JPH0231286A - Detecting method for special character row - Google Patents

Abstract

PURPOSE:To stably and exactly detect a special row by investigating row width and detecting the special row concerning two united character rows in which a character row with narrow width possible to have the special row not a row pitch, and the preceding and following character rows. CONSTITUTION:Row width Wi of the respective segmented character rows is obtained and a threshold Th1, which is determined from a character size to be defined as standard in a document, and the row width Wi of the respective character rows are compared. Then, the value of character row lower than this threshold is defined as the candidate of the special row. The row width of the united character row, in case that the preceding and following rows are respectively coupled and defined as one row to the row to be defined as the candidate of the special row, is obtained. According to the size relation of the obtained row width (in case the width of the row to be united with the preceding row is defined as WB and the width of the row to be united with the following row is defined as WA) and a threshold Th2 to be obtained from the character size which is defined as the standard in the document, it is decided how row is the special row candidate. For example, in the case of WB>=Th2 and WA>=Th2, the candidate is defined as a noisy row belonging neither the preceding or following rows.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention utilizes image processing technology to extract lines or columns of special characters (simply, (also referred to as speciality).

[Conventional technology]

As this type of detection method, the applicant first examines the pitch t between each character line, and compares the standard pitch obtained based on this with the pitch between each character line to find a set of character lines with a narrow pitch. , by checking the line widths of two character lines, if the width of one is less than the ratio of the standard line width in the document, replace the narrow character line with the wider one. A method has been proposed for detecting special characteristics consisting of rubies, underlines, side points, or lines attached to a character line (see Japanese Patent Application No. 1987-19595).

[vA problem that the invention attempts to solve]

However, with this method, as the proportion of special features in a document increases, the narrow pitch caused by the special features has a greater influence on the standard pitch calculation, and the obtained standard pitch and the special pitch increase. It is said that the difference between the narrow pitch and the normal row becomes smaller, and the detection accuracy of speciality decreases. I have one issue left.

For example, in an extreme case, if you try to read only one line with ruby, since there is only one pitch, it will not be possible to detect the ruby on that line.

Therefore, it is an object of the present invention to make it possible to stably and accurately detect peculiarities, regardless of the proportion of peculiarities contained in a document.

[Means to solve all problems]

When sentence 11 is subjected to image processing to cut out character lines, the width in the line direction (line width) of each line is calculated, and the line width is determined from the standard character size extracted in the line cutting process. Za that searches for a character line with a narrow line width compared to the threshold of 1, and integrates the narrow character line with the previous and subsequent character lines, respectively.
For integrated character lines of type I, each line width is compared with the second threshold value, which is determined from the standard character size and is composed of ruby, underline, side dots, and side line. Detect special character lines.

[Effect]

When detecting special characteristics such as ruby and underline,
By examining the width of each line and detecting the specialness t- for two integrated character lines, each of which is a combination of a narrow character line that has a possibility of speciality rather than the line pitch, and the character lines before and after it. To enable detection of special characteristics without being affected by the proportion of special characteristics included in a document.

〔Example〕

FIG. 1 is a 74-chard showing an embodiment of the present invention. As shown in the figure, this flowchart is as follows (1)
It consists of steps (4).

(1) Using the well-known IIIIlIgI@ technology, lines of text are extracted from the document. At this time, the maximum or most frequent value of the width in the line direction (line!I) is taken as the standard character size within the cutout area.

(2) Find the line Il1gWi of each extracted character line, compare the line @Wi f of each character line with the threshold Thl determined from the standard character size in the document, and find the line @Wi f of each character line below this threshold. Let the character line with , be a candidate for specialness.

(3) For a line that is a candidate for speciality, calculate the line width of the line (integrated character line) when it is combined with the previous line and the following line and regarded as one line.

(4) Obtained line m< line width integrated with previous line t WB
``Threshold value T'' obtained from the line width (WA is the line width integrated with the subsequent line) and the standard font size in the document.
Based on the magnitude relationship with h2, what kind of line is the specificity candidate is determined as follows.

■WB', 'I' h 2 Kawa'! , Th
In the case of 2 (case (41)) Noise line that is not attached to either line after deletion ■ Either WB2wA 10,000 (Th2
(Case (42)) Special characteristics attached to the normal rows that make up the row that satisfies the condition ■WB, Th2, and WA (In the case of T h 2, C(43)
) case] Line t with smaller line width - Specialities attached to the normal lines that constitute it, that is, WB and wAt - Compare, for example, WB>W
If so, it is determined that the speciality is attached to the following line.

A case will now be described in which the document shown in FIG. 2 is given as a target document for special line detection. Note that ηyYi ('=1 to 11) in the figure are the start and end marks of the cut out rows, respectively. Based on the coordinates of each of these cut out rows, row 11Wit-1W i -yl-Yl(i
-1 to 11) Find out. At the same time, a detection threshold value Th1i for special characteristics such as ruby and underline is determined from the standard character size C8 in the document using, for example, Tb1-αC8 (α: constant). This threshold Thx and the row width Wi of each row
, and find rows that are candidates for specificity. In the case of Figure 2, by using an appropriate value of α, it is easy to
and line 9 can be separated from other normal lines as candidates for speciality.

In this way, for rows 3 and 9, which were separated as candidates for specificity, the row width when integrated with the previous and following rows, for example, in the case of row 3, w3B-y, -Y2 ．． w3A-y4-y3 w
3B, find w3A'1. This situation is shown enlarged in FIG. Compare this W3B and W3Ath% with the threshold value Th2 obtained by, for example, Th2-βC8 (β: constant) to determine which row before or after this particularity candidate row 3 is attached to. ,
Or, it is determined whether it is a noise line that does not belong to either category. In this case as well, by using a-cut β, it is easy to determine that row 3 has a special characteristic attached to row 4. From a similar method, it can be determined that row 9 is a special feature attached to row 10.

Although the above targets are for horizontal writing, it is possible to detect targets for vertical writing using the same method.

However, in the case of vertical writing, there is no special feature attached to the 1iJ line (the left line in the case of vertical writing), such as an underline in the case of horizontal writing, so it is determined that it is attached to the previous line. The peculiarity is a noise line.

〔Effect of the invention〕

According to this invention, special characteristics such as ruby and underline can be detected without using line pitch.
Regardless of the proportion of #special lines in the document,
Stable detection becomes possible.

[Brief explanation of the drawing]

FIG. 1 is a flowchart showing the processing procedure of the present invention, and FIG. 2 is an explanatory diagram for specifically explaining the implementation 91t of the present invention, #! FIG. 3 is an enlarged view showing a part thereof. Code explanation: WA, WB, W3A, W3B...Line width, Thl
, Thz...Threshold value, Yl""Yll...
. . . Start coordinates of the row, y1 to ytt . . . End coordinates of the row. Agent Benhori Akio Namiki Agent Benhori Kiyotomi Matsuzaki III 21F No. 3! J

Claims (1)

[Claims] Image processing a document to cut out character lines, find the width in the line direction (line width) of each line, and calculate the line width using the first method determined from the standard character size extracted during the line cutting process. Find a character line with a narrow line width by comparing it with a threshold, and for each of the two merged character lines obtained by merging the narrow character line with the previous and subsequent character lines, calculate the line width from the standard character size. A method for detecting a special character line, comprising detecting a special character line consisting of ruby, underline, side dots, and side lines by comparing each with a second threshold value that is different from the determined first threshold value.