JPH0155510B2 - - Google Patents

Info

Publication number
JPH0155510B2
JPH0155510B2 JP55189131A JP18913180A JPH0155510B2 JP H0155510 B2 JPH0155510 B2 JP H0155510B2 JP 55189131 A JP55189131 A JP 55189131A JP 18913180 A JP18913180 A JP 18913180A JP H0155510 B2 JPH0155510 B2 JP H0155510B2
Authority
JP
Japan
Prior art keywords
character
scanning
background
length
projected
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.)
Expired
Application number
JP55189131A
Other languages
Japanese (ja)
Other versions
JPS57113185A (en
Inventor
Tozen Hai
Koya Fujita
Yoshihisa Fujii
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Computer Basic Technology Research Association Corp
Original Assignee
Computer Basic Technology Research Association Corp
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Computer Basic Technology Research Association Corp filed Critical Computer Basic Technology Research Association Corp
Priority to JP55189131A priority Critical patent/JPS57113185A/en
Publication of JPS57113185A publication Critical patent/JPS57113185A/en
Publication of JPH0155510B2 publication Critical patent/JPH0155510B2/ja
Granted legal-status Critical Current

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/10Character recognition
    • G06V30/18Extraction of features or characteristics of the image
    • G06V30/18086Extraction of features or characteristics of the image by performing operations within image blocks or by using histograms
    • G06V30/18095Summing image-intensity values; Projection and histogram analysis
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/10Character recognition

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Theoretical Computer Science (AREA)
  • Character Discrimination (AREA)

Description

【発明の詳細な説明】 本発明は、文字の背景部分に着目して文字の特
徴を抽出する文字認識方式に関する。
DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a character recognition method that extracts characteristics of characters by focusing on the background portion of the characters.

第1図a,bは、文字の背景に着目した従来の
文字認識方式を示すもので、背景のある点の上、
下、左、右の4方向を見たとき文字線がある方向
の数を数えて、0,1,2,3,4として符号化
して表示して背景の特徴としていた。すなわち、
第1図aにおいて、斜線で示された枠体1を文字
とすると、例えば左上隅部は、4方向に見たとき
文字線に当たらないので0と符号化され、この0
と同一行の右隣の位置は、下方に見たとき1回だ
け文字線と当たるから1、同様にして各隅部は
0、その間は1、かつ枠体1で囲まれた部分は4
と符号化される。ところが、この方式では、枠体
1の右下部に突き出し2が存在する場合、突き出
し2の直上部の点では、左方と下方の2方向に見
たとき文字線に当たるから2と符号化される。同
様な方式で、突き出し2を含む列は、図示のよう
に上から1,2,2,2,1と符号化されること
が解かる。これに対して、突き出しの存在しない
第1図aの場合は、前述の列に対応する列は0,
1,1,1,1,0と符号化される。第1図a,
bにおいて、これらの対応した列の数字には丸印
をつけて対比の便を図つた。第1図aとbの対比
より明らかなように、突き出し2の存在の有無と
いうわずかな相違しか、第1図aとbの文字パタ
ーンには存在しないにもかかわらず、符号化パタ
ーンでは大きな相違を生じてしまうので、例えば
同じ「ロ」の字でも筆記者のくせで上記の突き出
し2をもつた「ロ」を書いてしまうと文字「ロ」
とは全く別の文字と誤認識される惧れがある。こ
のような欠点を解決するための従来の方式として
第2図a,bに示すように文字パターンを縦方向
に走査し、この方向に沿つた、対向する辺の距離
の分布を抽出する方式がある。第2図aには数字
「3」を、この方式によつて走査した投影図を示
し、第2図bには、数字「5」を、この方式によ
つて走査した投影図を示す。この方式によれば、
前記の文字「ロ」の突き出しがあつても誤認識す
ることはないが第2図aとbの投影図では、数字
3と5が酷似してしまうので、両者の識別が難し
い。
Figures 1a and b show a conventional character recognition method that focuses on the background of characters.
When looking in four directions (bottom, left, and right), the number of character lines in each direction was counted, coded as 0, 1, 2, 3, and 4, and displayed as a feature of the background. That is,
In Figure 1a, if the frame 1 indicated by diagonal lines is a character, for example, the upper left corner does not hit the character line when viewed in four directions, so it is coded as 0, and this 0
The position next to the right on the same line as ``1'' is 1 because it hits the character line only once when viewed from below, and similarly, each corner is 0, the space between them is 1, and the area surrounded by frame 1 is 4.
is encoded as However, in this method, if protrusion 2 exists at the lower right corner of frame 1, the point directly above protrusion 2 is encoded as 2 because it hits the character line when viewed in two directions: left and downward. . In a similar manner, it can be seen that the column containing extrusion 2 is encoded as 1, 2, 2, 2, 1 from the top as shown. On the other hand, in the case of Fig. 1a where there is no protrusion, the columns corresponding to the aforementioned columns are 0,
It is encoded as 1, 1, 1, 1, 0. Figure 1a,
In b, the numbers in these corresponding columns are circled for convenience of comparison. As is clear from the comparison between Figure 1 a and b, although there is only a slight difference in the presence or absence of protrusion 2 in the character patterns in Figure 1 a and b, there is a large difference in the encoding pattern. Therefore, for example, even if the character "ro" is the same, if you write "ro" with the above protrusion 2 due to the habit of a scribe, it will become the character "ro".
There is a risk that it may be mistakenly recognized as a completely different character. A conventional method to solve this problem is to scan the character pattern vertically and extract the distribution of distances between opposing sides along this direction, as shown in Figure 2 a and b. be. FIG. 2a shows a projection of the number "3" scanned by this method, and FIG. 2b shows a projection of the number "5" scanned by this method. According to this method,
Even if the above-mentioned character "RO" is protruded, there will be no erroneous recognition, but in the projection views of FIGS. 2a and 2b, the numbers 3 and 5 look very similar, making it difficult to distinguish between them.

