JPH09179936A - Ruled-line erasing method and character recognition method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To securely erase ruled lines from an unknown document and recognize characters with high accuracy without being affected by the skew of an image. SOLUTION: Long black runs are extracted (108) from an OR-compressed binary image (106) of a binary image (102) of a document and the ruled-line rectangle formed by uniting the black runs is extracted (112). The coordinate values of the black runs in the ruled-line rectangle are converted (116) into coordinate values on the original binary image (102). The coordinate values after the conversion are used to convert (118) black pixels in an area corresponding to the black runs on the original binary image (102) or black pixels in an area obtained by enlarging the area of the black runs into white pixels (118), thereby obtaining a binary image from which the ruled lines are erased.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to image processing of documents or forms, and more particularly to erasing ruled lines and character recognition in documents or forms.

[0002]

2. Description of the Related Art Many documents or forms are provided with a frame in which characters are entered. As a method of reading characters written inside such a frame, the frame and the characters inside the frame are recorded in different colors, and the signal obtained by optically scanning the document or the form is separated into signals of each color. The method (JP-A-56-9877) was common. However, with this method, at the stage of creating a document or form, it is necessary to print a frame or the like in a specific color determined in advance in consideration of the characteristics of the scanner, which increases the cost of creating the document or form. However, there are inconveniences such as having to use a writing instrument of a specific color for writing characters in the frame.

In recent years, documents and forms printed on plain paper by a word processor or the like are often used without distinguishing between frames and characters by color. As a method of separating the frame and the characters in the frame in such a document or form, a histogram is created in the horizontal direction, and a ruled line is extracted from the peak position,
A method of recognizing characters after removing the ruled line portion
1-196382 and JP-A-2-7183) are known.

[0004]

However, Japanese Patent Application Laid-Open No.
In the method of 196382, since the position of the ruled line is detected from the histogram of the black pixels, there is a problem that the range of the ruled line cannot be accurately detected when the document has an inclination. Further, in the method disclosed in Japanese Patent Laid-Open No. 2-7183, since the area in which the ruled line is present is detected by dividing the area in the direction perpendicular to the line, the area to be divided is not determined for the unknown form. Therefore, there is a problem that the existence range of the ruled line cannot be accurately detected.

Therefore, it is an object of the present invention to provide a means for accurately erasing ruled lines for an unknown document or form even if the image is tilted. An object of the present invention is to provide means for performing high-accuracy character recognition on an unknown document or form having ruled lines (including underlines, the same applies below) even when the image is tilted. It is in. Another object of the present invention is to provide means for reducing the amount of memory required for the processing for erasing ruled lines.

[0006]

According to the first aspect of the present invention,
A black run extraction step of extracting a black run having a length equal to or greater than a certain value on the binary image stored in the first memory;
A ruled line rectangle extracting step of integrating black runs within a predetermined distance among the black runs extracted in the black run extracting step, and extracting a rectangle including all the integrated black runs as ruled line rectangles; A ruled line erasing step for converting black pixels in the binary image in the first memory into white pixels corresponding to the black runs included in the ruled line rectangle extracted in the ruled line rectangle extracting step. .

According to a second aspect of the present invention, a black run extraction step of extracting a black run having a length of a predetermined value or more on the binary image stored in the first memory, and the black run extraction step are performed. Ruled line rectangle extraction step of integrating black runs within a predetermined distance among the selected black runs and extracting a rectangle including all the integrated black runs as a ruled line rectangle, and a binary value in the first memory A black run in a ruled line rectangle extracting step for extracting a black run having a length equal to or more than a certain value inside the ruled line rectangle extracted in the ruled line rectangle extracting step from the image; and a black run in the binary image in the first memory. A ruled line erasing step of converting a black pixel corresponding to the black run extracted in the ruled line rectangle black run extraction step into a white pixel.

According to a third aspect of the present invention, in the ruled line erasing method according to the first or second aspect of the present invention, in the first memory, 2
Storing a binary image corresponding to the value image in a second memory, the black run extracting step including the second image.
The black run is extracted using the binary image in the memory.

