準点算出部
１により文字の基準点を先ず求める。基準点としては例
えば文字の重心を求める。これにより例えば図１（Ｂ）
に示すように基準点Ｇを得る。そしてこの基準点Ｇを通
るＸ軸、Ｙ軸を画く。そしてこのＸ軸より上の領域（図
１（Ｂ）の斜線部分）の部分の特徴点ＧU を上部特徴点
算出部２により求める。この特徴点として、文字の基準
点より上の部分の重心を求める。In order to achieve the above object, in the present invention, as shown in FIG. 1A, a reference point calculation unit 1 first obtains a reference point of a character. As the reference point, for example, the center of gravity of the character is obtained. As a result, for example, FIG.
A reference point G is obtained as shown in. Then, an X axis and a Y axis passing through the reference point G are drawn. Then, the characteristic point G U of the area above the X axis (hatched portion in FIG. 1B) is obtained by the upper characteristic point calculation unit 2. As this feature point, the center of gravity of the portion above the reference point of the character is obtained.
【０００９】同様にして下部特徴点算出部３によりこの
Ｘ軸より下の領域の部分の特徴点Ｇ D 、Ｙ軸より右の部
分の特徴点ＧR 、Ｙ軸より左の部分の特徴点ＧL を、そ
れぞれ下部特徴点算出部３、右部特徴点算出部４、左部
特徴点算出部５により求める。Similarly, the lower feature point calculation unit 3
Characteristic point G in the area below the X-axis D, The part to the right of the Y-axis
Minute feature point GR, G to the left of the Y axisLThe
Lower feature point calculator 3, right part feature point calculator 4, left part
It is obtained by the feature point calculation unit 5.
【００１０】傾き補正部６では、このようにして求めた
上部特徴点ＧU 、下部特徴点ＧD 、右部特徴点ＧR 、左
部特徴点ＧL と、Ｘ軸、Ｙ軸の交点である基準点Ｇをそ
れぞれ線で引き、各領域の傾きを求める。例えば図１
（Ｂ）に示す如く上部特徴点Ｇ U が算出されたとき、こ
のＧU とＧとの線とＹ軸との角度θU が上部領域の傾き
角度である。Inclination correction unit 6 obtains in this way
Upper feature point GU, Lower feature point GD, Right feature point GR,left
Department point GLAnd a reference point G, which is the intersection of the X-axis and the Y-axis,
Draw each line and find the slope of each area. Figure 1
Upper feature point G as shown in (B) UWhen is calculated,
GUThe angle θ between the line between G and G and the Y axisUIs the slope of the upper area
It is an angle.
【００１１】さらに傾き補正部６では、このようにして
それぞれ得られた上部領域の傾き角度θU 、下部領域の
傾き角度θD 、右部領域の傾き角度θR 、左部領域の傾
き角度θL の平均値を求めて文字全体の傾きを算出す
る。かくして得られた傾きに対して、例えば基準点Ｇを
中心に文字を回転させたり文字を分割すること等により
傾きの補正を行う。このようにして補正した後、辞書と
照合し、文字認識を行う。[0011] In further inclination correction unit 6, Thus the inclination angle of the upper region obtained respectively theta U, the inclination angle theta D of the lower region, the inclination angle theta R of the right region, the inclination angle of the left region theta The average value of L is calculated to calculate the inclination of the entire character. The inclination thus obtained is corrected by rotating the character around the reference point G or dividing the character. After correction in this way, character recognition is performed by collating with a dictionary.
【００１２】なお、図１（Ａ）においてＣは文字枠を示
している。In FIG. 1A, C indicates a character frame.
【００１３】[0013]
【作用】本発明によれば、個々の文字から傾きを検出
し、その傾きを補正することにより、より正確な手書き
文字の認識が可能になる。According to the present invention, it is possible to more accurately recognize a handwritten character by detecting the inclination of each character and correcting the inclination.
【００１４】[0014]
【実施例】本発明の一実施例を図２にもとづき説明す
る。図２において図１と同記号は同一部分を示し、７は
文字入力部、８は傾き算出部、９は文字回転部、１０は
文字認識部、１１は文字辞書である。DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT An embodiment of the present invention will be described with reference to FIG. 2, the same symbols as in FIG. 1 indicate the same parts, 7 is a character input unit, 8 is a tilt calculation unit, 9 is a character rotation unit, 10 is a character recognition unit, and 11 is a character dictionary.
