JPS5955581A - Region separating method of printed document picture - Google Patents

Region separating method of printed document picture

Info

Publication number
JPS5955581A
JPS5955581A JP57165582A JP16558282A JPS5955581A JP S5955581 A JPS5955581 A JP S5955581A JP 57165582 A JP57165582 A JP 57165582A JP 16558282 A JP16558282 A JP 16558282A JP S5955581 A JPS5955581 A JP S5955581A
Authority
JP
Japan
Prior art keywords
dimensional fourier
document
peak point
region
peak
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
Application number
JP57165582A
Other languages
Japanese (ja)
Inventor
Masahiko Hase
雅彦 長谷
Tanji Hoshino
星野 担之
Akihiro Shimizu
明宏 清水
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.)
Nippon Telegraph and Telephone Corp
Original Assignee
Nippon Telegraph and Telephone 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 Nippon Telegraph and Telephone Corp filed Critical Nippon Telegraph and Telephone Corp
Priority to JP57165582A priority Critical patent/JPS5955581A/en
Publication of JPS5955581A publication Critical patent/JPS5955581A/en
Pending 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
    • G06V30/40Document-oriented image-based pattern recognition
    • G06V30/41Analysis of document content
    • G06V30/413Classification of content, e.g. text, photographs or tables

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Artificial Intelligence (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)
  • Character Input (AREA)
  • Facsimile Image Signal Circuits (AREA)

Abstract

PURPOSE:To separate easily a graphic region from another region of a different pitch, by performing a two-dimensional Fourier conversion while operating the whole printed document at a prescribed local region and then observing the peak position on the Fourier conversion surface and the absolute value. CONSTITUTION:The signal obtained by picking up a document picture 1 is preprocessed 10, and then a two-dimensional Fourier conversion process 11 is carried out to the whole document picture. Then the 1st peak point close to an original point is detected in a Fourier converted space, and the row space and the position angle are detected 12 for the document picture as a whole. The position of a region having different character train and pitch, if any, is taken into consideration to obtain the space of the character row. A small matrix size is decided 13 after obtaining the row space of the 1st peak point, and the whole document is scanned 14 for each size of the small matrix. Then the peak point is detected 16 through a two-dimensional Fourier conversion 15. Thus a graphic region can be easily separated from another region having a different character train and pitch.

Description

【発明の詳細な説明】 (1)発明の属する分野の説明 本発明は、既存の本や印刷文書中の情報を自動的に入力
する装置において、ピッチの異々る文字列の領域および
図形領域Q切り分けをする方法に関するものである。
DETAILED DESCRIPTION OF THE INVENTION (1) Description of the field to which the invention pertains The present invention provides an apparatus for automatically inputting information in existing books and printed documents. It concerns the method of separation.

(2)従来の技術の説明 従来の印刷文書の領域を切υ分けする方法は。(2) Description of conventional technology What is the conventional method of dividing the area of a printed document?

第1図に示すように2文書画像を縦横方向に画像を投影
し、濃度のヒストグラムを作成し、ヒストグラムの極端
に変化する場所を判別することによシ1文字列領域と図
形領域の切υ分けを行う方法である。しかし本方法では
、第2図に示すように。
As shown in Figure 1, by projecting two document images vertically and horizontally, creating a density histogram, and determining the places where the histogram changes extremely, it is possible to This is a method of dividing. However, in this method, as shown in FIG.

ピッチの異なる文字列が存在する場合、または入力画像
が傾いている場合には1周辺分布を正確に取ることが不
可能となる。また、第3図に示すように9文書画像の黒
ラン白ランのランレングスの統計的性質まだは黒画素の
密度を加算することによシ、ピッチの異なる文字列およ
び図形領域の切り分けを行う方法がある。しかし本方法
には、処理時間およびランレングスの閾値の設定に問題
があった。
If there are character strings with different pitches, or if the input image is tilted, it is impossible to accurately obtain a one-marginal distribution. In addition, as shown in Figure 3, character strings and graphic areas with different pitches can be separated by adding the density of black pixels based on the statistical properties of the run lengths of black and white runs in nine document images. There is a way. However, this method has problems in setting thresholds for processing time and run length.