According to a fourth aspect of the present invention, in the ruled line erasing method according to the first aspect of the present invention, the binary image in the first memory is OR-compressed at a predetermined reduction ratio to generate a second binary image.
And storing the coordinate values of the black run extracted by the black run extraction step included in the ruled line rectangle extracted by the first ruled line rectangle extraction step on the binary image in the first memory. And a coordinate conversion step for converting the coordinate values of the black pixels to be converted into white pixels based on the coordinate values of the black run converted by the coordinate conversion step. And

According to a fifth aspect of the present invention, in the ruled line erasing method according to the second aspect of the present invention, the binary image in the first memory is OR-compressed at a predetermined reduction ratio to generate a second binary image.
And a coordinate conversion step of converting the coordinate values of the ruled line rectangle extracted in the first ruled line rectangle extraction step into coordinate values on the binary image in the first memory. The step of extracting black runs within a ruled line rectangle is characterized by recognizing a range in which a black run should be extracted based on the coordinate values of the ruled line rectangle converted by the coordinate conversion step.

According to a sixth aspect of the present invention, in the ruled line erasing method according to any one of the first to fifth aspects of the invention, the ruled line erasing step is such that black pixels inside a region obtained by enlarging a black run region are white. It is characterized by conversion into pixels.

According to a seventh aspect of the present invention, in the ruled line erasing method according to the sixth aspect of the invention, the enlargement amount of the area obtained by enlarging the area of the black run depends on the width of the ruled line rectangle including the black run. It is characterized by being controlled.

According to an eighth aspect of the present invention, in the ruled line erasing method according to any one of the first to fifth aspects of the present invention, the black run which is erased from the binary image after the ruled line is erased by the ruled line erasing step. Step for extracting an isolated black pixel connected rectangle smaller than a predetermined size within a certain distance from the position of, and a black pixel connected rectangle extracting step for the binary image after the ruled line is erased. And a step of converting white pixels inside the ruled line rectangle into black pixels inside the black pixel connected rectangle extracted by.

According to a ninth aspect of the present invention, a ruled line in a binary image is erased by the ruled line erasing method according to any one of the first to eighth aspects of the invention, and the ruled line is erased by the process. The character image of the value image is cut out and the character recognition is performed.

According to a tenth aspect of the present invention, in the character recognition method according to the ninth aspect, only the ruled lines parallel to the character lines are erased in the process of eliminating the ruled lines in the binary image. And

According to an eleventh aspect of the present invention, a frame recognition process for recognizing a frame surrounded by a ruled line on a binary image stored in the first memory and a character in the frame recognized by the frame recognition process. A character recognition process for cutting out an image from a binary image in the first memory and character recognition of the cut out character image; a ruled line erasing process for erasing ruled lines in the binary image in the first memory; When there is a character with a low certainty factor in the recognition result of the recognition process, a character image is again cut out from the binary image after ruled line elimination by the ruled line elimination process for at least the character with a low certainty factor, and character recognition is performed again. Character recognition processing is included.

According to a twelfth aspect of the present invention, in the character recognition method according to the eleventh aspect of the present invention, the ruled line erasing processing is performed by the ruled line erasing method according to any one of the first to eighth aspects. It is characterized by performing.

[0018]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings in order to clarify the embodiments of the present invention.

[First Embodiment] FIG. 1 is a block diagram showing a first embodiment of the present invention. With reference to this block diagram, the device configuration and processing contents of the present embodiment will be described according to the flow of processing. First, a binary image input unit 10 such as an image scanner
0 reads a document such as a document or form, and
The value image data is stored in the binary image memory 102.

Next, in the compressed image generation unit 104, 2
Generating compressed binary image data by compressing the binary image data in the value image memory 102 at a predetermined compression rate,
This is stored in the compressed binary image memory 106. At this time, a compression ratio of 1/4, 1/8, etc., which is easy to process in a digital operation system, is adopted. As the compression method, OR compression is adopted. Although a method other than the OR compression can be adopted, in order to ensure the extraction of the ruled line even when the ruled line is cut due to the influence of the resolution of the image scanner or the error of binarization, as will be described later, In this embodiment, OR compression is adopted. The reason why the black run extraction method is applied to the compressed binary image data instead of the original binary image is generated in the peripheral portion of the image due to the resolution of the image scanner, the error of binarization, etc. as described later. This is because it is easy to deal with fluctuations.