【００１５】文字入力部７は文字を入力するものであ
り、例えば光学的な文字読取り部とかタブレット等によ
り構成されるものである。傾き算出部８は文字の傾きを
算出するものであり、上部特徴点算出部２、下部特徴点
算出部３、右部特徴点算出部４、左部特徴点算出部５よ
りそれぞれ算出したＧU 、ＧD 、ＧR 、ＲL と基準点Ｇ
とを結ぶ直線と、Ｙ軸又はＸ軸とのなす角度θU 、
θD 、θR 、θL を求める。The character input unit 7 is for inputting characters, and is constituted by, for example, an optical character reading unit or a tablet. The inclination calculating unit 8 calculates the inclination of the character, and G U calculated by the upper characteristic point calculating unit 2, the lower characteristic point calculating unit 3, the right characteristic point calculating unit 4, and the left characteristic point calculating unit 5, respectively. , G D , G R , R L and reference point G
An angle θ U formed by the straight line connecting with and the Y axis or X axis,
Find θ D , θ R , and θ L.
【００１６】例えばＧU とＧとを結ぶ直線とＹ軸とのな
す角θU を反時計回りなら正、時計回りなら負の値とし
て求める。ＧD とＧについても同様にしてθD を求め
る。またＧR とＧとを結ぶ直線とＸ軸とのなす角θR を
反時計回りなら正、時計回りなら負の値として求める。
ＧL とＧについても同様にしてθL を求める。そしてこ
れらθU 、θD 、θR 、θL の平均値θｍを入力文字全
体の傾きとして求める。For example, the angle θ U formed by the straight line connecting G U and G and the Y axis is determined as a positive value if it is counterclockwise and a negative value if it is clockwise. Θ D is similarly obtained for G D and G. Further, the angle θ R formed by the straight line connecting G R and G and the X-axis is obtained as a positive value if counterclockwise and a negative value if clockwise.
Similarly, for G L and G, θ L is calculated. Then, the average value θm of these θ U , θ D , θ R , and θ L is obtained as the inclination of the entire input character.
【００１７】文字回転部９は、入力文字を基準点Ｇを中
心として−θｍ回転させて、その傾きの補正を行うもの
である。文字認識部１０は、このようにして調整された
入力文字など、文字辞書１１とを照合し、文字認識を行
うものである。The character rotating unit 9 rotates the input character about -θm about the reference point G and corrects the inclination thereof. The character recognition unit 10 collates with the character dictionary 11 such as the input character adjusted in this way to perform character recognition.
【００１８】文字辞書１１は入力文字を認識するため
の、文字毎の特徴等が格納されているもである。次に図
２の動作について説明する。The character dictionary 11 stores characteristics of each character for recognizing input characters. Next, the operation of FIG. 2 will be described.
【００１９】手書き文字は、文字入力部１において例
えばスキャナの如き光学読み取り部により読み取られ、
文字部分が「１」（黒）、他の部分が「０」（白）の２
値のイメージデータとして入力される。The handwritten character is read in the character input section 1 by an optical reading section such as a scanner,
2 where the character part is "1" (black) and the other part is "0" (white)
Input as image data of value.
【００２０】そして基準点算出部１において文字の重
心を基準点Ｇとして算出する。重心は次の様にして求め
られる。Then, the reference point calculation unit 1 calculates the center of gravity of the character as the reference point G. The center of gravity is calculated as follows.
【００２１】[0021]
【数１】 [Equation 1]
【００２２】すなわちＮにより「１」の部分の画素数が
得られる。そしてＸにより画素が「１」の部分のＸ方向
の距離を加算した値が得られ、これをＮで割ることによ
り重心ＧのＸ座標（Ｘｇ）が得られる。同様にＹにより
画素が「１」の部分をＹ方向の距離を加算した値が得ら
れ、これをＮで割ることにより重心ＧのＹ座標（Ｙｇ）
が得られる。That is, the number of pixels in the portion "1" is obtained by N. Then, a value obtained by adding the distance in the X direction of the portion where the pixel is "1" is obtained by X, and by dividing this by N, the X coordinate (Xg) of the center of gravity G is obtained. Similarly, a value obtained by adding the distance in the Y direction to the portion where the pixel is "1" is obtained by Y, and this value is divided by N to obtain the Y coordinate (Yg) of the center of gravity G.