(3)発明の目的 本発明の目的は、これらの欠点を解決することにあり1
文書画像全体を小領域に分割してスモールマトリックス
サイズでスキャンしながら空間的2次元フーリエ変換を
行い、2次元フーリエ変換のピーク値の絶対値の変化に
より、異なるピッチで書かれた文字領域および図形領域
を切り分けるものである。以下2図面について詳細に説
明する。
(3) Purpose of the invention The purpose of the present invention is to solve these drawbacks.
Split the entire document image into small areas and perform spatial two-dimensional Fourier transform while scanning with a small matrix size. Character areas and figures written at different pitches are determined by changing the absolute value of the peak value of the two-dimensional Fourier transform. It separates the area. The two drawings will be described in detail below.

(4)  発明の構成および作用の説明/(Z + y
)を入力された印刷文書画像の濃度分布とすると、2次
元フーリエ変換後の空間周波数成分g(ω工、ωV)は
次のようにあられすことができる。
(4) Explanation of the structure and operation of the invention/(Z + y
) is the density distribution of the input printed document image, the spatial frequency component g(ω, ωV) after two-dimensional Fourier transformation can be expressed as follows.

g(ω1ωy)=、/J′/(zIy)exp(−<(
ω1z+ωr v) ) dz tiy一般に、既存の
本や原稿の中の文字列は周期性をもっているために、そ
の周期性に対応したωX。
g(ω1ωy)=, /J'/(zIy)exp(-<(
ω1z+ωr v) ) dz tiyGenerally, character strings in existing books and manuscripts have periodicity, so ωX corresponds to that periodicity.

ωVの所にピークが生じる。文字ピッチの違う文字列が
印刷文書中に含まれる場合でも、その周期に対応したω
X、ωVの所にピークが生じる。
A peak occurs at ωV. Even if a printed document contains character strings with different character pitches, the ω
A peak occurs at X and ωV.

第4図は1文字列が傾いていない場合の文字列と2次元
フーリエ変換された平面の図形を示す。
FIG. 4 shows a character string and a two-dimensional Fourier-transformed plane figure when one character string is not tilted.

図中、1は入力画像、3は2次元フーリエ変換された像
、4は原点に近い第1ピーク点である。
In the figure, 1 is an input image, 3 is a two-dimensional Fourier transformed image, and 4 is a first peak point near the origin.

第1ピーク点に対応するωVの逆数が、入力画像におけ
る最大の文字列の行間隔となる。第2ピーク点に対応す
るωVの値は別のピッチの文字列の行間隔となる。
The reciprocal of ωV corresponding to the first peak point becomes the maximum line spacing of the character string in the input image. The value of ωV corresponding to the second peak point becomes the line spacing of character strings of different pitches.

印刷文書全体の行間隔が求められた後に局所的に見るだ
めのウィンドウの画素数を決定する。そのウィンドウの
マスクの大きさを決定する場合の条件は1文字列を2次
元フーリエ変換した場合にピークが十分に検出できる行
数を有することである。
After the line spacing of the entire printed document has been determined, the number of pixels in the locally visible window is determined. The condition for determining the size of the window mask is that it has a sufficient number of lines where a peak can be detected when one character string is subjected to two-dimensional Fourier transform.

スモールマトリクスのサイズが決定されれば。Once the size of the small matrix is determined.

第5図に示すように9文書画像全体をそのウィンドウサ
イズにしたがってスキャンし、タイミングをはかつて、
2次元フーリエ変換を行う。図中。
As shown in Figure 5, the entire nine document images are scanned according to their window size, and the timing is
Performs two-dimensional Fourier transform. In the figure.

5はマトリックスの移動方向、そして6はスモールマト
リックスを示す。このとき、スモールマトリックスの2
次元フーリエ変換値のピーク点の大きさと位置を監視し
ているととによって、変換面にピークが出現すれば文字
列領域であり、ピークが検出できなければ図1形領域で
あるものと判定する。
5 indicates the moving direction of the matrix, and 6 indicates the small matrix. At this time, 2 of the small matrix
By monitoring the size and position of the peak point of the dimensional Fourier transform value, if a peak appears on the transform surface, it is determined that it is a character string region, and if no peak is detected, it is determined that it is a Figure 1 type region. .

スモールマトリックスの2次元フーリエ変換面でのピー
ク点の位f6は、スモールマトリックスサイズが決定さ
れれば1文字列のピッチが同じ場合には、同じ位置に出
現するはずであるが1文字列のピッチが異なる場合は、
第6図のように2次元フーリエ変換面でのピークの位置
“はずれてくる。
The position f6 of the peak point on the two-dimensional Fourier transform surface of the small matrix should appear at the same position if the pitch of one character string is the same if the small matrix size is determined, but the pitch of one character string is If they are different,
As shown in FIG. 6, the position of the peak on the two-dimensional Fourier transform plane begins to shift.