The compressed binary image is used only for the black run extraction described later, and the target of the ruled line elimination described later is 2.
It is the original binary image data in the value image memory 102. The reason why the black run extraction target image memory and the ruled line erasing target image memory are separately prepared is to facilitate the erasure of both ruled lines in the case of a document in which vertical and horizontal ruled lines are mixed. That is, if one memory is shared by both processes, for example, if a horizontal ruled line is extracted and the horizontal ruled line is erased and then a vertical ruled line is extracted, a black run of a vertical ruled line that intersects with the horizontal ruled line is generated. Although a part cannot be extracted, such inconvenience can be avoided by dividing the memory. Since the compressed binary image data has a smaller data size than the original binary image data, the required memory amount does not increase so much even if two memories are provided. This point is also a benefit obtained by performing the black run extraction on the compressed binary image data.

When the compressed binary image data is generated,
The black run extraction unit 108 scans the compressed binary image data, extracts a black run having a length equal to or greater than a certain value that is likely to be a black run forming a ruled line, and extracts the data (coordinates of the start point and the end point). ) Is extracted to the black run memory 110. In the case of processing a document having a table area including vertical and horizontal ruled lines, both black runs in the horizontal direction (main scanning direction) and black runs in the vertical direction (sub scanning direction) are extracted. Although it is possible to extract black runs in both directions at the same time, in the present embodiment, black run extraction is performed in one direction, and ruled line extraction and ruled line deletion in the same direction are performed, and then the same is performed in the other directions. Processing is performed. When it is desired to erase only a line parallel to a character line such as an underline in a document, only a black run in a direction parallel to the character line is extracted.

When the black run data is obtained, the ruled line extraction unit 1
In step 12, the black run extracted is integrated into a rectangle (called a ruled line rectangle) that includes all black runs within a predetermined distance range. Then, the data of the ruled line rectangle (the diagonal vertex coordinates of the ruled line rectangle and the coordinates of the black run integrated in the ruled line rectangle are stored in the black run memory 110).
Information for further extraction) is stored in the ruled line memory 114.

When at least one ruled line rectangle has been extracted, the coordinate conversion unit 116 reads the diagonal vertex coordinate values of the ruled line rectangle from the ruled line memory 114 and integrates the ruled line rectangle based on the data in the ruled line memory 114. The coordinate values of the start point and the end point of each of the black runs are stored in the black run memory
Are read, and these coordinate values (which are the coordinate values on the compressed binary image) are converted (mapped) into the coordinate values on the original binary image before compression, and the conversion result is the ruled line erasing unit 118.
Send to

FIG. 2 schematically shows this coordinate conversion. Reference numeral 200 is a ruled line rectangle on the compressed binary image, and 201 is a ruled line rectangle mapped on the original binary image. L1, L2, L3, and L4 are black runs that form a ruled line and are integrated into a ruled line rectangle. Black run is OR
Since it is extracted from the compressed binary image, the black run mapped to the original binary image is slightly thicker by OR compression than the original black run on the original binary image. Also, on the original binary image, there may be a slight break in the black run of the ruled line due to the influence of the resolution of the scanner or the error of the binarization, but this break becomes a form connected by OR compression, so that An unbroken ruled black run is mapped to the original binary image data.

The ruled line erasing unit 118 converts the black pixel corresponding to the black run into a white pixel in the original binary image in the binary image memory 102 using the converted coordinate values provided by the coordinate converting unit 116. By converting, ruled lines are erased. In this ruled line erasing process, a plurality of processing modes can be selected in relation to the control of the erasing range. The control of the erase range in each mode will be described below.

1) Mode A The range of black runs after coordinate conversion is used as it is as an erasing range.
Even in this mode, the erasing range is wider than the black run range of the original ruled line on the original binary image by the amount that the black run of the ruled line is thickened by the OR compression. Therefore, even if there is some fluctuation in the original binary image, or even if there is a small break in the ruled line, the ruled line can be reliably erased.