Is obtained.
【００２３】次に上部特徴算出部２において、文字の
重心より上の部分に対して、その重心を上部特徴点ＧU
として求める。算出は、算出範囲が文字重心（Ｘｇ、Ｙ
ｇ）より上の部分となるだけで、前記と同様にして得ら
れる。下部特徴点算出部３では、上部特徴点算出部２と
同様に、文字重心より下の部分に対してその重心を下部
特徴点ＧD として求める。右部特徴点算出部４、左部特
徴点算出部５も、同様にしてそれぞれ重心の右部分、左
部分の重心を求め、右部特徴点ＧR 、左部特徴点ＧL と
する。Next, in the upper feature calculation unit 2, the center of gravity of the portion above the center of gravity of the character is determined as the upper feature point G U.
Ask as. In the calculation, the calculation range is the character center of gravity (Xg, Y
It is obtained in the same manner as described above, except for the part above g). Similar to the upper feature point calculating unit 2, the lower feature point calculating unit 3 obtains the center of gravity of the portion below the character center of gravity as the lower feature point G D. Similarly, the right-side feature point calculation unit 4 and the left-side feature point calculation unit 5 similarly obtain the right and left center-of-gravity points, respectively, and set them as the right-side feature point G R and the left-side feature point G L.
【００２４】傾き補正部６では、基準点Ｇと、前記Ｇ
U 、ＧD 、ＧR 、ＧL より文字の正規化を行う。まず傾
き算出部８で前記θU 、θD 、θR 、θL を求め、その
平均値θｍを入力文字の傾き角として求める。そしてこ
の傾き角にもとづき、文字回転部９では文字を基準点Ｇ
を中心にして−θｍだけ回転させ、傾きを補正する。In the inclination correction unit 6, the reference point G and the G
Character normalization is performed from U , G D , G R , and G L. First, the inclination calculator 8 obtains θ U , θ D , θ R , and θ L, and obtains the average value θm as the inclination angle of the input character. Then, based on this tilt angle, the character rotation unit 9 determines that the character is the reference point G.
Rotate by -θm around to correct the inclination.
【００２５】このようにして傾きの補正された文字デ
ータが文字認識部１０において文字辞書１１を参照して
文字認識を行う。なお、前記実施例では基準点、特徴点
に手書き文字の文字部分の重心を使用したが、図３
（Ａ）、（Ｂ）に示す如く、文字を囲む矩形または凸多
角形を考え、文字枠内部を「１」、文字枠外を「０」と
するようなＰXYを用いて、文字枠全体の重心を基準点と
し、基準点より上側の文字枠内の重心を上部特徴点とす
ることにより傾きを求めてもよい。また重心の代わりに
文字や文字枠の中心点を用いてもよい。The character data whose inclination has been corrected in this way is recognized in the character recognition unit 10 by referring to the character dictionary 11. Although the center of gravity of the character portion of the handwritten character is used as the reference point and the feature point in the above embodiment,
As shown in (A) and (B), consider a rectangle or a convex polygon that surrounds a character, and use P XY such that the inside of the character frame is “1” and the outside of the character frame is “0”, and The inclination may be obtained by using the center of gravity as the reference point and the center of gravity in the character frame above the reference point as the upper feature point. Further, the center point of a character or a character frame may be used instead of the center of gravity.
【００２６】本発明の第２実施例を図４、図５により説
明する。図４は本発明の第２実施例構成図であり、図５
は分割状態説明図である。図２により説明した第一実施
例では、文字の正規化として文字の傾きを算出し、回転
による補正を行ったが、第２実施例では認識処理におけ
る文字の分割処理の正規化を行うものである。A second embodiment of the present invention will be described with reference to FIGS. FIG. 4 is a configuration diagram of a second embodiment of the present invention, and FIG.
[Fig. 6] is an explanatory diagram of a divided state. In the first embodiment described with reference to FIG. 2, the inclination of the character is calculated and the correction by rotation is performed as the normalization of the character, but in the second embodiment, the division processing of the character in the recognition processing is normalized. is there.