図中、A、Bは2つの異なるピッチの文字列を表わし、
7u:2次元フーリエ変換のピーク位置を表わす。この
方式を利用することによって文字列のビツヂの異なる領
域も判別することが可能である(例えば、学会誌上の本
文およびあらましの部分)。
In the figure, A and B represent character strings with two different pitches,
7u: represents the peak position of two-dimensional Fourier transform. By using this method, it is possible to distinguish between areas with different character string bits (for example, the main text and summaries of academic journals).

次に第7図に示すように1文字列が傾いている場合では
、まず文書画像全体のフーリエ変換のピーク点の位置(
ωZ−u TωV=V>  により1文字列の傾いてい
る角度を求めることができる。具体的には1文字列が傾
いている場合でも、原点に近い第1ピーク点に対応する
ωπ、ωVの値$ + 11は2文字の繰り返し周波数
を示すから、その逆数が入力画像における文字列の行間
隔になる。
Next, if one character string is tilted as shown in Figure 7, first the position of the peak point of the Fourier transform of the entire document image (
The angle at which one character string is tilted can be determined by ωZ−u TωV=V>. Specifically, even if one character string is tilted, the value of ωπ and ωV corresponding to the first peak point near the origin $ + 11 indicates the repetition frequency of two characters, so the reciprocal is the character string in the input image. The line spacing will be .

捷た。第8図r(示すように、ピーク点の位置関係u 
+ 11より1文字列の妬いている角度θ−−−1−を
求めることができる。なお、Zは全画素数を表わす。そ
して、第9図に示すように、その角度θKしたがってス
キャンして、スモールマトリックスサイズで順次2次元
フーリエ変換を行い2文字列が傾いていない場合と同様
にピーク値を観察することによって文書画像の領域を判
別することが可能である。
I cut it. Figure 8 r (as shown, the positional relationship of the peak points u
+11, it is possible to find the jealous angle θ---1- of one character string. Note that Z represents the total number of pixels. Then, as shown in Fig. 9, the document image is scanned according to the angle θK, sequentially subjected to two-dimensional Fourier transform with a small matrix size, and the peak value is observed in the same way as when the two character strings are not tilted. It is possible to determine the area.

第10図に本発明の実施例のブロックダイアグラムを示
す。
FIG. 10 shows a block diagram of an embodiment of the present invention.

まず文書画像1をTVカメラ9で撮像してその信号を取
り込み、濃度補正ディジタルフィルタ等の前処理10を
行う。その後、2次元フーリエ変換処理11を文書画像
全体に行い、フーリエ変換された空間で原点付近の第1
ピーク点を見い出し。
First, a document image 1 is captured by a TV camera 9, its signal is taken in, and preprocessing 10 such as a density correction digital filter is performed. After that, two-dimensional Fourier transform processing 11 is performed on the entire document image, and the first point near the origin in the Fourier transformed space is
Find the peak point.

文書画像全体の行間隔および位置角度の検出12を行う
。文字列・ピッチの異なる領域が存在する場合は、その
点の位置も考慮し2文字列の間隔を求める。第1ピーク
点の行間隔が求まった後に。
The line spacing and position angle of the entire document image are detected 12. If there are regions with different character strings and pitches, the distance between the two character strings is determined by taking into account the position of the point. After finding the line spacing of the first peak point.

スモールマトリックスサイズ決定13を行い、そのサイ
ズで文書全体をスモールマトリックススキャン14でザ
ーチし、2次元フーリエ変換15を行い、そのピーク点
検出16をすることによって。
By determining a small matrix size 13, searching the entire document using the small matrix scan 14, performing a two-dimensional Fourier transform 15, and detecting peak points 16.

任意の文書の領域を判別することができる。Areas of any document can be determined.

2次元フーリエ変換は、ディジタル処理以外にも光学的
にレンズとディテクタによりリアルタイムで処理するこ
とも可能である(レンズによるリアルタイムで2次元フ
ーリエ変換装置の製品例:Deft Laborato
ries Inc、THE 5eries 200 )
In addition to digital processing, two-dimensional Fourier transform can also be processed optically in real time using a lens and detector (product example of a two-dimensional Fourier transform device using a lens in real time: Deft Laborato
ries Inc, THE 5eries 200)
.

