JP3720078B2 - Multicolor character image reader - Google Patents

Description

[0001]
[Industrial application fields]
The present invention relates to a multi-color character image reading apparatus that performs image extraction and character recognition from input of multi-color character images as compared with a form set as a standard.
[0002]
[Prior art]
A conventional form reading apparatus performs character recognition processing only on a recognition target by converting it into data to be dropped out using color designation information used at the time of form creation and data to be recognized.
FIG. 34 is a block diagram showing a conventional form reading apparatus disclosed in Japanese Patent Laid-Open No. 2-123488. In the figure, 511 is a color sensor, 512 is a form, 513 is a lens, 514 is a quantization unit, 515, 516 and 517 are image memories, 518 is a control unit, 519 is a color determination control unit, and 520 is a character recognition unit. .
FIG. 35 is a diagram illustrating an output example of the quantization unit for an image including a field frame, a character frame, and an entry character. In the figure, 531 is a field frame, 532 is a character frame, and 533 is an entry character. In the figure, solid lines, diagonal lines, and shading indicate color differences.
FIG. 36 is an explanatory diagram of the character segmentation operation. In the figure, 541 is a character cutout position.
FIG. 37 is a diagram showing a result of character segmentation. In the figure, 551, 552, 553, and 554 are individual character cutout results.
FIG. 38 is a diagram illustrating an output example of the quantization unit for an image including field frames, preprints, and entry characters. In the figure, 561 is a field frame, 562 is a preprint, and 563 is an entry character. In the figure, solid lines, diagonal lines, and shading indicate color differences.
FIG. 39 is a diagram illustrating a character string cutout result. In the figure, 571 is a character string for the preprint “year”, and 572 is a character string for the preprint “month”.
[0003]
Next, the operation of the apparatus having the above configuration will be described.
First, a form 512 in which a frame indicating a reading area is printed in an arbitrary color and characters are recorded in black, a code (color designation code) for designating a color written for each reading area, and a designated color are dropped out. Format information that stores a code (recognition designation code) that designates whether or not to perform recognition processing is prepared.
The color sensor 511 reads the form 512, decomposes it into red, blue, and green components, performs photoelectric conversion, and stores the data quantized by the quantization unit 514 in the image memories 515 to 517. The control unit 518 converts the color designation code set in the format information given from the outside into color component data for each color component, and outputs it to the color determination control unit 519. The color determination control unit 519 compares the color component data with the density information of the pixel positions in the reading area stored in the image memories 515 to 517, and generates a binary form image based on the determination result. The character recognition unit 520 performs character recognition processing on the binary image.
For example, when a color designation code indicating a print color of a character frame and a code indicating that the character frame is to be dropped out are set in the format information, the color determination control unit 519 deletes the color belonging to the color designation code into white data. Colors other than are converted to black data. The character recognition unit 520 recognizes characters for the image.
[0004]
The image shown in FIG. 35 is obtained from the quantizing unit 514 for the conventional apparatus operating as described above, and the color designation code indicating the print color of the character frame and the print color of the field frame is dropped in the format information. When the code shown is set, the color determination control unit 519 converts the character frame 532 and the field frame 531 into white data, and the entry character 533 into black data.
Further, when the entered character 533 is not entered in the center of the character frame as shown in FIG. 35 and the characters are entered close to each other or when the form image is shifted and input, the character segmentation operates normally. Instead, an incorrect character cutout position 541 shown in FIG. 36 is obtained. The original frame is dropped out and the section becomes unclear. As a result, erroneous character cutout results shown in 551 and 552 in FIG. 37 are obtained. Therefore, the character recognition unit 520 outputs an incorrect recognition result.
[0005]
Further, the image shown in FIG. 38 is obtained from the quantizing unit 514 for the conventional apparatus operating as described above, and the color designation code indicating the preprint print color and the field frame print color is dropped in the format information. When the code indicating this is set, the color determination control unit 519 converts the field frame 561 and the preprint 562 into white data, and the entry character 563 into black data.