本発明は上述の従来の欠点に鑑みて、筆記者特
有の個人差によるストロークのハネ突き出し等に
よる変動に対して別の文字と誤認識することのな
い安定した特徴を抽出でき、また、垂直、水平の
2方向のみの走査では区別しにくい字種も正確に
識別可能な文字認識方式を提供することを目的と
する。
In view of the above-mentioned conventional drawbacks, the present invention is capable of extracting stable features that will not be misrecognized as different characters despite fluctuations due to individual differences in strokes such as protrusion of strokes due to individual differences among scribes, It is an object of the present invention to provide a character recognition method that can accurately identify character types that are difficult to distinguish by scanning in only two horizontal directions.

本発明の文字認識方式の特徴とするところは文
字パターンを斜め方向に走査し該走査方向に沿つ
た、該文字パターンの第1の辺から第2の辺まで
の距離の分布を抽出して文字の認識を行うことで
ある。
The character recognition method of the present invention is characterized by scanning a character pattern in a diagonal direction and extracting the distribution of distances from the first side to the second side of the character pattern along the scanning direction. It is to recognize that

以下、図面を参照して本発明の一実施例を説明
する。
Hereinafter, one embodiment of the present invention will be described with reference to the drawings.

第3図aに示した、入力文字「文」に対して、
水平方向に走査し(b図)、最初の文字線の終り
b1を始点とし最後の文字線の始まりb1′を終点と
する背景の長さ、即ち走査方向に沿つた二辺間の
距離を抽出し、走査方向に垂直な軸10に投影1
1する。同様にして、垂直方向の走査による該背
景の長さを軸12に投影13する(c図)。さら
に、第3図dに示すように、入力文字「文」に対
して右上りたとえば45゜の角度を有して走査し、
最初の文字線の終りd1を始点とし、最後の文字線
の始まりd2を終点とする背景の長さを抽出し、走
査方向に垂直な軸14に投影15する。同様にし
て、入力文字「文」に対して右下りたとえば45゜
に走査して該背景の長さを抽出し、走査方向に垂
直な軸16に投影17する。
For the input character “sentence” shown in Figure 3a,
Scan horizontally (Figure b) and find the end of the first character line.
Extract the length of the background starting from b 1 and ending at the start of the last character line b 1 ', that is, the distance between the two sides along the scanning direction, and project it on the axis 10 perpendicular to the scanning direction.
Do 1. Similarly, the length of the background due to the vertical scan is projected 13 onto the axis 12 (figure c). Furthermore, as shown in FIG. 3d, the input character "sentence" is scanned upward to the right at an angle of, for example, 45 degrees,
The length of the background starting from the end d1 of the first character line and ending at the start d2 of the last character line is extracted and projected 15 onto the axis 14 perpendicular to the scanning direction. Similarly, the length of the background is extracted by scanning the input character "sentence" downward to the right, for example, at an angle of 45 degrees, and is projected 17 onto an axis 16 perpendicular to the scanning direction.