2) Mode B A range wider than the range of black runs after coordinate conversion by a certain width is set as an erasing range. Even if the image fluctuation is large, the black pixels on the ruled line can be reliably erased.

3) Mode C A range that is wider than the range of black runs after coordinate conversion by a certain ratio (for example, 1/4) of the ruled line rectangle width is set as an erasing range.
When the expansion width of the erase range is fixed (in mode B),
If the enlargement width is too small, the influence of fluctuation cannot be sufficiently removed, and if the enlargement width is too large, black pixels such as characters close to the ruled line may be excessively erased, making it difficult to select the enlargement width. On the other hand, in this mode C, the erasure range can be appropriately adjusted according to the size of the fluctuation by paying attention to the property that the size of the fluctuation of the image is proportional to the width of the ruled line.

Black run extraction in the required direction, ruled line rectangle extraction,
When the ruled line deletion is completed, the character cutout unit 120 performs character cutout on the binary image data after the ruled line deletion in the binary image memory 102, and gives the character image to the character recognition unit 122. The character recognition unit 122 performs character recognition on the character image and outputs the recognition result. The character cutting may be performed by a method similar to the conventional method, such as a method using projection or a method by extracting and integrating black pixel connected rectangles. Since the ruled lines that obstruct character extraction have been removed from the image in advance, it is possible to reliably cut out the characters, and therefore the characters in the table area and the underlined characters can also be recognized with high accuracy.

[Embodiment 2] FIG. 3 is a block diagram showing a device configuration of this embodiment. A big difference between this embodiment and the first embodiment is that the black run memory (110) existing in the first embodiment is omitted, as is clear by comparing FIG. 3 and FIG. The point is that one black run extraction unit 300 and its processing are added.

The processing content will be described. Binary image input unit 10
When the value is 0, the binary image data of the original is stored in the binary image memory 102, and the binary image data obtained by OR-compressing the binary image data at a predetermined compression rate is generated by the compressed image generation unit 104 to generate a compressed binary image. Memory 1
06. From this compressed binary image data, a black run having a length equal to or greater than a certain value is extracted by the black run extraction unit 108A. However, the extracted black run data is not stored in the memory and is directly sent to the ruled line extraction unit 112A. The ruled line extraction unit 112A uses the given black run data, integrates black runs within a predetermined distance into a rectangle, extracts a ruled line rectangle, and stores the ruled line data in the ruled line memory 114A. Note that, unlike the first embodiment, the black run memory (1
10) does not exist, the data of the ruled line rectangle stored in the ruled line memory 114A need only be the coordinates of the two corners of the diagonal of the ruled line rectangle and do not include information for referencing the black run. Since the value of the diagonal two-vertex coordinates of this ruled line rectangle is the coordinate on the compressed binary image, it is converted to the coordinate on the original binary image data before compression by the coordinate conversion unit 116A, and the coordinate value after conversion is It is sent to the added black run extraction unit 300.

In the black run extraction unit 300, the range of the ruled line rectangle of the original binary image data in the binary image memory 102 is scanned to extract black runs having a certain value or more, and the starting point of the extracted black runs is set. The coordinates of the end point are sent to the ruled line erasing unit 118. The ruled line erasing unit 118 uses the binary image memory 102 based on the given coordinate values of the black run.
The ruled line erasing process is performed on the binary image data inside. The specific contents of this ruled line deletion are the same as in the case of the first embodiment.

As described above, in this embodiment, it is possible to erase the ruled lines as in the first embodiment without using the memory (black run memory 110) for storing the black run data, and there is no memory space. This is advantageous when the function of each processing unit is realized by software using a personal computer or the like.

When ruled line elimination is completed, a character image is cut out from the binary image data after ruled line elimination obtained in the binary image memory 102 by the character cutout unit 120 as in the case of the first embodiment, and the character corresponding thereto is cut out. The recognition is performed by the character recognition unit 122.

[Third Embodiment] FIG. 4 is a block diagram showing the arrangement of the device according to the present embodiment. 2 by the binary image input unit 100
The device configuration and processing contents from the input of the value image data to the ruled line erasing unit 118 are the same as those in the first embodiment. However, the ruled line erasing unit 118 performs ruled line erasing in the mode A in the first embodiment. That is, only the black pixels corresponding to the black run range of the ruled line after the coordinate conversion are converted into white pixels.