Character recognition processing is performed on the binary image shown in FIG. 39 output from the color determination control unit 519. At this time, since the preprint 562 shown in FIG. 38 has already disappeared from the image, the separation of the preprint is not known. Therefore, as shown in FIG. 39, an image of “S61” 571 corresponding to “year” and an image of “2” 572 corresponding to “month” are extracted, and an erroneous character string cutout result is obtained. There is.
[0006]
[Problems to be solved by the invention]
The conventional form reading device is configured as described above, and the frame for classification is removed by dropout, and the relative relationship between the entered characters and the form components other than the entered characters is unknown, so image extraction and There was a problem that character recognition was not possible.
Furthermore, before reading the form, format information corresponding to the form to be read must be specified in advance, and a plurality of types of unregistered forms cannot be read together. Further, when a new reading form is designed or changed, it is necessary to set color designation codes in the format information one by one, and there is a problem that it takes time to set the format information.
[0007]
The present invention has been made in order to solve the above-described problems, and can accurately perform image extraction and character recognition with respect to input of a multicolor character image in which a plurality of types of fields are mixed. It is an object of the present invention to obtain a multicolor character image reading apparatus capable of easily setting color information for a setting field.
[0008]
[Means for Solving the Problems]
The multicolor character image reading device according to the present invention is a multicolor form reading format creation change in which a range is set for each designated color of character image input and a standard form is created in which the range position and color are grouped for each field. A form type determining means for comparing the standard form created by the multicolor form reading format creation / changing means and the character image input to determine the type of the input form, and the type determined by the form type determining means The character to be read from the image extraction means for extracting the target area from the character image input based on the above, the read area extracted by the image extraction means and the setting value of the corresponding field of the standard form of the type determined by the form type determination means And a character recognition means for performing character recognition processing on the character to be read for each area cut out by the character cutout means and recognizing the character.
[0009]
In addition to the basic configuration, color integration means is added to integrate pixels having similar color components with respect to character image input into a designated color.
[0010]
Further, a result output means for outputting the image extracted by the image extraction means and the recognition result of the character recognition means output is added to the basic configuration.
[0011]
Furthermore, the multi-color form reading format creation changing means extracts only the image corresponding to a specific color and outputs it, so that the range position and color setting can be changed.
[0012]
Furthermore, the form type determination means determines the form type based on the color information of each of the standard form and the character image input.
[0013]
Furthermore, the form type determining means determines the form type by associating data for each field for each color in addition to the color information of each of the standard form and the character image input.
[0014]
Further, the character cutout means cuts out the character image input using the relative position information between the field color information and the character image input.
[0015]
Furthermore, the image extraction means extracts a plurality of color information from the target area and outputs the extracted information.
[0016]
[Action]
In the multicolor character image reading device according to the present invention, the data of the range position and the color is collected for each field from the input character image information, and the standard form is stored, and the new character image input is compared with these standard forms. The type of the form is determined, a target area is extracted based on the determined standard form data, characters are further cut out from the standard form area, and character recognition is performed on the cut out characters.
[0017]
Furthermore, information having similar color components in the input character image information is collectively stored or compared as one color.
[0018]
Furthermore, the image of the reading area extracted from the character image input and the character recognition result are superimposed and output.
[0019]
Furthermore, an input character image or an image of a specific range of a standard form is extracted, and the setting of the range position and color can be freely changed.
[0020]
Furthermore, the color information of the standard form and each character image input is compared, and the form type is determined based on the comparison.
[0021]
Furthermore, in addition to the color information of the standard form and character image input, the data for each field for each color is compared to determine the form type.
[0022]
Furthermore, the character image input is cut out based on the relative position information between the field color information and the character image input.
[0023]
Furthermore, a plurality of pieces of color information are extracted from the target area of the character image information and are superimposed and output.
[0024]
【Example】
Example 1.