上述のようにして得られた「文」の4種類の投
影像11,13,15,17を用いてあらかじめ
辞書に貯えられた投影像と比較して、「文」の文
字を認識する。この認識回路20を第4図を用い
て説明する。
The four types of projection images 11, 13, 15, and 17 of "Bun" obtained as described above are compared with projection images stored in advance in a dictionary to recognize the characters of "Bun". This recognition circuit 20 will be explained using FIG. 4.

入力文字「文」21が加えられる観測部22の
出力はメモリ23に接続される。メモリ23には
4つの走査方向に対応する水平走査アドレスカウ
ンタ24、垂直走査アドレスカウンタ25、斜め
右より走査アドレスカウンタ26、斜め右下り走
査アドレスカウンタ27の出力が接続される。メ
モリ23の出力は水平方向検出部28、垂直方向
検出部29、斜め右上り方向検出部30、斜め右
下り方向検出部31をそれぞれ介して水平方向特
徴メモリ32、垂直方向特徴メモリ33、斜め右
上り方向特徴メモリ34、斜め右下り方向特徴メ
モリ35に入力される。各メモリ32,33,3
4,35の出力はマツチング回路36に接続され
る。このマツチング回路36には辞書37の出力
も加えられる。マツチング回路36の出力は認識
部38に加えられる。
The output of the observation section 22 to which the input character "sentence" 21 is added is connected to the memory 23. The outputs of a horizontal scanning address counter 24, a vertical scanning address counter 25, a diagonally rightward scanning address counter 26, and a diagonally downward rightward scanning address counter 27 corresponding to four scanning directions are connected to the memory 23. The output of the memory 23 is transmitted through a horizontal direction detection section 28, a vertical direction detection section 29, a diagonal upward right direction detection section 30, and a diagonal right downward direction detection section 31, respectively. The data is input to an upward direction feature memory 34 and a diagonally right downward direction feature memory 35. Each memory 32, 33, 3
The outputs of 4 and 35 are connected to a matching circuit 36. The output of a dictionary 37 is also added to this matching circuit 36. The output of the matching circuit 36 is applied to a recognition section 38.

上述のように構成された認識回路20におい
て、所定領域内に書かれた入力文字「文」21は
観測部22における一定方向の光学的な走査によ
つて前記領域内の濃淡を“1”,“0”に変換して
メモリ23に「文」のパターンに対応したデータ
として置数される。このデータは水平走査アドレ
スカウンタ24によつて水平方向に走査、即ちメ
モリ23において、前記所定領域内を水平方向に
走査するのに対応するアドレスを順次アクセスし
てゆき、水平方向における最初の文字線の終りと
最後の文字線の始めの間の背景の長さは水平方向
検出部28で検出され、水平方向特徴メモリ32
に前記背景の長さの投影像が格納される。同様に
して、垂直走査アドレスカウンタ25と垂直方向
検出部29の動作により垂直方向の前記背景の長
さの投影像が垂直方向特徴メモリ33に格納され
る。同様にして、斜め右上り走査アドレスカウン
タ26と斜め右上り方向検出部30の動作により
斜め右上り方向の前記背景の長さの投影像が斜め
右上り方向特徴メモリ34に格納される。さら
に、同様にして、斜め右下り走査アドレスカウン
タ27と斜め右下り方向検出部31の動作により
斜め右下り方向の前記背景の長さの投影像が斜め
右下り方向特徴メモリ35に格納される。
In the recognition circuit 20 configured as described above, the input character "sentence" 21 written in a predetermined area is optically scanned in a certain direction by the observation unit 22, so that the shading in the area is set to "1", It is converted to "0" and stored in the memory 23 as data corresponding to the "sentence" pattern. This data is scanned in the horizontal direction by the horizontal scanning address counter 24, that is, in the memory 23, the addresses corresponding to the horizontal scanning in the predetermined area are sequentially accessed, and the data is scanned in the horizontal direction by the horizontal scanning address counter 24. The length of the background between the end of the character line and the beginning of the last character line is detected by the horizontal direction detector 28, and is stored in the horizontal direction feature memory 32.
A projected image of the length of the background is stored. Similarly, a projected image of the length of the background in the vertical direction is stored in the vertical feature memory 33 by the operations of the vertical scanning address counter 25 and the vertical direction detection section 29. Similarly, a projected image of the length of the background in the diagonally upward right direction is stored in the diagonally upward right direction characteristic memory 34 by the operations of the diagonally upward right scanning address counter 26 and the diagonally upward right direction detecting section 30 . Furthermore, in the same manner, a projected image of the length of the background in the diagonally downward right direction is stored in the diagonally downward right direction characteristic memory 35 by the operations of the diagonally downward right scanning address counter 27 and the diagonally downward right direction detecting section 31 .

