JPH0433075B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a character line extraction device, and more particularly to a device for extracting a desired character line from a character line including ruby characters.

(Prior Art and its Problems) In general Japanese texts printed on books such as paperbacks, furigana and dots (hereinafter referred to as ruby) may be added to some characters. When characters with ruby attached in this manner are read using character recognition technology, the ruby may cause misreading or rejection. Therefore, it is necessary to separate characters or character lines from ruby. As a device for separating ruby characters, for example, as disclosed in Japanese Patent Publication No. 58-8024 (Japanese Patent Application No. 127855-1983), a device that separates ruby characters from one line of text is used. There is a method of separating based on the blank part of the projection information within a predetermined width from the edge of the side that is not
It is necessary to set a predetermined width in advance depending on the size of the printed characters. However, for ruby when characters of different sizes coexist, such as a chapter title with relatively large characters in a book and a main text line that is smaller than the chapter title, the font size is fixed in advance. cannot be determined. Furthermore,
If the character line and the ruby come into contact with each other due to the resolution of the image input device or the like, it becomes difficult to separate the ruby.

(Objective of the Invention) The present invention has been made to solve the above-mentioned drawbacks of the conventional art. A character with a ruby separation device that solves the above conventional drawbacks by focusing on the property that the position is not easily affected by the presence of ruby, but the width of the character line changes depending on the ruby. The object of the present invention is to provide a line extraction device.

(Structure of the Invention) According to the present invention, in a character line extracted from an optically scanned and quantized text image, the character line is divided into a plurality of partial areas having overlapping areas, and the character line is scanned in the character arrangement direction. means for detecting a projective distribution using
means for detecting both end positions and the average position of a plurality of partial regions according to a projection distribution; means for detecting a plurality of candidate sections in which the number of black pixels on the projection distribution is less than or equal to a certain value; and the average value. and a means for calculating a boundary area between a partial character line and a ruby area from a plurality of candidate areas based on the position of the character line and both end positions; A character characterized by: means for detecting a position where the number of black pixels on a projection distribution obtained in the direction of arrangement of character lines is minimum within a region; and extracting a desired character line according to a ruby separation position. The object of the present invention is to provide a line extraction device.

(Example) The present invention will be described below with reference to the drawings. FIGS. 1d to 1g are diagrams for explaining, as an example, a method of extracting a desired character line from character lines including ruby characters according to the present invention. Figures a, b, c, f, and g in the figure show part of character lines that include ruby, and d and e in the figure show part of character lines that do not include ruby. Further, a in FIG. 3 shows a state in which a part of ruby is in contact with a desired character. Incidentally, character lines as shown in FIGS. 1A and 2D can be extracted using conventional known techniques. It also shows the distribution of the number of black pixels, that is, the projection distribution, obtained by scanning in the direction of the arrangement of characters in character lines b, e, f, and g in the figure. Symbols R ₁ and R ₂ in Figures a and d are obtained when a character line is divided into multiple pieces with overlapping areas using, for example, a predetermined size set based on the width information of the character line. This shows the partial area that will be displayed. Therefore, for the partial region R ₁ shown in figure a, figure b,
It is possible to extract a projection distribution as shown in c for R ₂ and as shown in e in the same figure for R ₁ shown in c in the same figure and d in the same figure. The symbols B ₁ (l), B ₂ (l), and Bi(l) written on the projected distributions b, c, e, and f in the same figure indicate the left end position obtained for each projected distribution, and the same symbol B ₁
(U), B ₂ (U), and Bi(U) indicate the right end position. Further, the symbols μ ₁ , μ ₂ , μi indicate the average positions in the respective projection distributions. _{Therefore , in} the partial region R _{1 shown} in FIGS _.
is calculated using the formula D _l = μ ₁ − B ₁ (l), D _U = B ₁ (U) − μ _1. In the case of d in the same figure, which does not include ruby, the distance D _l and D _U can be considered to be almost equal, while in the case of a in the same figure that includes ruby, the distance D _U is larger than the distance D _l . This is the average value μ ₁ in a of the same figure
This is because it is not easily affected by whether or not ruby is included. Therefore, the vicinity of the boundary between the ruby line and the desired character line can be determined from, for example, the position S ₁ =μ ₁ +D _l shown in FIG. In addition, in Figure d, if the vicinity of the boundary between the ruby line and the desired character line as shown in Figure A is calculated using the average value μ ₁ shown in Figure d and the above-mentioned distance D _l , the above-mentioned The border will include the right edge of the character line, and it is easy to see that no ruby exists.