FIG. 1 is a block diagram showing the overall configuration of a multicolor character image reading apparatus according to an embodiment of the present invention. In the present embodiment, the configuration operation of the multi-color form reading format creation / changing means for determining the area range and its color for each partial area, which is referred to as a read form field, will be described.
In FIG. 1, reference numeral 1 denotes a multicolor form image obtained by inputting a multicolor form into an image input device. Reference numeral 2 denotes a memory that stores a multi-color form reading format group registered as a standard form by extracting an image from the multi-color form image and dividing the attributes into a number of fields as described later. 3 is a multi-color form reading format creation / change means for creating / changing data included in the multi-color form reading format, and 4 is a pixel having a color component similar to the input of the multi-color character image. Color integrating means for integrating and classifying the color component values of the multicolor form image into representative values, 5 includes information on reading areas and colors included in individual standard forms registered in the multicolor form reading format group, A form type determining means for comparing the image information integrated by the color integrating means and specifying the corresponding multi-color form reading format (standard form) to determine the form type, 6 is created by the color integrating means Image extracting means for extracting image information of a specific area defined in the multi-color form reading format from the obtained image information, and 7 cuts out characters from the image information extracted by the image extracting means. , Character segmentation means for outputting the image of the character region, 8 is a character recognition means for outputting character codes and color information recognizing an image obtained by said character segmentation means, corresponding. Reference numeral 9 denotes a result output means for editing and outputting the image output from the image extraction means and the character cutout means, the character code output from the character recognition means, and the color information.
[0025]
FIG. 2 is a detailed block diagram of the multicolor form reading format creation / change means. In other words, data such as form size, field frame, character frame, entry character, etc., such as the position, size, name, color, etc. in the form are obtained to recognize the character in the form or within a specific area. A form reading format (standard form) to be referred to extract an image is created or changed.
In the figure, a multi-color form image 1 is a form image input by an image input device. The multicolor image display control means 22 determines the display range in the form image as the input range of the multicolor form image 1. Usually, a decision is made to display the entire form image. The multicolor image display means 23 displays the range determined by the multicolor image display control means 22 in the form image 1. The image range designating unit 24 designates / selects a specific area or form component among the multicolor images displayed by the multicolor image display unit 23 using a coordinate input device such as a mouse.
The color information setting / changing means 25 for the designated range sets / changes color information for a specific area or form component in the multicolor form image 1 designated / selected by the image range designation means 24. The reading format structure setting means 26 is means for setting format data such as the number of fields, the field position, and the number of characters in the field to be referred to when recognizing characters in the multicolor form image 1 or extracting a specific area. . By this means, a reading format in which the color-related data is empty or has a predetermined value (arbitrary value) among specific data of the reading format (for example, the number of fields is 4) is created. The color information setting means 27 for the reading format sets the color information set / changed by the color information setting / changing means 25 for the designated range to the reading format created by the reading format structure setting means 26.
[0026]
FIG. 3 is an explanatory diagram showing an example of data in the multicolor form reading format. In the figure, 31 is an item for specifying a field, 32 is an item for designating an attribute element of a field frame color, 33 is a specified value for 32 attribute elements, 34 is an item for specifying a field, 35 is a character frame color in the field , 36 is a specified value for 35 attribute elements. 37 is the specified value of the field 1 character extraction condition, 38 is the character recognition target of the field 1, 39 is the background color specification value, 301 is the specified value of the field 2 character extraction condition (1), and 302 is the field 2 character The specified value of the clipping condition (2), 303 is the position specified value of the preprint (1) of field 2, 304 is the position specified value of the preprint (2) of field 2, and 305 is the specified value of the preprint color of field 2 , 306 is the designated value of the character recognition target in field 2, 307 is the designated value of the entry character color, 308 is the designated value of the image cropping condition of field 3, 309 is the designated value of the field frame color of field 3, and 310 is the field 3 311 is the specified value of the image cropping target color (1), 311 is the specified value of the image cropping target color (2), 312 is the multicolor form reading format, 31 Is the color information.
