JP2963465B2 - Space detection method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention relates to a method of detecting a space on a character line (not a space between characters but a space character to which a space code is to be assigned. The same applies hereinafter) in a character recognition device. About.

[Conventional technology]

2. Description of the Related Art Conventionally, there is the following method as a space detection method in a character recognition device.

One is that the printing pitch of the document is given in advance as a constant or as a ratio to the character height, and by comparing the given pitch with the pitch between characters that are subject to space detection, This method determines whether there is a space between characters.

The other is a method of taking a histogram of the margin width between characters in a line, setting a valley portion as a threshold, and setting a space when the margin width between characters is wider than this threshold.

[Problems to be solved by the invention]

In the former method, a space may not be detected in the case of a proportional pitch printing (hereinafter, abbreviated as proportional printing) document in which the pitch changes depending on the character width. For example, in a proportional print manuscript, the space between characters with narrow character widths is not so large compared to the pitch when they do not exist,
Detection often fails.

The latter scheme does not work well for fixed pitch printed originals.

Therefore, an object of the present invention is to specify a printing mode (fixed pitch printing / proportional printing) of a document without specifying in advance.
An object of the present invention is to provide a space detection method capable of accurately detecting a space for a document in any print mode.

[Means for solving the problem]

The present invention takes the ratio or difference between the maximum value and the minimum value of the pitch from the pitch distribution measured from the document, determines the document mode (constant pitch printing / proportional printing) based on the ratio or difference, If judged, P
Space is detected based on the ratio of (the pitch between the current character of interest and the previous character) and SW (standard character width). If proportional printing is determined, MS (between the current character of interest and the previous character) The space is detected based on the ratio between the width of the character and the SW. SW is defined by MP (maximum value of vertical projection in one row) × α (predetermined constant).

The mode determination of constant pitch printing or proportional printing is based on the distribution of pitches.If the ratio or difference between the maximum value and the minimum value of the pitch is small, it is determined to be constant pitch printing.
Judge as proportional printing.

(Operation)

Generally, models such as those shown in FIGS. 2 and 3 can be set for a fixed-pitch print original and a proportional print original, respectively. That is, as shown in FIG. 2, the pitch is constant in a fixed-pitch printing document,
The margin width between characters changes depending on the width of characters before and after the margin.
On the other hand, as shown in FIG. 3, in a proportionally printed original, the margin between characters is constant, but the pitch varies depending on the width of the preceding and following characters.

Therefore, if space detection is performed by a method that is inappropriate for the original print mode, as shown in FIGS. 4 and 5,
Obviously, a detection error will occur. That is, as shown in FIG. 4, in a fixed-pitch document, the margin between narrow characters is widened. Therefore, when the document is treated as a proportional print document, the margin between characters is erroneously regarded as a space. Will be detected. Conversely the fifth
As shown in the figure, in a proportionally printed original,
A space between characters having a small width cannot be detected when handled as a fixed-pitch printing document because the pitch between the characters before and after the space is not much different from the case where there is no space between the characters.

As described above, it is understood that it is necessary to appropriately switch the space detection method depending on the document print mode.

According to the present invention, as described above, a space detection method using a pitch when a document is determined to be a fixed-pitch print document, and a space detection method using a margin between characters when a document is determined to be a proportional print document. It is apparent from the above description that these space detection methods are methods suitable for each print mode.

Here, the distribution of the pitch and the margin width between characters will be considered for a fixed-pitch print original and a proportional print original.

In the case of a document printed at a constant pitch, the pitch is constant as is clear from FIG. 2, and the pitch histogram is as shown in FIG. The first peak corresponds to the original pitch, the second peak corresponds to the pitch between characters with one space between them, and the third peak corresponds to the space between two characters.
Letters correspond to the pitch between letters. Since the margin width between characters varies depending on the width of the characters before and after, the histogram is as shown in FIG.