【００２７】図４において、基準点算出部１、上部特徴
点算出部２、下部特徴点算出部３、右部特徴点算出部
４、左部特徴点算出部５、文字入力部７、文字認識部１
０、文字辞書１１等は図２のそれと同様のものである。In FIG. 4, a reference point calculation unit 1, an upper characteristic point calculation unit 2, a lower characteristic point calculation unit 3, a right characteristic point calculation unit 4, a left characteristic point calculation unit 5, a character input unit 7, character recognition. Part 1
0, the character dictionary 11 and the like are the same as those in FIG.
【００２８】傾き補正部６は分割線生成部２１と領域分
割部２２を有する。分割線生成部２１では、図５（Ａ）
に示す如く、基準点Ｇから各特徴点ＧU 、ＧD 、ＧR 、
ＧLを通る４本の分割線Ｄ1 、Ｄ2 、Ｄ3 、Ｄ4 を求め
る。領域分割部２２では、生成されたこの４本の直線Ｄ
1 、Ｄ2 、Ｄ3 、Ｄ4 によって文字を複数の領域Ａ1、
Ａ2 、Ａ3 、Ａ4 に分割する。The inclination correction unit 6 has a dividing line generating unit 21 and a region dividing unit 22. In the dividing line generation unit 21, FIG.
As shown in, each of the feature points G U , G D , G R from the reference point G,
Four dividing lines D 1 , D 2 , D 3 and D 4 passing through G L are obtained. In the area division unit 22, the four generated straight lines D
1 , D 2 , D 3 and D 4 are used to define a character into a plurality of areas A 1 ,
Divide into A 2 , A 3 , and A 4 .
【００２９】文字特徴抽出部２３では、分割された領域
毎に文字の特徴として、その輪郭線の方向コードを抽出
する。ここで方向コードとは、領域内の水平な画素列、
垂直な画素列、斜めの画素（例えば４５度、１３５度、
２２５度、３１５度等）列を数えて、その数を特徴量と
するものである。The character feature extraction unit 23 extracts the direction code of the contour line of each divided area as a character feature. Here, the direction code means a horizontal pixel row in the area,
Vertical pixel rows, diagonal pixels (eg 45 degrees, 135 degrees,
(225 degrees, 315 degrees, etc.) columns are counted and the number is used as a feature amount.
【００３０】文字認識部１０では、これにより得られた
特徴量と、文字辞書１１とを対応させて文字の認識を行
う。なお、前記説明では文字を４つの領域に分割した例
について説明したが、分割線をさらに分割することによ
り、図５（Ｂ）に示す如く、多数の領域に分割でき、き
め細かな照合を行うことができる。The character recognition unit 10 recognizes a character by associating the feature amount thus obtained with the character dictionary 11. In the above description, an example in which a character is divided into four areas has been described, but by dividing the dividing line further, it is possible to divide into a large number of areas as shown in FIG. 5B, and fine matching can be performed. You can
【００３１】本発明の第３実施例を図６、図７にもとづ
き説明する。図６は本発明の第３実施例構成図、図７は
その動作状態説明図である。前記説明では、４つの特徴
点ＧU 、ＧD 、ＧR 、ＧL はそれぞれ独立に文字の４つ
の部分から求めていたが、この実施例では、上部特徴点
ＧU と下部特徴点Ｇ D の２つの特徴点を求めた後に、こ
れらを用いて右部特徴点ＧR と左部特徴点Ｇ L を求める
ものである。A third embodiment of the present invention will be described with reference to FIGS. 6 and 7.
Explain. FIG. 6 is a block diagram of the third embodiment of the present invention, and FIG.
It is an operation state explanatory view. In the above description, four features
Point GU, GD, GR, GLAre four letters each independently
In this example, the upper feature point
GUAnd lower feature point G DAfter finding the two feature points of
Using these, the right feature point GRAnd the left feature point G LAsk for
It is a thing.