In this embodiment, a black pixel concatenated rectangle extraction section 400 and its processing, a black pixel concatenated rectangular memory 402 related thereto, a rectangular white pixel conversion section 404 and its processing are added. The black pixel concatenated rectangle extraction unit 400 scans the binary image after the ruled line is erased by the ruled line eraser 118 in the binary image memory 102 within a certain distance from the position of the erased black run, An isolated black pixel connected rectangle smaller than a certain size is extracted, and the coordinates of the two diagonal vertices of the rectangle are stored in the black pixel connected rectangle memory 402. The in-rectangle white pixel conversion unit 404 refers to the data of the black pixel connection rectangle in the black pixel connection rectangle memory 402, and extracts the extracted black pixel connection of the binary image data after ruled line deletion in the binary image memory 102. The process of replacing the inside of the rectangle with a pixel is performed.

By extracting such a black pixel connected rectangle and replacing it with a white pixel, black pixels of a ruled line (due to image fluctuations) that cannot be removed only by replacing only the black run range with white pixels. Can also be reliably removed. FIG. 5 is an explanatory diagram thereof. FIG. 5A shows a ruled line rectangle and its black run. (B) shows the result of erasing ruled lines in the range of black runs, but black runs r1 and r2 due to image fluctuations are left unerased. Therefore, an isolated black pixel connected rectangle within a range of a distance Δd from the position of the black run of the erased ruled line is detected, and the inside of the rectangle is replaced with a white pixel, thereby remaining black runs r1 and r2.
Can also be erased.

In the ruled line erasing B or C of the above-described embodiment, all black pixels in a range wider than the black run are replaced with white pixels. Therefore, for example, some pixels of a character string close to the ruled line are excessively erased. There is a risk of doing it. On the other hand, according to this embodiment, only small black pixel connected rectangles are extracted and erased. Therefore, only black pixel connected components such as black runs r1 and r2 due to image fluctuation or the like should be erased and should not be erased. It is possible to prevent excessive erasure of black pixels.

Character extraction and character recognition for the binary image data after the above ruled line erasing processing are the same as in the first embodiment.

In this embodiment, similarly to the second embodiment, the ruled line rectangle is extracted without using the black run memory 110, and then the black run in the ruled line rectangle is extracted, which is described above. It may be modified so as to execute ruled line removal, extraction of black pixel connected rectangles, and conversion into white pixels.

[Fourth Embodiment] FIG. 6 is a block diagram showing the arrangement of the device according to the present embodiment. With reference to this block diagram, the device configuration and processing contents of the present embodiment will be described along the processing flow.

Binary image data of the original is input by the binary image input unit 100 and stored in the binary image memory 102. Based on this binary image data, the ruled line extraction unit 6
00 extracts black runs in the main-scanning direction and the sub-scanning direction having a length of a certain length or more, and black runs in each direction within a certain distance are integrated into a rectangle, and integrated blacks (ruled line rectangles) Data (coordinates of two diagonal corners of a ruled line rectangle) is stored in the ruled line memory 602. The frame extracting unit 604 refers to the ruled line rectangle data in the ruled line memory 602 and recognizes the frame surrounded by the ruled lines in the main scanning direction and the sub-scanning direction using the coordinates outside the ruled line rectangular frame. , Frame data (coordinates of two diagonal corners of the frame) are stored in the frame memory 606. Regarding the recognition of such a frame, the applicant's Japanese Patent Application No. 7-3
It is described in detail in the specification and drawings attached to the application No. 20066.

Next, the characters inside the recognized frame are recognized. That is, the character cutout unit 608 cuts out the image data of the character inside each frame from the binary image data in the binary image memory 102, and stores it in the character image memory 610. The method of cutting out a character image may be a method of scanning an image in a frame, extracting black pixel connected rectangles, and cutting out a character area by integrating them. At this time, the applicant's Japanese Patent Application No. 7-32
As described in detail in the specification and drawings attached to the application of No. 0066, a character entered from an adjacent frame such as a black pixel connecting rectangle in contact with a frame or a black pixel connecting rectangle included in a ruled line rectangle. It is also possible to adopt a method of cutting out characters by excluding the black pixel connected rectangle determined to be. Also,
It is also possible to adopt a method in which the line area is cut out by projecting the image in the frame in the row direction, and the character area is cut out by projecting each line area in the vertical direction.

