JPH02125389A - Space detecting method - Google Patents

Abstract

PURPOSE:To generate a necessary number of space codes by calculating line standard character width and line minimum character width from the maximum value of vertical projection in a line direction and determining the number of spaces of a blank part by using the line standard character width, line minimum character width, and the length of the space part. CONSTITUTION:A CPU 103 refers to the line projection of a line (current line) to be processed which is stored in a line projection storage area 102b according to a line standard character width and line minimum character width calculation program 104c to find the maximum value (line maximum projection value), calculates the current line standard character width and current line minimum character width by using the line projection maximum value, and stores the result in storage areas 102c and 102d in a RAM 102. Then the line standard character width, line minimum character width, and the length of the blank part are used to determine the number of spaces of the blank part. Consequently, the number of spaces can be determined properly.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a method for detecting spaces in each line of a document in a character recognition processing device. [Prior Art] The general flow of processing in a character recognition device is shown in FIG. 10, in which an image of a document to be processed is input (step 100), and lines are cut out from the document image (step 100).
2), step 1003) of cutting out a character from a line, step 1004) of recognizing the cut out character, and step 1005) of generating one or more space codes for a blank part of the line, and the recognized character code and generation. The resulting space code is saved as a recognition result (step 1006). In the space detection step 1005, the number of spaces is determined from the length of the blank part of the line. Conventionally, the number of spaces is determined by a specific calculation using a value corresponding to the standard character width set in advance from the outside and the length of the blank part. The number of spaces was determined. [Problems to be Solved by the Invention] According to the above-mentioned conventional space detection method, even if the document has a fixed pitch, there is the trouble of having to reset the standard character width every time the pitch of the document changes. . Furthermore, most English documents have irregular pitches, and in the case of such documents, it is difficult to appropriately determine the number of spaces using the above conventional method. SUMMARY OF THE INVENTION An object of the present invention is to provide a space detection method suitable for English documents, regardless of fixed pitch, irregular pitch, or character size. [Means for Solving the Problems] The space detection method according to the present invention calculates the line standard character width and line minimum from the maximum value of the projection (number of black pixels) in the direction perpendicular to the line direction in each line of the document to be processed. The character width is calculated, the line standard character width, the line minimum character width, and the length of the blank area are used to determine the number of spaces in the blank area by a specific calculation. Other features of the present invention will be explained with reference to Examples. [Function] Consider the letter h shown in Figure 9 as a model. Assuming that only this character exists in a line, there is a proportional relationship between the height Lh of the character and the maximum value of the vertical projection (number of black pixels) of the line (referred to as the line maximum projection value). °The ratio between Lw and Lh has the property that it does not change greatly even if the font and character size change. Therefore, for appropriate characters Lh, Lw and Ls
By measuring and setting , it is possible to calculate an appropriate line standard character width and line minimum character width for each line using the line maximum projection value determined for each line, for example, using the following equation. Line standard character width = line maximum projection value XLw/Lh (1)
Line minimum character width = line maximum projection value XLs/Lh (2)
In addition, the alphabet has letters that extend upward like h,
Because there are characters that extend downward, such as y and j, and a certain amount of skew occurs in the scanned image of the document, the line width (width in the range where the horizontal projection exceeds the threshold) is determined by the height of the characters. It does not necessarily reflect accurately and is unstable depending on the content of the document and the reading conditions. On the other hand, if the maximum projection value per line is used, such instability can be eliminated and an appropriate line standard character width and line minimum character width can be calculated. By performing specific operations (such as size comparison and division) using the standard line character width and minimum line character width, it is possible to accurately determine the number of spaces even in English documents with irregular pitches. Become. [Embodiments of the Invention] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a schematic block diagram showing a common device configuration according to each embodiment of the present invention. 101 is a scanner for inputting document images; 102 is a RAM for storing images, data related to processing, and processing results; 103 is a CPU that executes processing; and 104 stores programs for each process. It is a ROM. The contents of processing performed by the above device configuration will be explained below for each embodiment. For the image of the document input by the scanner 101,
The CPU 103 performs line cutting according to the line cutting program 104a, and stores the cut out line image in the line image storage area 102a in the RAM 102. This line segmentation can be done, for example, by measuring the horizontal projection (projection on the vertical axis) of a document (in this case, a horizontally written document), and then cutting out the range where the projection value exceeds a certain threshold as the range of lines. done by. Note that the line extraction method itself may be arbitrary, such as dividing a line into a plurality of blocks and performing line extraction by horizontal projection for each block. At the same time as this row extraction, the CPU 103 measures the vertical projection (number of black pixels) of the row image, and stores the measurement result, that is, the row projection, in the storage area 102b. Reference will now be made to FIGS. 2 and 3. According to the line standard character width/line minimum character width calculation program 104c, the CPU 103 refers to the line projection of the line currently being processed (current line) stored in the line projection storage area 102b, and calculates its maximum value (line maximum Projection value) (step 201 in Figure 2), and using this maximum line projection value, calculate the standard character width of the current line and the minimum character width of the current line using equations (1) and (2) above, and use the results. Stored in the respective storage areas 102c and 102d in the RAM 102 (