In the case of a proportionally printed original, as is clear from FIG. 3, since the pitch varies considerably depending on the width of the character, the pitch histogram is as shown in FIG.
Since the inter-character margin width is polarized when there is a space between the characters and when there is no space therebetween, the histogram of the inter-character margin width is as shown in FIG.

According to the present invention, the print mode of the document is determined from the pitch distribution as described above. Considering only the possible pitch range when there is no space between characters in a fixed-pitch printed document, the pitch histogram of the fixed-pitch printed document is shown in FIG. 10, and the pitch histogram of the proportional-printed document is As shown in FIG. 11, it can be seen that the respective distributions are significantly different. Therefore,
According to the present invention, it is possible to reliably determine the print mode of a document. As a specific determination method, as is clear from comparison between the two histograms, when the pitch is measured within the pitch search range, the ratio or difference between the maximum value and the minimum value of the pitch is small in a fixed-pitch print document, Since a proportionally printed original is large, a method of comparing the ratio or difference with a predetermined threshold value is possible.

Thus, according to the present invention, accurate space detection can be performed by a space detection method suitable for the print mode of the document without previously specifying the print mode of the document.
In the case of a fixed-pitch print document, space detection is performed using the measured pitch, so that appropriate space detection can be automatically performed for documents of different pitches.

〔Example〕

 Hereinafter, embodiments of the present invention will be described with reference to the drawings.

Example 1 FIG. 12 shows a flow of the entire processing. First, the maximum value MP of the vertical projection in one line of interest is obtained (Process 10
0). This MP is a measure of the character height in a line.

Standard character width as a guide for character width in space calculation
SW is calculated by equation (1) using MP and a predetermined constant α (process 102).

SW = α × MP (1) After the initial setting of the variable (process 104), the pitch data in the pitch search range (in this case, the range not exceeding 2 × SW) is considered to have no space in the fixed-pitch print document. The processing is sequentially performed (processing 106 to processing 114).

FIG. 13 shows a detailed flow of the processing 106 for measuring the pitch P of one character of current interest. The pitch is calculated by the following equation: P = (CME + CMS-OME-OMS) / 2 (2), where CMS and CME are the start position (left end) and end position (right end) of the current character, respectively, as shown in FIG.
Which is detected from the data of the vertical projection. Also, OMS and OME are the start position and end position of the character before the current character as shown in FIG. 14, and in fact, as is clear from FIG. 13, the CMS obtained for the immediately preceding character is The value of CME is used.

When the pitch P measured in this way exceeds twice the standard character width SW (when it exceeds the pitch search range)
, The pitch P is invalidated (process 108). 2 x SW
A pitch P less than is considered valid and is used to perform the necessary update of the maximum value PMAX or the minimum value PMIN of the pitch in the row (steps 110 to 116). It is clear that PMAX and PMIN correspond to the maximum value and the minimum value of the pitch shown in FIGS. 10 and 11, respectively.

When the processes 106 to 116 are executed up to the last character in the line of interest, the print mode of the line of interest is determined by comparing the ratio between PMAX and PMIN with a predetermined threshold TH1 (process 118). As is clear from the description related to FIGS. 10 and 10, the print mode determination result is constant pitch printing when the ratio is small and proportional printing when the ratio is large. Note that the same determination can be made using a difference instead of the ratio between PMAX and PMIN, as described above.

When the printing mode is determined to be constant pitch printing, space detection using the pitch is performed (process 120), and when the printing mode is determined to be proportional printing, the space detection using the margin between characters is performed (process 122). .

FIG. 15 shows a detailed flow of the processing 120 for space detection for constant pitch printing. In this processing, TP = (PMAX + PMIN) / 2 (3) β = TP / SW (4) is calculated.

Next, the pitch P of the current character of interest is measured, and S = β × (P / TP) −1 (5) is calculated. Then, S space codes are set before the current character. That is, S (S characters) spaces are detected between the current character and the preceding character. This is repeated one character at a time until the last character in the line.