Next, with respect to the cut out character image, the character recognition unit 612 performs character recognition such as feature extraction and matching with a recognition dictionary, and language processing by language knowledge or grammatical knowledge. An evaluation value (for example, a value obtained by normalizing the distance from the dictionary by the number of character contours) is stored in the recognition memory 614. Next, the certainty factor calculation unit 616 calculates the certainty factor, which is the degree of certainty of each recognition result, and stores it in the recognition memory 614. The method of determining the certainty factor is, for example, Japanese Patent Application No. 7-29422 filed by the present applicant.
The detailed description is omitted here because it is described in detail in the specification and drawings attached to the application of No. 3. The evaluation value itself is also a kind of certainty factor although it is based on the information on a character-by-character basis, and thus it can be used as the final certainty factor.

When the certainty factor of the recognition result of all the characters in all the frames thus obtained is higher than a certain level, the recognition result obtained up to this stage is the final recognition result. It is output to the outside and is not executed because the processing for deleting ruled lines described below is unnecessary. Of course, in parallel with the above character recognition processing, the following ruled line erasing processing may be unconditionally executed. The process for deleting the ruled line is at least for one character in one frame,
If the certainty factor is lower than a certain level, it is always executed.

The processing for erasing the ruled lines is the same as that in the first embodiment, and the compressed image generation unit 104 generates compressed binary image data and the compressed binary image memory 106.
To the compressed binary image data, the black run extraction unit 108 extracts a black run having a length equal to or more than a certain value, the black run data is stored in the black run memory 110, and the ruled line extraction unit 112 integrates the black runs. Extraction of ruled line rectangle by
The ruled line rectangular data is stored in the ruled line memory 114, the ruled line rectangular data is converted by the coordinate conversion unit 116, and the ruled line erasing unit 118 erases the ruled line based on the conversion result. However, this ruled line erasing process may be performed by the same method as in the second or third embodiment. Further, the ruled line erasing process may be performed only on the range related to the frame including the character with low confidence.

When there is a character with a low certainty factor in at least one frame, the character cutout unit 608 extracts the character within the frame containing the low certainty factor with respect to the binary image data after the ruled line deleting process described above. The character recognition unit 616 executes again the character recognition for the cut out character image, and the reliability calculation unit 616 executes the reliability calculation for the recognition result again. Then, the character recognition unit 612 compares the certainty factor calculated after each recognition result with the certainty factor calculated before, and when the certainty factor calculated afterwards is higher, the character recognition unit 612 is in front of the recognition memory 614. The recognition result of is rewritten with the recognition result of later.

As described above, when the certainty factor of the character recognition result using the binary image data in which the ruled line rectangle is not erased is low, the binary image data after the ruled line erase is performed at least for the frame including the character with the low certainty factor. The characters in the table area of the document or the form can be recognized more accurately because the recognition result is corrected again by using.

[0050]

According to the first aspect of the present invention, since ruled lines are erased in units of black runs that form ruled lines, even if there is an image skew, accurate ruled line elimination is possible without being affected by it. Is.

According to the second aspect of the present invention, since the ruled lines are erased in units of black runs, the ruled lines can be erased accurately without being affected by the image skew. According to the first aspect of the present invention, since the black run data extracted by the black run extraction process is used in the ruled line erasing step, the black run data is stored until the ruled line erasing step is completed. You need to prepare. According to the present invention, the black run in the ruled line rectangle is extracted immediately before the ruled line erasing step, and the extracted black run data is used for the ruled line erasing. Therefore, a memory for saving the black run data is unnecessary, The amount of memory required can be reduced.

According to the third aspect of the present invention, since the black run extraction target image memory and the ruled line erasing target image memory are prepared separately, the vertical and horizontal ruled lines are different from those when the same memory is used for both processes. The process of erasing ruled lines on mixed documents becomes easy.