Step 202). Next, the CPU 103 cuts out characters from the current line image based on the current line projection according to the character cutting program 104b (step 203). This character extraction is performed, for example, by a method in which a range in which the current line projection value exceeds a specific threshold value is defined as a character range, and the other range is defined as a margin, but other methods may also be used. The CPUIO 3 performs character recognition on the cut out character image using the character recognition program 104f (step 204). Note that there is a dictionary necessary for this character recognition, but it is not shown in the figure. The code of the recognized character is stored in the recognition result storage area 102e (step 205). Note that a configuration may be adopted in which feature extraction for character recognition, matching with a dictionary, and the like are executed by dedicated hardware. In addition, regarding the blank part of the current printout extracted by character cutting, the CPU 103 uses the space processing program 104.
Step 206
). This process is as shown in FIG. 3.6 First, the length L of the blank space is compared with the minimum character width of the current line (step 301). When L≦(current line minimum character width), the number N of spaces in the blank section is finally set to 0 (step 302). L> (@minimum forward character width), the value obtained by adding the minimum character width of the current line to L is set again as the blank length length (step 303), and this Ll' is calculated by dividing the current semi-standard character width. Set the value to N (step 3o4). And the N
and 2 are compared (step 305), and when N>2, the value of N (the value rounded off or rounded off after the decimal point) is finally set as the number of spaces N. On the other hand, when it is determined in step 305 that N is less than or equal to 2,
Finally, the number of spaces N is set to 1 (step 306
). In this way, the optimum number of spaces can be determined by calculation using the length of the blank section, the standard character width of the current line, and the minimum character width of the current line, regardless of whether the pitch is irregular or constant, or the character size. Next, the CPU 103 stores a number of space codes corresponding to the determined number of spaces N in the recognition result storage area 102e (step 2o7). The process described above is executed for all lines of the document, and the process for one page is completed. Note that the code and space code of the recognized character are temporarily stored in the RAM 102, and at the end of processing the current line, the character code and space code are merged and stored in the recognition result storage area 102e. It's okay. In addition, in the first embodiment, processing was performed without distinguishing between the leading blank part of each line and other blank parts, but in the present embodiment 2, only the leading blank part of each line was processed compared to other blank parts. Distinguish from
Special processing is performed as described below. ° Other processes are the same as in the first embodiment. In the second embodiment, the current line character standard width and current line minimum character width determined for the first line of the document are used as the standard character width and the minimum character width. This is the width of each storage area 1
02i, 102j. The processing content for determining the blank space is basically the same as that shown in FIG. However, in the process of determining the number of spaces in the blank area at the beginning of each line (the margin area that appears before the characters), instead of the current line standard character width and current line minimum character width determined for the line, storage areas 102i, 1
The standard character width and minimum character width stored in 02j are used. That is, in steps 301 and 303 of FIG. 3, the minimum character width is used instead of the minimum character width of the current line, and in step 304, the standard character width is used instead of the standard character width of the current line. To determine the blank space of
As in the first embodiment, each current line standard character width and current line minimum character width are used. As a natural result, for the first line of the document, the process for determining the number of blank spaces is exactly the same as in the first embodiment. Effects specific to the second embodiment are as follows. In the first embodiment, when the character size differs for each line, even if the length of the margin at the beginning of one line is the same, the standard character width of the current line and the minimum character width of the current line used to determine the number of spaces are different. As a result, the number of spaces for the margin at the beginning of a line varies, and as a result, when the recognition result is output, the length of the margin at the beginning of each line varies, resulting in an unnatural result. On the other hand, in this embodiment 2, for the blank space at the beginning of all lines, the current line display character width and current line minimum character width of one specific line (in this case, the first line of the document) are set to the same width. , leading blanks of the same length will generate the same number of space codes. Therefore, as described above, the unnaturalness associated with the blank space at the beginning of the line can be eliminated. In this embodiment, as in the second embodiment, the blank space at the beginning of each line is distinguished from other blank spaces, and special processing as described below is performed. Processing other than this is the same as in the first embodiment. In this third embodiment, the number of spaces is not determined for the blank section at the beginning of each line of the document, but the length information is
The information is sequentially stored in the storage area 102f on the AM 102. Regarding the margins other than the beginning of the line, the number of spaces is determined using the standard character width of the current line and the minimum character width of the current line, as in the first embodiment. In addition, information on the current line standard character width and current line minimum character width calculated for each line is
The data is sequentially stored in the storage area 102g and LO2h on the M102. After completing character recognition for the last line of the document and generating space codes for blank areas other than the beginning, program 1
04d, calculate the average value of all the current line standard character widths stored in the storage area 102g, store this as the standard character width in the storage area 102i, and
Find the average value of all the current line minimum character widths stored in 02h, and use this as the minimum character width in the storage area 102j.