(5)効果の説明 以上説明したように、印刷文宵中の文字列の周期性に着
目し、全体に2次元フーリエ変換を適用し2行間隔を求
め、それをもとにスモールマトリックスのサイズを決定
し、そのマトリックスサイズのウィンドウによって文1
画像全体をスキャンして2次元フーリエ変換のピーク点
を調べることによってピーク点の位置および大きさによ
って。
(5) Explanation of the effect As explained above, we focused on the periodicity of the character strings in the printed text, applied two-dimensional Fourier transform to the whole, found the two-line spacing, and based on that, the size of the small matrix. , and by the window of that matrix size, sentence 1
By the position and size of the peak points by scanning the entire image and examining the peak points of the two-dimensional Fourier transform.

印刷文書中の領域を判別することが可能である。It is possible to determine regions in a printed document.

文字列が傾いている場合では、全体のフーリエ変換時に
角度を検出することが可能であるので、同様な方式で領
域を判別することが可能である。
If the character string is tilted, it is possible to detect the angle during Fourier transformation of the entire character string, so it is possible to determine the area using a similar method.

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

第1図は従来の領域判別法による入力画像とそのヒスト
グラム、第2図は従来の領域判別法による文字列が傾い
た場合の入力画像とヒストグラム。 第3図は従来の文書領域判別法、第4図は入力画像と2
次元フーリエ変換面、第5図はスモールマトリックスの
サイズと移動方向、第6図1スモールマトリツクスの位
ti6と2次元フーリエ変換のビ一りの位置、第7図は
文字列が傾いている場合の原画像と2次元フーリエ変換
像、第8図は文字列が傾いている場合の傾き検出方法、
第9図は文字列カ傾いている場合のスモールマトリック
スと移動方向のそれぞれの説明図である。そして第10
図は実施1列のブロック図である。 図中、1け入力画像、2はヒストグラム、3は2次元フ
ーリエ変換像、4は第1ピーク点、5はマトリックス移
動方向、6はスモールマトリックス、7は2次元フーリ
エ変換ピーク位置、8は文字列の位置、8αは文字列の
傾き、10は画像前処理、11は2次元フーリエ変換、
12は文字列の行間隔・位置・角度の検出、13はスモ
ールマトリックスサイズ決定、14はスモールマトリッ
クススキャン、15は2次元フーリエ変換、16はピー
クの位置・大きさ検出、をそれぞれ表わす。 特許出願人 日本電信電話公社 代理人弁理士 森 1)  寛 一4ζ 第1図 茹 2 図 (台 男、ン 第 5 (2) 第 9 図 第 6 図 第 7(21 第 8 図
Fig. 1 shows an input image and its histogram obtained by the conventional area discrimination method, and Fig. 2 shows an input image and histogram obtained when a character string is tilted by the conventional area discrimination method. Figure 3 shows the conventional document area discrimination method, and Figure 4 shows the input image and 2
Dimensional Fourier transform surface, Figure 5 shows the size and movement direction of the small matrix, Figure 6 shows the small matrix position ti6 and the position of the two-dimensional Fourier transform, and Figure 7 shows the case where the character string is tilted. The original image and the two-dimensional Fourier transform image, Figure 8 shows the method of detecting the inclination when the character string is incline.
FIG. 9 is an explanatory diagram of the small matrix and the movement direction when the character string is tilted. and the 10th
The figure is a block diagram of the first implementation. In the figure, 1 is the input image, 2 is the histogram, 3 is the 2D Fourier transform image, 4 is the first peak point, 5 is the matrix movement direction, 6 is the small matrix, 7 is the 2D Fourier transform peak position, 8 is the character column position, 8α is the slope of the character string, 10 is image preprocessing, 11 is two-dimensional Fourier transform,
Reference numeral 12 indicates line spacing, position, and angle detection of a character string, 13 indicates small matrix size determination, 14 indicates small matrix scanning, 15 indicates two-dimensional Fourier transformation, and 16 indicates peak position/size detection. Patent Applicant Patent Attorney for Nippon Telegraph and Telephone Public Corporation Mori 1) Kanichi 4ζ Figure 1 Boiled Figure 2 (Daio, N, Figure 5 (2) Figure 9 Figure 6 Figure 7 (21 Figure 8)

Claims (3)