マツチング回路36は、メモリ32,33,3
4,35に貯えられている4方向の走査によつて
得られた各特徴と辞書37の出力信号とを比較
し、その比較結果に基づいて認識部38で文字の
認識を行い、認識結果を出力する。
The matching circuit 36 includes memories 32, 33, 3
Each feature obtained by scanning in four directions stored in 4 and 35 is compared with the output signal of the dictionary 37, and based on the comparison result, character recognition is performed in the recognition unit 38, and the recognition result is Output.

第5図a,bには、上記本発明による文字認識
方式を用いて、第1図a,bに示した従来の文字
認識方式との対比を示したものである。水平、垂
直、斜め右上り方向の走査に対しては、符号化さ
れる領域は第5図a,bともに領域Aとなり、し
かも全く同様のパターンが得られる。斜め右下り
方向の走査に対して符号化される領域は、第5図
bでは領域A,Bとなり、突き出し2が存在する
にもかかわらず、第5図aとbの符号化領域の差
は領域Bのみである。したがつて、第1図bに示
した従来の方式に比べて、本発明の文字認識方式
では突き出し2の存在によつて、同一の文字を別
の文字として誤認識することを回避できる。
FIGS. 5a and 5b show a comparison between the character recognition method according to the present invention and the conventional character recognition method shown in FIGS. 1a and 1b. For scanning horizontally, vertically, and diagonally upward to the right, the area to be encoded is area A in both a and b of FIG. 5, and exactly the same pattern is obtained. The regions encoded for scanning in the diagonally downward right direction are regions A and B in FIG. 5b, and despite the presence of protrusion 2, the difference between the encoding regions in FIG. Area B only. Therefore, compared to the conventional method shown in FIG. 1b, the character recognition method of the present invention can avoid erroneously recognizing the same character as different characters due to the presence of the protrusion 2.

また、第6図には、数字「3」(a図)、「5」
(b図)とを本発明による斜め右上り方向の走査
によつて文字線に囲まれた部分の背景の長さを抽
出し、それぞれ投影軸41,42に投影像43,
44として得ると、「3」と「5」に対応する投
影像43と44とは、その特徴に大きな差異を生
じる。したがつて、水平、垂直方向の走査では判
別がつきにくい字種についても斜め方向の走査を
行うことによつて正確に判別できる。
Also, in Figure 6, the numbers ``3'' (Figure a) and ``5'' are shown.
(B figure) is scanned diagonally upward to the right according to the present invention to extract the length of the background of the part surrounded by the character line, and the projected images 43 and 43 are projected onto the projection axes 41 and 42, respectively.
44, the projected images 43 and 44 corresponding to "3" and "5" have a large difference in their characteristics. Therefore, character types that are difficult to distinguish by scanning in the horizontal and vertical directions can be accurately discriminated by scanning in the diagonal direction.

【図面の簡単な説明】[Brief explanation of drawings]