[0027]
FIG. 4 shows an example of a multicolor form image. In the figure, 40 is a preprinted character of “employee number”, 41 is also a preprint “date of employment”, 42 is a preprint “name (mark)”, 43 is a preprint “periphery of home”, 44 is An entry field frame corresponding to the employee number, 45 an entry field frame corresponding to the entry date, 46 an entry field frame for the name (mark), 47 a field frame for the home surroundings diagram, and 48 an in-field character for the employee number A frame 49 is an in-field preprint for the hire date.
[0028]
Next, a procedure for creating the multi-color form reading format 28 will be described with reference to FIG. Here, it is assumed that the multicolor image display control means 22 performs a normal operation and determines to display the entire multicolor image.
An example of a reading format created by the reading format structure setting means 26 is shown in FIG. The reading format is set in one line per item. For example, item 31 describes data specifying a field, “field 1”. The item 32 is a detailed item representing an attribute relating to the item 32, and describes that data relating to “field frame color” of “field 1” is set. The item 33 represents specific data for the item 31 and the item 32, and color names or color component values such as “black” and “red” are described, but the item 33 is not yet set in FIG.
[0029]
An example in which the multicolor form image 1 is read and displayed by the multicolor image display means 23 is shown in FIG. The operator selects or designates a part of the image using the image range designating unit 24 while viewing the multicolor image shown in FIG. For example, it is assumed that the form component shown in the field frame 45 in FIG. 4 is selected.
The color information acquisition unit 29 from the image acquires the color information of the image in the designated range designated by the image range designation unit 24. When the operator selects the field frame color setting value 33 portion of FIG. 3 while the field frame 45 of FIG. 4 is selected, the color information setting means 27 for the reading format is acquired by the color information acquisition means 29 from the image. The set color information is set to the field frame color setting value 33 in FIG.
[0030]
In the present embodiment, an example in which the color designation is one type has been described.
As described above, according to this embodiment, the operator can easily set the color information in the multicolor form reading format simply by selecting / specifying a specific area or a form component in the multicolor form image. It is possible to set a multi-color form reading format (standard form).
[0031]
Example 2
In the present embodiment, operations of the color integration unit and the form type determination unit in the overall configuration will be described.
FIG. 5 is an example of a multicolor form image, in which data is entered in the multicolor form image of FIG. In the figure, an area 50 surrounded by a dotted line is an area of an employee number entry column.
FIG. 6 is an enlarged view of the frame 50 shown in FIG. 5 for explaining the operation of the color integration means 4 and shows the distinction between colors. That is, this is an example of processing for the area 50 in the employee number entry column.
FIG. 7 is a diagram for explaining the operation of the form type determination unit 5. Color classification element data obtained by classifying element data according to color information described in a standard form for character image input different from that in FIG. It is. In the figure, 71 is color classification element data corresponding to the color A, 72 is color classification element data corresponding to the color B, and 73 is color classification element data corresponding to a color not registered in the multicolor form reading format.
FIG. 8 is a diagram for explaining the operation of the form type determination means 5 and explains the collation operation between the multicolor form reading format and the color classification element data. In the figure, reference numeral 81 denotes a multicolor form reading format.
FIG. 9 is a diagram of a character image different from FIG. 8 for explaining the operation of the form type determination means 5. In the figure, 90 is a character described in a multi-color form reading format that does not correspond to characters on the color classification element data, and 91 is described in a multi-color form reading format that does not correspond to a table area on the color classification element data. Table area 92 is a table area on the color classification element data that does not correspond to the table area described in the multi-color form reading format, and 93 corresponds to the character described in the multi-color form reading format but has a different color. The character on the color classification element data 94 is a multi-color form reading format different from 81.
[0032]
Next, the operation will be described.
The operations of the color integration unit 4 and the form type determination unit 5 will be described with reference to the multicolor form image shown in FIG.
When the color integration unit 4 is applied to the partial area 50 of the multicolor form image shown in FIG. 5, the color integration unit 4 examines the color component value for each pixel on the multicolor form image, and displays similar color components. A pixel having a value is classified into several representative values. This process can be realized, for example, by performing clustering based on color component values.