Store in. Then, the length of the blank section at the beginning of each line stored in the storage area 102f and the storage area 102i. Using the standard character width and minimum character width stored in 102j, the number of spaces at the beginning of each line is determined by the same process as shown in Figure 3, a space code of that number is generated, and the recognition result is It is stored in the required position in the storage area 102e. According to the third embodiment, although the processing is slightly more complicated than that of the second embodiment, the processing result of the blank space at the beginning of a line is more stable than the above embodiment. In Example 3, the average value of the standard character width of each current line and the minimum character width of each current line was used to calculate the number of spaces in each blank space at the beginning of each line. On the other hand, in this embodiment 4, the value that appears most frequently among the standard character widths of each current line is used as the standard character width in determining the number of spaces in the blank area at the beginning of a line, and similarly The value with the highest frequency of occurrence is used as the minimum character width in determining the number of spaces at the beginning of a line. Other than this, the processing is the same as in the third embodiment. According to the fourth embodiment, although the processing is a little more complicated than the third embodiment, a more natural recognition result can be obtained for the blank space at the beginning of one line. The point of this embodiment is that, in addition to the line standard character width and line minimum character width, the horizontal length of the blank area where the vertical projection of each line is less than a certain threshold is also used in the calculation to determine the number of spaces. This is different from each of the embodiments described above. FIG. 4 is a flowchart showing the overall flow of processing. The difference from the processing in each of the above embodiments shown in FIG. The only difference is that the content of step 405 for determining the number of spaces has been changed, and the content of the other processing steps is the same. FIG. 5 is a flowchart showing the processing contents of step 403. In this process, it is determined from the distribution of margin lengths of the current line whether the line is printed with narrow spacing or wide spacing, and the minimum line spacing used in the calculation to determine the number of spaces is determined. Determine the width. Specifically, the minimum line blank width is determined to be a small value for lines printed with narrow space intervals and a large value for lines printed with one wide space interval. That is, from step 501 to step 507, within the character amount blank area (the blank space between the preceding and succeeding characters, excluding the blank areas at both ends of the line) of the current line, 2 of the line standard character width is
The maximum value M of the width of the character amount blank area excluding the character amount blank area which is twice the width or more (this is originally considered to be a blank area with two or more spaces) is determined. Then steps 508,50
At step 9, lines with wide spacing and lines with narrow spacing are determined by comparing the maximum width value M with the line standard character width. However, since there are also lines without spaces, when the maximum width value M is less than the line standard character width, the line without spaces is also determined by comparing it with the line minimum character width. For lines with wide spacing or lines without spaces, step 510 sets the line minimum blank width to half the sum of the line standard character width and the line minimum character width. For other lines, in step 511, one-half of the sum of the line minimum character width and the maximum width value M is set as the line minimum blank width. The determined minimum line blank width is area 10 of RAM 102.