【００３２】図６において、右部特徴部４′及び左部特
徴部５′以外は前記図２等と同様である。基準点算出部
１により基準点Ｇが得られ、上部特徴点算出部２により
上部特徴点ＧU が得られ、下部特徴点算出部３により下
部特徴点ＧD が得られたとき、右部特徴点算出部４′で
は、図７に示す如く、ＧU 、ＧD を通る直線ＬV に関
し、これに垂直でかつＧを通る直線ＬH を引き、ＬH 上
でＧより右側にある任意の点を右部特徴点とする。また
左部特徴点算出部５′ではこの直線ＬH 上でＧより左に
ある任意の点を左部特徴点とする。6 is the same as that shown in FIG. 2 except for the right characteristic portion 4'and the left characteristic portion 5 '. When the reference point calculation unit 1 obtains the reference point G, the upper feature point calculation unit 2 obtains the upper feature point G U , and the lower feature point calculation unit 3 obtains the lower feature point G D , the right part feature at point calculator 4 ', as shown in FIG. 7, G U, relates straight line L V through G D, a straight line is drawn L H that passes through the vertical and and G to, any of the right from G on L H Let the point be the right-side feature point. In the left-side feature point calculation unit 5 ', any point on the straight line L H to the left of G is set as the left-side feature point.
【００３３】このようにして傾きを求めてこれを補正し
て文字認識を正確に行うことができる。なお、図６、図
７では、上部特徴点ＧU 、下部特徴点ＧD から右部特徴
点ＧR、左部特徴点ＧL を求めたが、逆に右部特徴点Ｇ
R と左部特徴点ＧL を先に求めてそれを用いて上部特徴
点ＧU 、下部特徴点ＧD を求めてもよい。In this way, the inclination can be obtained and corrected to accurately perform character recognition. 6 and 7, the right characteristic point G R and the left characteristic point G L are obtained from the upper characteristic point G U and the lower characteristic point G D , but conversely, the right characteristic point G
It is also possible to obtain R and the left-side feature point G L first and use them to obtain the upper feature point G U and the lower feature point G D.
【００３４】また文字によって、幅より高さの方が長い
文字（例えば１の字）ならばＧU 、ＧD からＧR 、ＧL
を求め、高さより幅の方が長い文字（例えば−の字）な
らば逆にするように適応させてもよい。Depending on the character, if it is a character whose height is longer than its width (for example, the character 1), G U , G D to G R , G L
May be obtained, and if the character has a width longer than the height (for example, a minus character), it may be reversed.
【００３５】図８はこの例を示す。縦横長判別部２５に
より文字の縦横長のいずれが大きいか判別し、切換制御
部２６、第１算出部２７、第２算出部２８を制御する。
切換制御部２６は基準点算出部１からのデータを第１算
出部２７に送出するのか、第２算出部２８に送出するの
かを選択するものであり、例えば幅より高さの方が長い
文字については、第１算出部２７に送出する。FIG. 8 shows this example. The vertical / horizontal determination unit 25 determines which of the vertical and horizontal characters is larger, and controls the switching control unit 26, the first calculation unit 27, and the second calculation unit 28.
The switching control unit 26 selects whether to send the data from the reference point calculation unit 1 to the first calculation unit 27 or the second calculation unit 28. For example, a character whose height is longer than its width is selected. Is sent to the first calculator 27.
【００３６】第１算出部２７は、上部特徴点算出部２−
１と下部特徴点３−１を有する。縦横長判別部２５の制
御信号により幅より高さの方が長い文字の場合には、こ
れらの算出部は切換制御部２６から伝達されたＧを含む
文字データにより前記の如くＧU とＧD を算出しこれら
をＧとともに第２算出部２８に送出する。逆に幅の方が
長い文字の場合には、第２算出部２８から伝達されたＧ
とＧR 、ＧL により先ずＬH を作成し、次にＧを通るＬ
V を作成してＧU 、ＧD を決定する。The first calculator 27 is an upper feature point calculator 2-.
1 and a lower feature point 3-1. In the case of a character whose height is longer than width according to the control signal of the vertical / horizontal determination unit 25, these calculation units use the character data including G transmitted from the switching control unit 26 as described above for G U and G D. And sends them to the second calculator 28 together with G. On the other hand, when the character has a longer width, the G transmitted from the second calculator 28 is transmitted.
First, L H is created from G R and G L , and then L passing G
Create V to determine G U and G D.