Next, the pixels classified into the same representative value are grouped by obtaining the continuity of the region. Thereby, pixel groups classified into the same representative value and connected on the image can be extracted. This process can be realized, for example, by performing a labeling process using the representative value of each pixel. Thereby, the pixel group indicated by 61 shown in FIG. 6 is extracted.
Finally, element data (position and shape for each element, color value of representative value) is calculated for the pixel group indicated by 61. This is stored as data of the corresponding field in FIG.
The actual processing is performed on the entire multicolor form image. That is, all fields of the input form are processed and data is stored.
[0033]
After the color integration unit 4 ends, the form type determination unit 5 is executed.
The form type determination means 5 determines the form type of the character image 1 of the input multicolor form based on the element data for each field extracted by the multicolor form reading format group 2 and the color integration means 4.
First, one standard form is selected from the multicolor form reading format group 2.
Next, element data for each field is classified according to the color information described in the selected standard form, and color classification element data is created.
Here, when N types of colors are described in the selected standard form, the element data is classified into N + 1 pieces (N types of colors + other colors). For example, the element data of the input character image is element A (RGB color component values are R100, G50, B70), element B (R100, G60, B70), element C (R50, G50, B50), element D (R20). , G20, B10), while color A (R100 ± 20, G50 ± 10, B70 ± 10) and color B (R50 ± 5, G50 ± 10, B50 ± 10) are described in the standard form , Elements A and B are classified as color A, element C is classified as color B, and element D is classified as an unregistered color in the standard form.
Specifically, the color classification element data shown in FIG. 7 can be created based on the input character image example shown in FIG.
[0034]
Next, characters, table regions, and ruled lines are extracted based on the created color classification element data.
That is, as shown in FIG. 8, the characters and table areas extracted as color classification element data are associated with the characters and table areas described in the standard form 81. Here, the correspondence is performed so as to be optimized by using information such as the position, size, color, character code, and table structure of characters and table areas drawn on the form.
Furthermore, the similarity between the standard form and the input multicolor form image 1 is determined based on the result of the association.
This similarity is obtained from the difference in position, size, color, character code, table structure, etc. when characters and table regions are associated, the remaining characters without correspondence, the number of table regions, etc. . Here, the difference in the structure of the table is obtained by a method of associating the ruled lines constituting the table described in the multicolor form reading format with the ruled lines extracted as the color classification element data. This can be realized by, for example, application No. 6-76795, “form processing apparatus including table”.
For example, FIG. 9 shows a character / table area association result of color classification element data for each field in a multi-color form reading format 94 different from that shown in FIG. Here, the character 90 and the table area 91 on the multi-color form reading format do not correspond to the known character and table area of the color classification element data. Further, the table area 92 extracted as the color classification element data cannot be associated with the table area on the standard form. Further, the character 93 extracted as the color classification element data corresponds to the character on the standard form, but the color is different.
Therefore, for example, the similarity between the other standard form 94 shown in FIG. 9 and the input multicolor form image 1 in FIG. 5 is lower than the similarity with the standard form 81 in FIG.
[0035]
Thereafter, the above process is performed for a plurality of standard forms stored in the multicolor form reading format group 2, and the similarity is calculated for each standard form. Assume that the standard form having the highest similarity is the type of the input multicolor form image (character image) 1.
However, even if the degree of similarity is the highest, if the degree of similarity is lower than a predetermined value, or if (the highest degree of similarity) − (second highest degree of similarity) is small, the input multicolor form image 1 Since the type determination is impossible, the input multicolor form image 1 is rejected.
If not rejected here, the created color classification element data, the number of the standard form with the highest degree of similarity, and the result of associating characters and table areas are output.
[0036]
In this embodiment, an example in which the form type determination is performed in units of pixels in each field has been described. However, it is also possible to perform form type determination in units of small areas composed of a plurality of pixels.