The details of the pitch P measurement process here are as shown in FIG.

FIG. 16 shows a detailed flow of the process 122 for space detection of proportional printing. In this process, the inter-character space width MS = CMS-OME (6) between the current character of interest and the preceding character is calculated, and the number of spaces S = MS / SW (7) is calculated. CMS and OME are the positions shown in FIG. 14, and SW is the standard character width.

Then, S space codes are set before the current character.

The same processing is repeated one character at a time until the last character in the line.

Embodiment 2 The flow of the entire process is the same as that of Embodiment 1, and can be shown in FIG.

The difference between the present embodiment and the first embodiment is that when the character width of the current character or the character before it is larger than a certain threshold in the pitch measurement process for determining the print mode (process 106 in FIG. 12). Such a wide character is likely to be a combination of two or more characters. That is, the pitch measurement is not performed (actually, pitch data that is excluded in the process 108 is measured). Do).

FIG. 17 shows a detailed flow of the pitch measurement processing. As is apparent from this figure, a value twice as large as the standard character width SW is used as a threshold for determining the possibility of character connection, and the pitch P = 2 × SW when the character width exceeds 2 × SW. Is not used for updating PMAX and PMIN.

By introducing such a character width check, the stability of character mode determination in a line where a connected character exists is improved.

 The other processing contents are the same as in the first embodiment.

Embodiment 3 The flow of the entire process is the same as that of Embodiment 2, and can be shown as in FIG.

The difference between the present embodiment and the second embodiment is that when the pitch measurement is performed in the constant pitch space detection process (process 120 in FIG. 12), the character width of the current character or the preceding character is wider than a certain threshold. By changing the character width to an appropriate value and then performing pitch measurement, the accuracy of the pitch measurement at the connected character portion is increased.

The overall flow of this space detection process is shown in FIG. 15, but the detailed flow of pitch measurement during this process is shown in FIG.
As shown in FIG. As is apparent from this flowchart, here, a double value of the standard character width is used as a threshold for character width check, and when the character width exceeds the threshold, the character width is set to SW and the pitch is calculated. .

This character width correction is shown in FIG. The other processing contents are the same as in the second embodiment.

Embodiment 4 The flow of the entire process is the same as that of Embodiment 3, and can be shown as in FIG.

The difference between the present embodiment and the third embodiment is that the pitch measured in the pitch measurement for the print mode determination (process 106 in FIG. 12) is stored, and when it is determined that the printing is at the constant pitch, the space detection (the That is, the pitch stored in the process 120) in FIG. 12 is used.

The detailed flow of pitch measurement for print mode determination
It is shown as in Figure 20. Here, MP [I] is the stored pitch. In addition, the same correction of the character width as described in the third embodiment is performed. It should be noted that CH = 2 × SW is added to the data of the pitch P of the portion where the character width correction is required, so that the data is not used for updating PMAX and PMIN.

The detailed flow of space detection processing for constant pitch printing is
It is shown as shown. Other processing contents are the same as those in the third embodiment.

Example 5 The flow of the entire process is as shown in FIG. The processes with the same numbers as in FIG. 12 have the same contents as the corresponding processes in the fourth embodiment. The difference from the fourth embodiment is only the contents of a pitch measurement process 106A for print mode determination and a determination process 108A of whether to use the measured pitch.

The detailed flow of the pitch measurement processing 106A is as shown in FIG. That is, as is clear from comparison with FIG. 20, the pitch P is measured and stored as MP [I] after correcting the character width as necessary as in the fourth embodiment. WG = CMS-OME + CH (7).

Then, in process 108A, the inter-character margin width WG and SW
The pitch data of the character width exceeding 2 × SW is excluded from the data for updating PMAX and PMIN.