Store in 2k. FIG. 6 is a flowchart showing the processing contents of the space number determination step 405. The difference from the processing in each of the embodiments shown in FIG. 3 is that in step 601, instead of the minimum line sentence width, the minimum line blank width is compared with the length of the blank section. Regarding the blank part at the beginning of the line, the number of spaces may be determined without distinguishing it from the blank part of the character amount as in the first embodiment, or the number of spaces may be determined without distinguishing it from the blank part of the character amount as in the above embodiment The number of spaces may be determined separately from the section. The overall process flow of this embodiment is the same as that of the fifth embodiment shown in FIG. 4, but the content of the space number determination process (step 405 in FIG. 4) is different. FIG. 7 is a flowchart of this space number determination process. In step 701, a character amount blank portion having a width less than the minimum line character width is detected, and in step 702, the number of spaces is determined to be zero. On the other hand, for blank spaces with a character amount exceeding the line minimum character width, the number of spaces is tentatively determined in steps 703 and 704 in the same way as in each of the embodiments described above. Next, compare the total width of the preceding and succeeding characters with the line standard character width (step 705), and if it is less than the line standard character width, the number of spaces is determined by subtracting 1 from the tentatively determined number of spaces (step 706). ), and for other rows, the number of spaces is determined to be O or 1 (steps 708, 70
9). If the number of spaces determined in step 706 is smaller than 0, the number of spaces is set to 0 (step 702). In other words, in this embodiment, when characters with a fixed pitch and a small character width are printed one after another, the width of the blank space is smaller than when the characters before and after have a standard character width. Focusing on the increase in character width, in step 705, character space spaces sandwiched between characters with narrow character widths are identified, and such space space spaces are tentatively determined based on the line minimum character width and line standard character width. Correct the number of spaces by one space. The processing for the blank space at the beginning of the line is as described in Example 1.2 above.
It may be the same as 3.4. The number of spaces is determined by a method that combines the methods of Examples 5 and 6 described above, and a flowchart thereof is shown in FIG. That is, step 8 corresponding to step 701 in FIG.
In No. 01, the difference from the sixth embodiment is that the width of the character amount blank part and the minimum line blank width are compared, but the other processing contents are the same as the sixth embodiment. According to the space detection method of each embodiment described above, spaces can be detected easily and quickly regardless of whether the pitch is fixed or irregular, or the character size is different, and each parameter of the character set by the model is processed. Even if each parameter obtained from the target character differs, the error is absorbed during the processing process and an accurate number of spaces can be obtained. Further, according to the space detection methods of the fifth and seventh embodiments, more accurate space detection is possible by reflecting the width distribution of the character amount blank portion for each line in determining the number of spaces. Furthermore, according to the space detection methods of Embodiments 6 and 7, the number of spaces is determined by considering the character widths before and after a blank space between characters116. Accurate space detection is possible. [Effects of the Invention] As is clear from the above description, according to the present invention, the number of blank spaces can be accurately determined and the necessary number can be achieved regardless of the fixed pitch, irregular pitch, or character size of the document to be processed. It is possible to generate a space code, and it is also possible to prevent the length of a blank section at the beginning of a line from becoming unnatural after recognition, and space processing is relatively simple and can be processed at high speed.

[Brief explanation of the drawing]

FIG. 1 is a schematic block diagram showing the device configuration according to each embodiment of the present invention, FIGS. 2 and 3 are detailed flowcharts of overall processing and space number determination processing in one embodiment of the present invention, and FIG. 4 , FIG. 5 and FIG. 6 are flowcharts of overall processing, minimum blank width determination processing, and space number determination processing in another embodiment of the present invention, and FIG. 7 is a flow chart of space number determination processing in another embodiment of the present invention. 8 is a flowchart of the space number determination process in another embodiment of the present invention, and FIG. 9 is a character flowchart for explaining the determination of the line standard character width and line minimum character width based on the line maximum projection value. FIG. 10, a diagram showing the model, is a flowchart of general processing of the character recognition device. 101...Scanner 102...RAM, 10
3=-CPU, 1104-RO. Figure 2 Figure 4 Figure 6 Section Figure 9

Claims (6)

[Claims] (1) For each line of the document to be processed, calculate the line standard character width and line minimum character width from the maximum value of the projection in the direction perpendicular to the line direction, and calculate the line standard character width, the line minimum character width, and the blank area. 1. A space detection method in a character recognition device, characterized in that the number of spaces in the blank portion is determined using the length of . (2) A claim characterized in that, in the calculation for determining the number of blank spaces at the beginning of each line of a document to be processed, a line standard character width and a line minimum character width calculated for one line of the document are used. The space detection method described in item (1). (3) Use the respective average values of the line standard character width and line minimum character width calculated for each line of the document to be processed in the calculation for determining the number of blank spaces at the beginning of each line of the document. The space detection method according to claim 1, characterized in that: (4) The line standard character width and line minimum character width that have the highest frequency of occurrence calculated for each line of the document to be processed are used in the calculation for determining the number of blank spaces at the beginning of each line of the document. Claim (1) characterized in that the use of
) space detection method described. (5) The space detection method according to claim (1), characterized in that the horizontal distribution of blank areas in each row is used in calculation for determining the number of spaces. (6) The space detection method according to claim 1 or 5, characterized in that the number of spaces in the blank area is corrected based on the widths of the character parts before and after the blank area.