Next, the i-th (i=1, 2,
A method of detecting a boundary area between a ruby line and a desired character line using the detected projection distribution for the partial area (3, . . . ) will be described. First, as shown by the dotted line in the figure, sections a, b, and c in which the number of black pixels on the projection distribution is equal to or less than a predetermined slice level Ts are detected. Next, when the above-mentioned position S ₁ (=μ ₁ + D _l ) is calculated using the above-mentioned average value μ ₁ and distance D _l and a boundary candidate section having a predetermined allowable width is set from position S ₁ , the above-mentioned The area indicated by the logical product of the areas a, b, c and the boundary candidate area can be determined as the boundary area in the i-th area Ri. For example, in the case of figure f, it is obtained from section b, and in the figure
The interval indicated by Li and Ui is determined as the boundary interval. In the figure, Li represents the left end position of the boundary section, and Ui represents the right end position of the boundary section. Further, the method of calculating the projection distribution based on the average value μi described above is not limited to the method using the predetermined slice level Ts described above.

From the left end position Li and right end position Ui (i=1, 2,...) of the boundary section detected for the plurality of partial regions obtained in this way, the entire character line as shown in g in the same figure is The left end position L of the boundary section for
and U can be easily calculated. Finally, for the projection distribution for the entire character line, from the left end position L to the right end position U of the boundary section, as shown by the dotted line arrow in g in the same figure, the position where the number of black pixels on the projection distribution is the smallest is set to the ruby line. It can be determined as the separation position between the character line and the character line. Note that the projection distribution used when calculating the separation position described above may be detected again for the entire character line, or, for the purpose of improving processing speed, it may be substituted by accumulating the projection distribution of partial areas. It's okay. FIG. 2 is a logic block diagram illustrating a specific embodiment of the present invention. In the figure, 1 is an image memory, and the image memory 1 includes:
Document images such as books and forms are optically scanned and quantized and stored as image information. 2 is a character line detection unit that sequentially detects character lines from the image information stored in the image memory 1 and stores them in the line information storage unit 3;
Stores area information such as the position and size of the character line. Note that the character line detecting unit 2 that inputs the document image described above and sequentially extracts character lines from the image input device that stores it in the image memory 1 and the image information stored in the image memory 1 can be configured by using a known technique. realizable. The partial area calculation unit 4 converts the character lines stored in the line information storage unit 3 into
n (n≧1) partial regions Ri (i=1, 2,...
n), sequentially in the character arrangement direction (for example, the first
After transmitting information on the vertical direction (in figure a) and the position and size of the partial region Ri to the projection distribution extraction unit 5,
The partial region Ri obtained by the projective distribution extraction unit 5
The projection distributions for the partial projections are sequentially stored in the partial projection storage unit 6. The projection distribution extraction unit 5 extracts the projection distribution in the set direction for the set area by scanning and reading out the image memory 1. The average position calculation unit 7 calculates the average position μi explained in FIG. 1 from the projection distributions sequentially transferred from the partial projection storage unit 6. The candidate section detecting section 8 includes a partial projection storage section 6
As explained in FIG. 1, a plurality of sections in which the number of black pixels on the projection distribution is less than or equal to a predetermined slice level Ts are determined for the projection distribution to be transferred in sequence. The boundary section calculation section 9 calculates the boundary sections Li, Ui between the ruby line and the desired character line in the partial area using the average position μi output from the average position calculation section 7 and the candidate section, as explained in FIG. A plurality of sections a, b, c which are the output of the detection unit 8 and the left end position of the partial area
Calculated based on Bi(l) and right end position Bi(U). still,
The left end position and right end position of the partial area described above are detected in the partial area calculation unit 4 when the projection distribution is transferred from the projection distribution extraction unit 5 to the partial area calculation unit 4, and are transferred to the boundary section calculation unit 9. Suppose that it is done.
Next, the boundary section of the partial area Ri (i=1, 2,...n) sequentially output from the boundary section calculation unit 9 (as shown in FIG. 1 f, the left end position of the boundary section is Li, the right end position (denoted by Ui) are transferred to the minimum value calculation unit 10 and maximum value calculation unit 12, respectively. That is,
The left end position Li of the boundary section is sent to the minimum value calculation unit 10,
The right end position Ui value of the boundary section is transferred to the maximum calculation unit 12. Reference numeral 11 denotes a left end storage section, which stores the left end position L of the boundary section between ruby and a desired character line. Note that the left end storage section 11 is set to a very large value as an initial value. Reference numeral 13 denotes a right end storage section, which stores the right end position U of the boundary section between ruby and a desired character line. Note that the right end storage section 13 is set to 0 as an initial value. The minimum value calculation section 10 compares the left end position Li of the boundary section of the partial area Ri (i=1, 2,...n) sequentially output from the boundary section calculation section 9 with the content stored in the left end storage section 11. and the left end position which is the output value of the boundary section calculation unit 9
If Li is smaller than the contents of the left end storage section 11, the left end position Li is transferred to the left end storage section 11, and the contents of the left end storage section 11 are updated. The maximum value calculation section 12 compares the right end position Ui of the boundary section of the partial area Ri (i=1, 2,...n) sequentially output from the boundary section calculation section 9 with the content stored in the right end storage section 13. and the right end position which is the output value of the boundary section calculation unit 9
If Ui is larger than the contents of the right end storage section 13, the right end position Ui is transferred to the right end storage section 13, and the contents of the right end storage section 13 are updated. By performing the above operation for n partial regions Ri (i=1, 2,...n), the left end position L and right end position U of the boundary section between ruby and a desired character line are respectively 11 and the right end storage section 13. The ruby separation position determination unit 14 includes the left edge storage unit 11
The left end position L and right end position U, which are the contents of the right end storage unit 13, and the area information of the character line for which the left end position L and right end position U are to be calculated are stored in the line information storage unit 3.
The region information and the character alignment direction of the character line are transferred to the projection distribution extraction unit 5, and the ruby and The position where the text is separated from the desired character line (hereinafter referred to as ruby separation position) is calculated. That is, in the boundary section indicated by the left end position L to the right end position U, the position where the number of black pixels on the projection distribution in the entire character line is the minimum is calculated as the ruby separation position. In the above explanation, it is assumed that the projective distribution of the entire character line is extracted, but the projective distribution extracting unit 5 extracts the projective distribution of the boundary section indicated from the left end position L to the right end position U of the entire character line. By doing so, the processing time for calculating the projection distribution can be shortened, and furthermore, the partial projection storage unit 6
It is also possible to calculate the cumulative projection distribution of the partial regions Ri (i=1, 2, . . . n) stored in , and use it as a substitute for the projection distribution of the entire character line described above. Based on the ruby separation position obtained by the ruby separation position determination unit 14 and the area information of the character line stored in the line information storage unit 3, a desired character line from which the ruby area has been removed from the character line is extracted from the image memory 1. The extracted characters are transferred to a character segmentation device (not shown in the figure) that can be realized using conventional technology, and separated into individual characters. The above-described operation is stored in the line information storage section 3 and applied to all character lines. Note that among the character lines stored in the line information storage unit 3, in the case of a character line that does not include ruby, all the combinations of ruby and the desired character line are Either the boundary section is not detected, or the right end position U of the detected boundary section includes the right end position of the entire character line, so it is stable and desirable even when applied to a character line that does not include ruby. It is possible to extract a line of characters.