Embodiment 6 The flow of the entire process is as shown in FIG. The difference from the fifth embodiment is that the pitch measurement (the
This is the content of the process 108A) in FIG. 22 and the content of the space detection of the proportional printing (the process 122 in FIG. 22).

The detailed flow of pitch measurement for print mode determination
As shown in FIG. As is apparent from a comparison between FIG. 24 and FIG. 23, in the present embodiment, the data of the margin width between characters is stored as WG [I], which is different from the processing content in the fifth embodiment.

The detailed flow of space detection for proportional printing is as shown in FIG. As is apparent from a comparison between FIG. 25 and FIG. 16, in the present embodiment, the number of spaces is calculated using WG [J] stored as a margin between characters at the time of pitch measurement. The other processing contents are the same as in the fifth embodiment.

In each of the embodiments described above, for example, a document image that can be realized by a hardware configuration as shown in FIG. 1 is input by a scanner or the like, and is stored in the image memory 1. The CPU 2 executes line segmentation and character segmentation for the document image, space detection according to the present invention, and the like, and data such as vertical projection extracted at the time of character segmentation is stored in the data memory 3. . Next, space detection described in detail in each of the above embodiments is performed, and data related to this processing and processing result data are also stored in the data memory 3. A program for such processing of the CPU 2 is stored in the program memory 4.

〔The invention's effect〕

As described above in detail, according to the present invention, an appropriate method is automatically selected for both fixed-pitch printing originals and proportional printing originals without any need to specify a printing mode in advance, and a space is reliably detected. It is possible,
In addition, since the space detection is performed using the measured pitch in the case of a fixed-pitch print original, appropriate space detection can be automatically performed in response to the difference in pitch.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an example of a hardware configuration according to each embodiment of the present invention, FIG. 2 is a diagram showing a fixed-pitch printing original model, FIG. 3 is a diagram showing a proportional-printing original model, FIG. And FIG. 5 are diagrams for explaining the relationship between the print mode and space detection of the document, respectively. FIGS. 6 and 7 are diagrams respectively showing the histogram of the pitch and the margin between characters of a fixed-pitch print document. 8 and 9 show histograms of the pitch and the margin between characters of the proportionally printed original, respectively. FIGS. 10 and 11 show the fixed and proportionally printed originals when the pitch search range is limited, respectively. FIG. 12 is a diagram showing a pitch histogram, FIG. 12 is a schematic flowchart of the entire process, and FIG. 13 is a flowchart of pitch measurement for print mode determination. , FIG. 14 is a schematic view for illustrating a character position, the
FIG. 15 is a flowchart of space detection for a fixed-pitch printed document, FIG. 16 is a flowchart of space detection for a proportional-printed document, FIG. 17 is a flowchart of pitch measurement for print mode determination, and FIG. Flowchart of pitch measurement for space detection, FIG. 19 is an explanatory diagram of character width correction, FIG. 20 is a flowchart of pitch measurement for print mode determination, FIG.
FIG. 21 is a flowchart of space detection of a fixed-pitch print original, FIG. 22 is a schematic flowchart of the entire process, FIG. 23 is a flowchart of pitch measurement for print mode determination,
FIG. 24 is a flowchart of pitch measurement for determining a print mode, and FIG. 25 is a flowchart of space detection of a proportionally printed original. 1 ... Image memory, 2 ... CPU, 3 ... Data memory, 4 ... Proportional memory.

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int.Cl. ⁶ , DB name) G06K 9/34 G06K 9/62

Claims (1)

(57) [Claims] 1. A ratio or a difference between a maximum value and a minimum value of a pitch is determined from a distribution of pitches measured from a document, and a document mode is determined based on the ratio or the difference. If space is detected by the ratio of SW and proportional pitch printing is determined, MS and SW
Where P = pitch between the current character of interest and the previous character, SW = MP (maximum value of vertical projection in one line) x α (predetermined constant) MS = current character of interest A space detection method, characterized by a space between characters between a character and a character preceding the character.