In this embodiment, an example is shown in which characters and table areas are associated with each other in order to determine the form type. However, it is possible to associate characters, photo areas other than the table areas, and the like at the same time.
As described above, according to the present embodiment, the type of a form can be determined even in reading with a mixture of a plurality of types of forms, and accurate even for a multicolor form image having poor image quality in which uneven printing of the form has occurred. The form type can be determined.
[0037]
Example 3 FIG.
In the present embodiment, an operation of adding an additional mechanism to the multi-color form reading format creation changing means to extract only a specific color image or changing the color information will be described.
FIG. 10 is a diagram showing operation buttons for creating / changing the multi-color form reading format. In the figure, 11 is a color range reduction button, and 12 is a color range enlargement button.
[0038]
A procedure for changing the color information of the reading format when the color information is set in the standard form and a problem occurs in image extraction / character recognition because the set color information is not accurate will be described.
For example, it is assumed that the multi-color form reading format is determined as shown in FIG. 3, the color information 33 is already determined as “color B”, and the color information 36 is already determined as “color C”.
Next, the form to be read is input to the image input device to obtain a multicolor form image 1. Only the standard form storing the information shown in FIG. 3 is stored in the multi-color form reading format group 2 and this image data is input to the color integrating means 4 and then the form type determining means 5 is operated. In this way, color classification element data for each field in which the correspondence between the multicolor form image 1 and the standard form is obtained. This color classification element data is input to the multicolor image display control means 22.
The multi-color image display control means 22 has a color information correction mode instead of a normal mode for making a decision to display the entire image, and displays only color classification element data whose color information value is “color C”. Control.
Under the control of the multicolor image display control means 22, only the color classification element data whose color information value is “color C” is displayed by the multicolor image display means 23. According to the reading format shown in FIG. 3, the form component whose color information value is “color C” is the “character frame color” of “field 1”. Only 48 should be displayed.
[0039]
The operator checks on the screen whether or not only the character frame 48 of FIG. 4 is displayed. If the character frame 48 is not displayed, a form component or a specific area other than the character frame 48 is displayed. In this case, the mode of the multicolor image display control means 22 is changed so that the entire multicolor form image is displayed.
For example, it is assumed that not only the character frame 48 but also the field frame 44 is displayed because “color C” and “color B” are similar. The operator changes the color range of the color C by the color information setting / changing means 25 for the designated range. In this case, pressing the color range reduction button 11 to reduce the color range of the color C can prevent confusion with the color B.
If the character frame 49 is not displayed, the color is adjusted so that the character frame 49 can be extracted by depressing the color range expansion button 12 and expanding the color C color range.
[0040]
In this example, the example of setting the color range by reduction / enlargement was shown, but direct input of numerical values, adjustment of individual color attributes such as red, blue, green, etc., only for the maximum value and only the minimum value It is also possible to make this change.
As described above, according to this embodiment, it is possible to easily check whether the color information set in the multicolor form reading format actually works effectively, and to easily correct even if there is a problem.
[0041]
Example 4
In this embodiment, the operation of the character cutout means will be described.
FIG. 11 is a flowchart for explaining the operation of the character cutout means 7 when there is a character frame.
FIG. 12 is a diagram showing image information corresponding to the field frame color. In the figure, 131 is the same as the image 531 shown in FIG.
FIG. 13 is a diagram showing image information corresponding to the character frame color. In the figure, 132 is the same as the image 532 shown in FIG.
FIG. 14 is a diagram showing image information corresponding to colors that are not registered in the standard form 133. In the drawing, reference numeral 133 is the same as the image 533 shown in FIG.
FIG. 15 is a diagram for explaining the operation of character frame extraction for image information corresponding to the character frame color. In the figure, reference numeral 151 denotes a character frame cutout position.
FIG. 16 is a diagram showing the relationship between the character recognition target image information and the character frame cutout position. In the figure, reference numeral 151 denotes the same image as that shown in FIG.
FIG. 17 is a diagram showing a result of character cutout.
[0042]
Next, the operation of the character cutout means when there is a character frame will be described.