According to the invention described in claim 4 or 5, as in the invention described in claim 3, the process of erasing mixed vertical and horizontal ruled lines becomes easier as compared with the case of one memory, and the second memory is compressed. Since the stored image is stored, the size of the second memory can be reduced, and thus the memory required for processing can be reduced. In addition, a black run is extracted by using OR-compressed binary image data, and this black run is
Since it maps on the value image and converts black pixels to white pixels,
Accurate ruled line deletion is possible even if a small break occurs in the ruled line or if there is image fluctuation around the ruled line.

According to the sixth aspect of the invention, by making the area for converting the black pixel into the white pixel wider than the area of the black run, the image fluctuation occurs due to the influence of the resolution of the image scanner and the binarization error. In this case, the ruled lines can be surely erased.

Image fluctuations are likely to occur in the peripheral portion of the ruled lines due to the influence of the resolution of the image scanner and the binarization error, but the magnitude of this fluctuation tends to be proportional to the width (thickness) of the ruled lines. According to the invention of claim 7,
By adjusting the size of the area for converting the black pixels into the white pixels according to the width of the ruled line rectangle, unnecessary black pixels due to image fluctuation can be appropriately erased.

According to the eighth aspect of the present invention, by replacing the black pixel connecting rectangles within a fixed distance from the position of the erased black run with white pixels, only the black run range becomes white pixels. Black pixels of a ruled line (generated by image fluctuation) that cannot be completely removed only by replacement can be surely removed. Further, when all black pixels in a range wider than the black run are replaced with white pixels, for example, some pixels of a character string close to a ruled line may be excessively erased. However, the invention according to claim 8 According to this, only small black pixel connected rectangles are extracted and erased, so that it is possible to prevent excessive erasure of black pixels that should not be erased.

According to the ninth aspect of the invention, after the ruled lines which obstruct the character segmentation are removed from the image in advance,
Since character cutting is performed, reliable character cutting is possible, and therefore characters close to a ruled line and characters with underlines can also be recognized with high accuracy.

According to the tenth aspect of the invention, even in the state where the ruled line exists, a character which comes into contact with or approaches the frame where a clipping error is likely to occur can be recognized with high accuracy by re-recognizing after the ruled line is erased.

According to the invention of claim 11, claim 9 is provided.
Alternatively, the accuracy of character recognition can be further improved by improving the ruled line removal accuracy in the tenth invention.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a device configuration according to a first embodiment of the present invention.

FIG. 2 is a diagram illustrating mapping of ruled lines and a black run on a compressed image onto an original image.

FIG. 3 is a block diagram showing a device configuration of a second embodiment of the present invention.

FIG. 4 is a block diagram showing a device configuration according to a third embodiment of the present invention.

FIG. 5A is a diagram showing an example of a ruled line in which a black run due to image fluctuation is present. (B) It is a figure which shows the state which erased the black run inside the example of the same ruled line.

FIG. 6 is a block diagram showing a device configuration according to a fourth embodiment of the present invention.

[Explanation of symbols]

 100 binary image input unit 102 binary image memory 104 compressed image generation unit 106 compressed binary image memory 108, 108A black run extraction unit 110 black run memory 112, 112A ruled line extraction unit 114, 114A ruled line memory 116, 116A coordinate conversion unit 118 Ruled Line Erasing Unit 120 Character Cutout Unit 124 Character Recognition Unit 300 Black Run Extraction Unit 400 Black Pixel Concatenated Rectangle Extraction Unit 402 Black Pixel Concatenated Rectangle Memory 404 Rectangle White Pixel Conversion Unit 600 Ruled Line Extraction Unit 602 Ruled Line Memory 604 Frame Extraction Unit 606 Frame Memory 608 Character cutout unit 610 Character image memory 612 Character recognition unit 614 Confidence factor calculation unit 614 Recognition memory

Claims (12)