【００３７】第２算出部２８は、右部特徴点算出部４−
１と左部特徴点５−１を有する。縦横長判別部２５の制
御信号により幅の方が高さより長い文字の場合には、こ
れらの算出部は、切換制御部２６から伝達されたＧを含
む文字データによりＧR とＧ L を算出してこれらをＧと
ともに第１算出部２７に送出する。逆に高さの方が長い
文字については、第１算出部２７から伝達されたＧとＧ
U 、ＧD によりまずＬ V を作成し、次にＧを通るＬH を
作成し、ＧR 、ＧL を決定する。The second calculating section 28 includes a right-side feature point calculating section 4-
1 and the left part feature point 5-1. Control of portrait / landscape determination unit 25
If the width of the character is longer than the height due to the signal,
These calculation units include G transmitted from the switching control unit 26.
G by character dataRAnd G LAnd calculate these as G
Both are sent to the first calculator 27. On the contrary, the height is longer
Regarding the characters, G and G transmitted from the first calculation unit 27
U, GDFirst L V, Then L through GHTo
Create and GR, GLTo decide.
【００３８】このようにして傾きを算出して補正を行い
文字認識を正確にすることができる。In this way, the inclination can be calculated and corrected to make the character recognition accurate.
【００３９】[0039]
【発明の効果】このように、本発明によれば、傾いた手
書き文字に対しても正規化により傾きを補正することが
可能になる。このため本発明を手書き文字認識の前処理
として使用することにより、一層正確な文字の認識が可
能になる。As described above, according to the present invention, the inclination can be corrected by normalization even with respect to the inclined handwritten character. Therefore, by using the present invention as a pre-process for handwritten character recognition, more accurate character recognition is possible.
【図１】本発明の原理説明図を示す。FIG. 1 is a diagram illustrating the principle of the present invention.
【図２】本発明の一実施例構成図を示す。FIG. 2 shows a configuration diagram of an embodiment of the present invention.
【図３】文字枠説明図を示す。FIG. 3 shows a character box explanatory diagram.
【図４】本発明の第２実施例構成図を示す。FIG. 4 shows a configuration diagram of a second embodiment of the present invention.
【図５】分割状態説明図を示す。FIG. 5 shows an explanatory diagram of a divided state.
【図６】本発明の第３実施例構成図を示す。FIG. 6 shows a configuration diagram of a third embodiment of the present invention.
【図７】動作状態説明図を示す。FIG. 7 shows an explanatory diagram of an operating state.
【図８】本発明の第４実施例構成図を示す。FIG. 8 shows a configuration diagram of a fourth embodiment of the present invention.
【図９】文字分割状態説明図を示す。FIG. 9 is an explanatory diagram of a character division state.
１ 基準点算出部 ２ 上部特徴点算出部 ３ 下部特徴点算出部 ４ 右部特徴点算出部 ５ 左部特徴点算出部 ６ 傾き補正部 1 Reference Point Calculation Section 2 Upper Feature Point Calculation Section 3 Lower Feature Point Calculation Section 4 Right Side Feature Point Calculation Section 5 Left Side Feature Point Calculation Section 6 Tilt Correction Section
───────────────────────────────────────────────────── フロントページの続き (72)発明者 伊崎 保直 神奈川県川崎市中原区上小田中1015番地 富士通株式会社内 (72)発明者 諏訪 美佐子 神奈川県川崎市中原区上小田中1015番地 富士通株式会社内 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Hoshinao Izaki 1015 Kamiodanaka, Nakahara-ku, Kawasaki City, Kanagawa Prefecture, Fujitsu Limited (72) Inventor, Misako Suwa 1015, Kamedotachu, Nakahara-ku, Kawasaki City, Kanagawa Prefecture, Fujitsu Limited
Claims (11)
Ｇを算出する基準点算出部（１）と、 この基準点Ｇの上部の中心となる上部特徴点を算出する
上部特徴点算出部（２）と、 基準点Ｇの下部の中心となる下部特徴点を算出する下部
特徴点算出部（３）と、 基準点Ｇの右部の中心となる右部特徴点を算出する右部
特徴点算出部（４）と、 基準点Ｇの左部の中心となる左部特徴点を算出する左部
特徴点算出部（５）と、 基準点Ｇと特徴点を結ぶ直線から文字の傾きを求め、こ
れにより傾きを補正する傾き補正部（６）を具備し、 手書き文字の正規化を行うことを特徴とする手書き文字
正規化方式。1. A reference point calculation unit (1) for calculating a reference point G that is the center of normalization of handwritten characters, and an upper feature point calculation unit for calculating an upper feature point that is the center of the upper part of the reference point G. (2), a lower feature point calculation unit (3) that calculates a lower feature point that is the center of the lower part of the reference point G, and a right part feature that calculates a right feature point that is the center of the right part of the reference point G A point calculation unit (4), a left-side feature point calculation unit (5) that calculates a left-side feature point that is the center of the left side of the reference point G, and a character inclination from a straight line connecting the reference point G and the feature point. A handwritten character normalization method comprising a tilt correction unit (6) for obtaining and correcting the tilt according to the obtained value, and normalizing the handwritten character.