Based on the flowchart of FIG. 11, the operation on the image area of FIG. 35 will be described with reference to the image information examples of FIGS. Further, it is assumed that the standard form is determined as shown in FIG. 3, the color information 33 is already determined as “color B”, and the color information 36 is already determined as “color C”.
In addition, when recognizing an input character image, the position of each region is relatively recognized.
[0043]
Character segmentation is performed on the image information corresponding to the area obtained from the form type determination means 5. The image information corresponding to the area obtained from the form type determination means 5 is classified by the color registered in the standard form. The image shown in FIG. 12 for color B, the image shown in FIG. 13 for color C, and the image shown in FIG. 14 for colors not registered in the standard form are color integration means 4 and form type determination means. 5 is obtained.
First, in step S111, a character cut-out target image is extracted. In this case, as shown in FIG. 3, the character frame conformity 37 is selected as the character cutout condition for the standard form. Furthermore, the character frame color is set to color C36 by the standard form. Therefore, the color C is selected as the character cutout target image, and image information for the color C shown in FIG. 13 is extracted.
[0044]
In step S112, a character frame cutout process is performed on the image of FIG. In character frame cutout, an image histogram is extracted in the vertical direction, and candidate points are set at equal pitches based on the character frame pitch stored in the multicolor form reading format. Further, this candidate cutout point is shifted to the left and right, and a candidate point sequence that minimizes the sum of the histogram values on the candidate cutout point is set as a character frame cutout position. Thereby, the character frame cutout position 151 shown in FIG. 15 is obtained with respect to FIG.
[0045]
In step S113, a character recognition target image is extracted. In this case, 38 (other than the field frame color, other than the background color, and other than the character frame color) 38 is selected as the character recognition target in the multicolor form reading format. The background color is set to color F39, the field frame color is set to color B33, and the character frame color is set to color C36. Therefore, a color other than color F, color B, and color C is selected as a character recognition target. Thereby, the image of FIG. 14 is obtained as a character recognition target image.
[0046]
In step S114, since the relative position of each image is previously recognized on the character recognition target image obtained in step S113, the character frame cutout position obtained in step S112 is overlapped to cut out the character. FIG. 16 shows the relationship between the character recognition target image and the character cutout position, and FIG. 17 shows the character cutout result.
In character recognition, the character recognition means 8 recognizes individual cutout results of the character cutout results of FIG. Since this is an extremely common means, detailed description of the configuration and operation will be omitted.
[0047]
In this embodiment, the shape of the character frame and the field frame is a solid rectangle, but other shapes such as a circle and a rhombus, a broken line, a wavy line, and the like may be used.
Further, in the present embodiment, an example in which the character frame color, the field frame color, and the character recognition target color designation are one type is shown, but the present invention is not limited to this, and a plurality of types of colors can be designated.
In the present embodiment, an example of finally recognizing a character has been described. However, the present invention is not limited to this, and a character cutout image can also be output.
As described above, according to this embodiment, since the relative position is determined by obtaining color information and position information for each field, the characters entered in the multicolor form are not entered in the center of the character frame, and the characters are close to each other. Even when it is entered or when the form image is shifted and input, it is possible to cut out the character accurately, and thus the character can be recognized accurately.
[0048]
Example 5 FIG.
In this embodiment, the operation of the character cutout means will be described.
FIG. 18 is a flowchart for explaining the operation of the character cutout means 7 when there is a preprint.
FIG. 19 is a diagram illustrating an example of a form image corresponding to the area obtained from the form type determination unit 5. In the figure, 191 is a field frame, 192 is a preprint image (1), 193 is a preprint image (2), and 194 is an entry character.
FIG. 20 is a diagram showing image information corresponding to the field frame color in the region shown in FIG. In the figure, 191 is the same image as that of FIG.
FIG. 21 is a diagram showing image information corresponding to the preprint colors in the area shown in FIG. In the figure, reference numerals 192 and 193 denote the same year and month preprints as those in FIG.
