JP2569132B2 - How to determine special character lines - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention uses an image processing technique to extract a special character line or column (simply, Special rows.)
The present invention relates to a method for detecting a character line having a narrow line width and determining whether the character line is attached to any one of the preceding and following character lines or a noise line that does not belong to any of the character lines.

[Conventional technology]

As a detection and discrimination method of this kind, the applicant first examines the pitch between each character line, compares the standard pitch obtained based on this with the pitch between each character line, and finds a set of character lines with a narrow pitch. By examining the line widths of two character lines, if one of the widths is less than a certain ratio compared to the standard line width in the document, the narrower character line is Ruby underline attached to the character line
A method of detecting and determining a special line consisting of a side point or a side line has been proposed (see Japanese Patent Application No. 63-19595).

[Problems to be solved by the invention]

However, in such a method, when the ratio of special lines included in a document is high, the influence of the narrow pitch caused by the presence of the special lines on the calculation of the standard pitch becomes large, and the obtained standard pitch and the special pitch are compared. There is a problem that the difference between the narrow pitch between the row and the normal row is reduced, and the detection accuracy of the special row is reduced.

For example, in an extreme case, even if it is attempted to read only one line with ruby, since there is only one pitch, ruby on that line cannot be detected.

Therefore, the present invention detects the special line stably and accurately regardless of the ratio of the special line included in the document, and determines whether the special line is attached to the preceding or following character line, or An object of the present invention is to be able to determine whether a noise row belongs to neither of them.

[Means for solving the problem]

In order to achieve the above object, according to the present invention, a character line is cut out by performing image processing on a document, a width (line width) in each line direction is obtained, and the line width is determined by a standard character size extracted in a line cutting process. After searching for a character line with a narrow line width by comparing with a first threshold value determined from the above, two integrated character lines obtained by integrating the character line with the narrow line width with the character lines before and after the character line are obtained. Each line width is compared with a second threshold value different from the first threshold value determined from the standard size, and based on the comparison result, a character line with a narrow line width is attached to either the preceding or following character line. It is now determined whether the line is a special character line or a noise line.

[Action]

Detects special lines without any influence on the percentage of special lines included in the document, and determines whether they are attached to the previous or next character line or a noise line that does not belong to either be able to.

〔Example〕

FIG. 1 is a flow chart showing an embodiment of the present invention. As shown in the figure, this flowchart comprises the following steps (1) to (4).

(1) A character line is cut out from a document using a well-known image processing technique. At this time, the largest or most frequent value in the line direction (line width) is set as the standard character size in the cutout area.

(2) cut out to determine the line width Wi of character line, to compare the line width Wi of each character row with a threshold Th _1, which is determined from the character size that is standard in the document, this threshold The following character lines are considered as special line candidates.

(3) The line width of a line (combined character line) in the case where a line that is a special line candidate is combined with the preceding and subsequent lines and regarded as one line is determined.

(4) Obtained from the obtained line width (the line width integrated with the previous line is W _B , and the line width integrated with the subsequent line is W _A ) and the character size standardized in the document the magnitude relation between the threshold value Th ₂ for judges as follows what kind of line is the special line candidate.

For W _B Th ₂ and W _A Th ₂ nozzle rows W _B of [the case (41)] unattached to either row adjoining either 1-way of W _A <For Th ₂ [Case of (42)] Special row attached to the normal row constituting the row that satisfies the condition When W _B <Th ₂ and W _A <Th ₂ [Case (43)] Normal with the smaller row width determining special row associated with the row that is, it compares the W _B and W _a, for example, W _B> W _a special line associated with row after if.

Here, the case where the document shown in FIG. 2 is given as the target document for the special line detection will be described. It should be noted that Y _i and y _i (i = 1 to 11) in the figure are the start and end coordinates of the cut-out line. Based on the coordinates of each of the extracted lines, a line width Wi is obtained by Wi = y _i −Y _i (i = 1 to 11). At the same time, from the standard character size C _S in the document, for example, a threshold value for detecting a special line such as ruby underline by Th ₁ = αC _S (α: constant)
Seek Th _1. Comparing the threshold Th ₁ and row line widths Wi, it detects a candidate to become the line of the special lines. In the case of FIG. 2, by using an appropriate value of α, the rows 3 and 9 can be easily separated from other normal rows as special row candidates.

For the special row candidate obtained in this way, it is determined whether the candidate is a ruby underline or a noise row. That is, the line width when the lines 3 and 9 separated as the special line candidates are integrated with the lines before and after them,
For example, in the case of row 3, W _3B = W ₃ -Y ₂ and W _3A = Y ₄ -Y ₃ are used to determine W _3B and W _3A . This is shown in an enlarged manner in FIG. Then, W _3B and W _3A are compared with, for example, a threshold Th ₂ obtained by Th ₂ = βC _S (β: constant), and the row 3 of this special row candidate It is determined whether the noise line is a noise line that does not belong to any of them.
Again, by using the appropriate β, row 3 becomes row 4
It is easy to determine that this is a special line associated with. By the same procedure, row 9 can be determined to be a special row accompanying row 10.

The above is an example of horizontal writing, but it is also possible to detect vertical writing using the same method. However, in the case of vertical writing, there is no special line attached to the previous line (the left line in the case of vertical writing), such as the underline in the case of horizontal writing, so it is determined to be attached to the previous line. The special row is a noise row.

〔The invention's effect〕

According to the present invention, special lines such as ruby and underline are detected without using line pitch, so that stable detection can be performed regardless of the ratio of the special lines included in the document. Become.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a flowchart showing the processing procedure of the present invention, FIG. 2 is an explanatory view for specifically explaining an embodiment of the present invention, and FIG. FIG. Description of symbols W _A , W _B , W _3A , W _3B …… line width, Th ₁ , Th ₂ …… threshold, Y _{1 to} Y
₁₁ ...... start coordinates of the line, y ₁ ~y ₁₁ ...... end coordinates of the line.

Claims (1)

(57) [Claims] An image processing of a document cuts out a character line, obtains a width in each line direction (line width), and determines the line width from a standard character size extracted in a line cutting process. After searching for a character line with a narrow line width in comparison with the threshold value, for each of two integrated character lines in which the narrow character line is integrated with the preceding and following character lines, the respective line widths are standardized. Are compared with a second threshold value different from the first threshold value determined based on the first and second threshold values. Alternatively, a special character line determination method characterized by determining whether the line is a noise line.