ようにしたことを特徴とする請求項１記載の手書き文字
正規化方式。2. The handwritten character normalization method according to claim 1, wherein the center of gravity of the character portion is calculated as a reference point.
分、左の部分の文字部分の重心を特徴点として算出する
ようにしたことを特徴とする請求項１記載の手書き文字
正規化方式。3. The handwritten character normal according to claim 1, wherein the center of gravity of the character portion of the portion above, below, right, and left of the reference point is calculated as the feature point. Method.
による文字枠の重心又は中心を基準点として算出するこ
とを特徴とする請求項１記載の手書き文字正規化方式。4. The handwritten character normalization method according to claim 1, wherein the center of gravity or the center of a character frame formed by a minimum rectangle or a convex polygon including a character is used as a reference point.
による文字枠の基準点より上の部分、下の部分、右の部
分、左の部分の重心を特徴点として算出するようにした
ことを特徴とする請求項１記載の手書き文字正規化方
式。5. The center of gravity of a portion above, below, right, and left of a reference point of a character frame of a minimum rectangle or a convex polygon including a character is calculated as a feature point. The handwritten character normalization method according to claim 1, wherein the handwritten character normalization method is used.
字の傾きを求め、その角度の分だけ文字を回転させるこ
とにより文字の傾きの補正を行うことを特徴とする請求
項１記載の手書き文字正規化方式。6. The handwriting according to claim 1, wherein the inclination of the character is calculated based on the angle formed by the reference point and the feature point, and the character is rotated by the angle. Character normalization method.
割して、正規化を行うことを特徴とする請求項１記載の
手書き文字正規化方式。7. The handwritten character normalization method according to claim 1, wherein the character is divided by a straight line connecting the reference point and each feature point to perform normalization.
た点より分割線に平行な直線を引き、それらの直線で文
字を分割して、正規化を行うことを特徴とする請求項１
記載の手書き文字正規化方式。8. A normalization is performed by drawing straight lines parallel to a dividing line from a point obtained by equally dividing the dividing line connecting the reference point and each feature point, dividing the character by these straight lines. Item 1
The handwritten character normalization method described.
分の２つの部分の重心を上下の特徴点とし、上下の特徴
点を結ぶ直線に垂直で基準点を通る直線上で、基準点の
左右の任意の位置に左右の２つの特徴点を求めることを
特徴とする請求項１記載の手書き文字正規化方式。9. A barycenter of two parts, a part above and below a reference point of a character part, is defined as upper and lower feature points, and a reference line is drawn on a straight line which is perpendicular to a straight line connecting the upper and lower feature points and which passes through the reference point. The handwritten character normalization method according to claim 1, wherein two right and left feature points are obtained at arbitrary positions on the left and right sides of the point.
部分の２つの部分の重心を左右の特徴点とし、左右の特
徴点を結ぶ直線に垂直で基準点を通る直線上で、基準点
の上下の任意の位置に上下の２つの特徴点を求めること
を特徴とする請求項１記載の手書き文字正規化方式。10. The center of gravity of two parts, the right part and the left part of the character part, is defined as the left and right feature points, and the reference point is on a straight line that is perpendicular to the straight line connecting the left and right feature points and passes through the reference point. 2. The handwritten character normalization method according to claim 1, wherein two upper and lower characteristic points are obtained at arbitrary positions above and below the point.
場合は、文字部分の基準点より上の部分と下の部分の２
つの部分の重心を上下の特徴点とし、上下の特徴点を結
ぶ直線に垂直で基準点を通る直線上で基準点の左右の任
意の位置に左右の２つの特徴点を求め、また横の方が大
の場合は、文字部分の基準点より右の部分と左の部分の
２つの部分の重心を左右の特徴点とし、左右の特徴点を
結ぶ直線に垂直で基準点を通る直線上で基準点の上下の
任意の位置に上下の２つの特徴点を求めることを特徴と
する請求項１記載の手書き文字正規化方式。11. An aspect ratio of a character is calculated, and when the height is larger, 2 of the part above and below the reference point of the character part is calculated.