【特許請求の範囲】[Claims] (1)印刷文書中のピッチの異なる文字列領域および図
形領域を切り分ける方法において、印刷文書全体を決め
られた局所領域(マトリックスサイズ)で操作しながら
2次元フーリエ変換を、行い。 そのフーリエ変換面でのピーク点を観察しながら。 そのピーク点の位置−驚よび絶対値をもとにしてピッチ
の異たる文字列領域および図形領域を切シ分けることを
特徴とする領域切シ分は方法。
(1) In a method of separating character string regions and graphic regions having different pitches in a printed document, a two-dimensional Fourier transform is performed while operating the entire printed document in a predetermined local region (matrix size). While observing the peak point on the Fourier transform surface. A method for dividing an area into a character string area and a graphic area having different pitches based on the position and absolute value of the peak point.
(2)第1項において2次元フーリエ変換を行う場合に
文書画像の濃度情報をディジタル的に人、力し、フーリ
エ変換面でのピークを検出することを特徴とする領域切
り分は方法。
(2) An area segmentation method characterized in that, in the case of performing two-dimensional Fourier transformation in item 1, density information of a document image is digitally inputted and peaks on the Fourier transformation plane are detected.
(3)  第1項において、2次元フーリエ変換を行う
場合に文書画像の濃度情報を光学的に入力し。 瞬時に7−リエ変換を行い、その変換面での情報を撮像
素子を周込て入力し、ピーク点を検出することを特徴と
する領域切シ分は方法。
(3) In item 1, when performing two-dimensional Fourier transformation, the density information of the document image is optically input. A region cutting method characterized by instantaneously performing a 7-lier transform, inputting information on the transform plane through an image sensor, and detecting a peak point.
JP57165582A 1982-09-22 1982-09-22 Region separating method of printed document picture Pending JPS5955581A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP57165582A JPS5955581A (en) 1982-09-22 1982-09-22 Region separating method of printed document picture

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
JP57165582A JPS5955581A (en) 1982-09-22 1982-09-22 Region separating method of printed document picture

Publications (1)

Publication Number Publication Date
JPS5955581A true JPS5955581A (en) 1984-03-30

Family

ID=15815090

Family Applications (1)

Application Number Title Priority Date Filing Date
JP57165582A Pending JPS5955581A (en) 1982-09-22 1982-09-22 Region separating method of printed document picture

Country Status (1)

Country Link
JP (1) JPS5955581A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS6180969A (en) * 1984-09-28 1986-04-24 Fuji Xerox Co Ltd Picture signal processor

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS6180969A (en) * 1984-09-28 1986-04-24 Fuji Xerox Co Ltd Picture signal processor
JPH0457274B2 (en) * 1984-09-28 1992-09-11 Fuji Xerox Co Ltd

Similar Documents

Publication Publication Date Title
US6970606B2 (en) Automatic image quality evaluation and correction technique for digitized and thresholded document images
US4656665A (en) Thresholding technique for graphics images using histogram analysis
JP5844783B2 (en) Method for processing grayscale document image including text region, method for binarizing at least text region of grayscale document image, method and program for extracting table for forming grid in grayscale document image
US5563403A (en) Method and apparatus for detection of a skew angle of a document image using a regression coefficient
JP3950777B2 (en) Image processing method, image processing apparatus, and image processing program
JPH0737103A (en) Inclination angle detection device
CA1316605C (en) Method of searching binary images to find search regions in which straight lines may be found
JPS5955581A (en) Region separating method of printed document picture
Gopakumar A survey on image splice forgery detection and localization techniques
JPH0718812B2 (en) Foreign object detection method
JPH03126181A (en) Area dividing method for document image
JPS58201182A (en) Character and graph demarcating method
US5764788A (en) Strand orientation sensing
JP2841373B2 (en) Pattern inspection equipment
JPS60153567A (en) Method for extracting area in printed document picture
JP2647458B2 (en) Halftone dot area separation device
JPH10176997A (en) Method and apparatus for detecting cyclic flaw
JP2853140B2 (en) Image area identification device
KR100537829B1 (en) Method for segmenting Scan Image
JP3024929B2 (en) Pixel feature extraction method for corresponding point search and corresponding point search method for stereo image
JPH04157577A (en) Positioning device
JP4458210B2 (en) Object detection method
JPH0564396B2 (en)
JPH03126180A (en) Image processing device
JPH0760459B2 (en) Corner detector