[Claims] 1. A black run extraction step of extracting a black run having a length equal to or greater than a predetermined value on a binary image stored in a first memory, and a black run extracted in the black run extraction step. Of the black runs within a predetermined distance, and a ruled line rectangle extraction step of extracting a rectangle that includes all the integrated black runs as a ruled line rectangle, and a ruled line rectangle extracted in the ruled line rectangle extraction step. A ruled line erasing step of converting black pixels in the binary image in the first memory into white pixels corresponding to a black run. 2. A black run extraction step of extracting a black run having a length equal to or greater than a certain value on a binary image stored in a first memory, and a black run extracted in the black run extraction step. A ruled line rectangle extracting step of integrating black runs within a predetermined distance of, and extracting a rectangle including all the integrated black runs as a ruled line rectangle; and the ruled line rectangle from the binary image in the first memory. A black run within a ruled line rectangle extracting a black run having a length equal to or greater than a certain value inside the ruled line rectangle extracted in the extracting step; and a black run within the ruled line rectangle in the binary image in the first memory. A ruled line erasing step of converting a black pixel corresponding to the black run extracted in the extracting step into a white pixel, and a ruled line erasing method. 3. The ruled line erasing method according to claim 1, further comprising a step of storing a binary image corresponding to the binary image in the first memory in a second memory, and the black run extracting step. Is a ruled line erasing method characterized in that a black run is extracted using a binary image in the second memory. 4. The ruled line erasing method according to claim 1,
A step of storing in the second memory a binary image obtained by OR-compressing the binary image in the first memory at a predetermined reduction ratio; and including the ruled line rectangle extracted in the first ruled line rectangle extraction step, A coordinate conversion step of converting the coordinate values of the black run extracted by the black run extraction step into the coordinate values on the binary image in the first memory, and the ruled line erasing step is converted by the coordinate conversion step. A ruled line erasing method characterized in that the position of a black pixel to be converted into a white pixel is recognized based on the coordinate value of the black run. 5. The ruled line erasing method according to claim 2,
A step of storing a binary image obtained by OR-compressing the binary image in the first memory at a predetermined reduction ratio in a second memory; and a coordinate value of a ruled line rectangle extracted in the first ruled line rectangle extraction step, A coordinate conversion step of converting into a coordinate value on a binary image in one memory, and the black run extracting step within the first ruled line rectangle extracts a black run based on the coordinate value of the ruled line rectangle converted by the coordinate conversion step. A ruled line erasing method characterized by recognizing a range to be extracted. 6. The ruled line erasing method according to claim 1, wherein the ruled line erasing step converts black pixels inside a region obtained by enlarging a black run region into white pixels. How to erase ruled lines. 7. The ruled line erasing method according to claim 6, wherein the enlargement amount of the region obtained by enlarging the region of the black run is controlled according to the width of the ruled line rectangle including the black run. Ruled line erasing method. 8. The ruled line erasing method according to claim 1, wherein the binary image after the ruled line is erased by the ruled line erasing step is within a certain distance from the position of the erased black run. A black pixel connected rectangle extracting step for extracting an isolated black pixel connected rectangle smaller than a predetermined size, and a black pixel connected rectangle extracted by the black pixel connected rectangle extracting step for the binary image after the ruled line is erased. A ruled line erasing method comprising: converting a black pixel inside the pixel into a white pixel; 9. A ruled line in a binary image is defined by claim 1.
A character recognizing method comprising: a process of erasing by the ruled line erasing method described in any one of 1. above, and a process of cutting out a character image of a binary image in which the ruled lines are erased and performing character recognition. 10. The character recognition method according to claim 9, wherein in the process of deleting the ruled lines in the binary image, only the ruled lines in the direction parallel to the character line are deleted. 11. A frame recognition process for recognizing a frame surrounded by ruled lines on a binary image stored in a first memory, and a character image in the frame recognized by the frame recognition process for the first memory. Of the recognition result of the character recognition process of cutting out the character image of the extracted character image from the binary image in the first memory, the ruled line erasing process of erasing the ruled line in the binary image in the first memory, and the recognition result of the character recognition process. If there is a character with a low certainty factor, at least for the character with a low certainty factor,
A character recognition method, comprising: re-character recognition processing for cutting out a character image from the binary image after ruled line erasing by the ruled line erasing processing and performing character recognition. 12. The character recognition method according to claim 11, wherein the ruled line erasing process is performed by the ruled line erasing method according to any one of claims 1 to 8.