The center of gravity of two parts is used as the upper and lower feature points, and two feature points on the left and right are obtained at arbitrary positions on the left and right of the reference point on a straight line that passes through the reference point and is perpendicular to the line connecting the upper and lower feature points. When is large, the center of gravity of the two parts, the right part and the left part of the reference point of the character part, is the left and right feature points, and the reference is made on a straight line that passes through the reference point perpendicular to the line connecting the left and right feature points. 2. The handwritten character normalization method according to claim 1, wherein two upper and lower characteristic points are obtained at arbitrary positions above and below the point.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP4217714A JPH0668304A (en) | 1992-08-17 | 1992-08-17 | Handwritten character normalization system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP4217714A JPH0668304A (en) | 1992-08-17 | 1992-08-17 | Handwritten character normalization system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| JPH0668304A true JPH0668304A (en) | 1994-03-11 |
Family
ID=16708590
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| JP4217714A Pending JPH0668304A (en) | 1992-08-17 | 1992-08-17 | Handwritten character normalization system |
Country Status (1)
| Country | Link |
|---|---|
| JP (1) | JPH0668304A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6115506A (en) * | 1997-05-07 | 2000-09-05 | Nec Corporation | Character recognition method, character recognition apparatus and recording medium on which a character recognition program is recorded |
| KR101725501B1 (en) * | 2016-07-13 | 2017-04-13 | 한동대학교 산학협력단 | Method and apparatus for recognizing character |
-
1992
- 1992-08-17 JP JP4217714A patent/JPH0668304A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6115506A (en) * | 1997-05-07 | 2000-09-05 | Nec Corporation | Character recognition method, character recognition apparatus and recording medium on which a character recognition program is recorded |
| AU753499B2 (en) * | 1997-05-07 | 2002-10-17 | Nec Corporation | Character recognition method, character recognition apparatus and recording medium on which a character recognition program is recorded |
| KR101725501B1 (en) * | 2016-07-13 | 2017-04-13 | 한동대학교 산학협력단 | Method and apparatus for recognizing character |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US7437001B2 (en) | Method and device for recognition of a handwritten pattern | |
| RU2621601C1 (en) | Document image curvature eliminating | |
| EP0750415A2 (en) | Image processing method and apparatus | |
| RU2251738C2 (en) | Method for synchronizing filled machine-readable form and its template in case of presence of deviations | |
| JP3471578B2 (en) | Line direction determining device, image tilt detecting device, and image tilt correcting device | |
| WO1991020055A1 (en) | Self-centering character stroke thickening for optical character recognition | |
| JPH0668304A (en) | Handwritten character normalization system | |
| CN113537216B (en) | Dot matrix font text line inclination correction method and device | |
| JP3303246B2 (en) | Image processing device | |
| Hadi et al. | A novel approach of skew estimation and correction in persian manuscript text using radon transform | |
| JP2635582B2 (en) | Pattern recognition method | |
| JP3196291B2 (en) | Character recognition method and device | |
| JPH10207978A (en) | Character pattern collating method and device therefor | |
| JPH09179982A (en) | Specific pattern detecting method | |
| JP2755299B2 (en) | Image processing method | |
| JPH09179938A (en) | On-line handwritten character recognition device | |
| JPH0573718A (en) | Area attribute identifying system | |
| JP2792656B2 (en) | Restoration method of interruption of thinned data in handwritten digit recognition | |
| JP2616994B2 (en) | Feature extraction device | |
| JP2732752B2 (en) | Fuzzy pattern recognition device | |
| JPH0656625B2 (en) | Feature extraction method | |
| JP3368807B2 (en) | Character recognition method and apparatus, and storage medium storing character recognition program | |
| JP2024139561A (en) | Character recognition device, method, and program | |
| JP2918363B2 (en) | Character classification method and character recognition device | |
| JPS6379193A (en) | Character reader |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| A02 | Decision of refusal |
Free format text: JAPANESE INTERMEDIATE CODE: A02 Effective date: 20010925 |