(Effects of the Invention) As explained above, according to the character line extracting device including ruby characters of the present invention, it is possible to extract characters with ruby characters in cases where it is difficult to fix the character size in advance or when a desired character line and ruby characters are extracted. It becomes possible to stably separate the ruby line and the character line even when they touch each other.

[Brief explanation of drawings]

FIGS. 1a to 1g are diagrams for explaining, as an example, a method of extracting a desired character line from character lines including ruby characters according to the present invention. FIG. 2 is a logic block diagram showing a specific embodiment of the present invention. In the figure, 1 is an image memory, 2 is a character line detection unit, 3 is a line information storage unit, 4 is a partial area calculation unit, 5 is a projection distribution extraction unit, 6 is a partial projection storage unit, 7 is an average position calculation unit, 8 is a candidate section detection section, 9 is a boundary section calculation section,
10 is a minimum value calculation section, 11 is a left end storage section, 12 is a maximum calculation section, 13 is a right end storage section, and 14 is a ruby separation position determination section.

Claims (1)

[Claims] 1 Separate the ruby from the text line to which ruby has been added,
An apparatus for extracting a desired character line, comprising: means for extracting a character line from optically scanned and quantized text image information; and a means for extracting a character line from optically scanned and quantized text image information; means for extracting a projective distribution in the direction of arrangement of characters in the character line; means for detecting both end positions and an average position of a plurality of partial regions according to the projective distribution; and black pixels on the projective distribution. means for detecting a plurality of candidate sections whose number is equal to or less than a certain value; and means for calculating a boundary section between a partial character line and a ruby area from the plurality of candidate sections based on the end positions and the average position. Then, the boundary area between the character line and the ruby line is calculated from the boundary sections of the plurality of partial areas, and within the boundary area, the number of black pixels on the projection distribution obtained in the direction of the arrangement of characters in the character line is calculated. A character line extracting device comprising means for detecting a minimum ruby separation position, and extracting a desired character line according to the ruby separation position.