第1図a,bは従来の文字認識方式のパターン
図、第2図a,bは同じく従来の文字認識方式の
他の例を示すパターン図、第3図は本発明による
文字認識方式の一実施例を説明するためのパター
ン図、第4図は上記実施例のブロツク図、第5図
a,bは本発明による文字認識方式が適用された
パターン図であり、従来の第1図a,bに示した
パターン図と対比するものであり、第6図a,b
は本発明による文字認識方式が適用されたパター
ン図であり、従来の第2図a,bに示したパター
ン図と対比するものである。 22…観測部、23…メモリ、24,25,2
6,27…アドレスカウンタ、28,29,3
0,31…検出部、32,33,34,35…メ
モリ、36…マツチング、37…辞書。
Figures 1a and b are pattern diagrams of a conventional character recognition system, Figures 2a and b are pattern diagrams showing other examples of the conventional character recognition system, and Figure 3 is a pattern diagram of a character recognition system according to the present invention. FIG. 4 is a block diagram of the above embodiment, and FIGS. 5a and 5b are pattern diagrams to which the character recognition method according to the present invention is applied. This is to be compared with the pattern diagram shown in Figure 6a and b.
is a pattern diagram to which the character recognition method according to the present invention is applied, and is compared with the conventional pattern diagrams shown in FIGS. 2a and 2b. 22...Observation unit, 23...Memory, 24, 25, 2
6, 27...address counter, 28, 29, 3
0, 31...Detection unit, 32, 33, 34, 35...Memory, 36...Matching, 37...Dictionary.

Claims (1)

【特許請求の範囲】 1 文字パターンを水平、垂直並に斜め方向に走
査し、文字線ではさまれた文字背景の長さを抽出
し、上記各方向に垂直な軸に投影し、各投影像を
記憶する記憶手段と、 各投影像を記憶した該記憶手段の出力と基準の
辞書に予め貯えられた文字毎の各投影像を比較す
る比較手段とを具備し、 上記比較手段の出力により各投影像の差異に基
づき文字の認識を行うことを特徴とする文字認識
方式。
[Claims] 1. A character pattern is scanned horizontally, vertically, and diagonally, the length of the character background sandwiched between the character lines is extracted, and the length of the character background sandwiched between the character lines is extracted and projected onto an axis perpendicular to each of the above directions, and each projected image is and a comparison means for comparing the output of the storage means storing each projection image with each projection image for each character stored in advance in a reference dictionary, and the output of the comparison means A character recognition method characterized by recognizing characters based on differences in projected images.
JP55189131A 1980-12-29 1980-12-29 Character recognition system Granted JPS57113185A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP55189131A JPS57113185A (en) 1980-12-29 1980-12-29 Character recognition system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
JP55189131A JPS57113185A (en) 1980-12-29 1980-12-29 Character recognition system

Publications (2)

Publication Number Publication Date
JPS57113185A JPS57113185A (en) 1982-07-14
JPH0155510B2 true JPH0155510B2 (en) 1989-11-24

Family

ID=16235915

Family Applications (1)

Application Number Title Priority Date Filing Date
JP55189131A Granted JPS57113185A (en) 1980-12-29 1980-12-29 Character recognition system

Country Status (1)

Country Link
JP (1) JPS57113185A (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0609474A1 (en) * 1993-02-03 1994-08-10 International Business Machines Corporation Method and apparatus for transforming an image for classification or pattern recognition
US5872725A (en) * 1994-12-05 1999-02-16 International Business Machines Corporation Quasi-random number generation apparatus and method, and multiple integration apparatus and method of function f

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS5357927A (en) * 1976-11-04 1978-05-25 Norprint Ltd Method and device for identifying character

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS5357927A (en) * 1976-11-04 1978-05-25 Norprint Ltd Method and device for identifying character

Also Published As

Publication number Publication date
JPS57113185A (en) 1982-07-14

Similar Documents

Publication Publication Date Title
JP3302147B2 (en) Document image processing method
US5799115A (en) Image filing apparatus and method
JP3343864B2 (en) How to separate words
JPH0546591B2 (en)
JPH0155510B2 (en)
US4232290A (en) Handwritten character identification
US5267325A (en) Locating characters for character recognition
JP2825072B2 (en) String recognition device
JPH0514952B2 (en)
JP2902097B2 (en) Information processing device and character recognition device
JPH07141465A (en) Method for detecting inclination of document image
JPS5953983A (en) Detecting and segmenting method of character
JPH07168911A (en) Document recognition device
JP3276554B2 (en) Format recognition device and character reader
JP2727755B2 (en) Character string recognition method and apparatus
JPH03126188A (en) Character recognizing device
JP2659182B2 (en) Character recognition device
JPS603676B2 (en) Intersection extraction method
JPS5831028B2 (en) character recognition device
JPH0467674B2 (en)
JPS62126485A (en) Character segmentation system
JPH08235311A (en) Character recognition device
JPS58207183A (en) Character discriminating system
JPH05114048A (en) Method and device for recognizing character
JPS62175889A (en) Character feature extracting method