FIG. 22 is a diagram showing image information corresponding to the entered character color in the area shown in FIG. In the figure, 194 is the same as that of FIG.
[0049]
FIG. 23 is an image diagram of a frame and the like for explaining the processing contents of character entry area extraction. In the figure, reference numeral 231 denotes a character entry area (1) for the preprint (1), and 232 denotes a character entry area (2) for the preprint (2).
FIG. 24 is an image diagram for explaining the operation of character string segmentation. In the figure, reference numeral 194 denotes entry characters, and reference numerals 231 and 232 denote frames shown in FIG.
FIG. 25 is a diagram illustrating a result of character string segmentation. In the figure, 251 is the character string cutout result corresponding to preprint (1), 252 is the character string cutout result corresponding to preprint (2), and 194 is the entry characters shown in FIG.
FIG. 26 shows the result of character segmentation. In the figure, 261 is the result of character cutout for the first character corresponding to preprint (1), 262 is the result of character cutout for the second character corresponding to preprint (1), and 263 is corresponding to preprint (2). The character cutout result 264 for the first character is a character cutout result for the second character corresponding to preprint (2).
[0050]
Next, the operation of the character cutout means when there is a preprint will be described.
The operation corresponding to the image input in FIG. 19 will be described based on the flowchart in FIG. 18 with reference to the image information examples in FIGS.
Character segmentation is performed on the image information corresponding to the area obtained from the form type determination means 5. The image information corresponding to the area obtained from the form type determination means 5 is classified by the color registered in the standard form. The image shown in FIG. 20 is obtained for the color B, the image shown in FIG. 21 is obtained for the color D, and the image shown in FIG.
First, in step S181, a preprint cutout target image is extracted. In this case, in the multi-color form reading format shown in FIG. 3, preprint character (1) compliant 301 is selected as the character cutout condition (1), and preprint character (2) compliant 302 is selected as the character cutout condition (2). Yes. Further, the color D305 is set as the preprint color together with the entry coordinates 303 of the preprint character (1) and the entry coordinates 304 of the preprint character (2). Therefore, the color D is selected as the preprint cutout target image, and image information for the color D shown in FIG. 21 is extracted.
[0051]
In step S182, preprint cutout processing is performed on the image information shown in FIG. In the preprint cutout, an image histogram is extracted in the vertical direction, and cut out using the preprint position stored in the multicolor form reading format.
In this case, there are two preprint characters, and the fact that the preprint (1) entry coordinates 303 are to the left of the preprint (2) entry coordinates 304, the preprint image (1) 192 in FIG. Preprint image (2) 193 corresponds to preprint (2) to preprint (1).
Next, the character entry area is extracted using the image information for the field frame color of FIG. 20 and the image information of FIG. 23 in which the image for the preprint color of FIG. 21 is superimposed. The area to the left of the preprint image (1) 192 associated with the preprint (1) is extracted as the character entry area (1) 191 and the preprint image (2) 193 associated with the preprint (2) is extracted. The area on the left side of the preprint image (1) 192 associated with the preprint (1) is extracted as the character entry area (2) 232. The upper and lower ends of the character entry area are obtained based on the upper and lower end positions of the field frame 191.
[0052]
In step S183, a character recognition target image is extracted. In this case, the entry character color 306 is selected as the character recognition target in the multicolor form reading format. Further, the color E307 is set as the entry character color. Therefore, a color corresponding to the color E is selected as the character recognition target image. Thereby, the image of FIG. 22 is obtained as the character recognition target image.
In step S184, as shown in FIG. 24, the character entry area obtained in step S182 is superimposed on the character recognition target image obtained in step S183, and the character string is cut out. The character string is divided into a character string cutout result 251 corresponding to the preprint (1) shown in FIG. 25 and a character string cutout result 252 corresponding to the preprint (2).
[0053]
In step S185, each character is cut out from the character string cutout result obtained in step S184. In character cutout, an image histogram is extracted in the vertical direction, and a cutout position in the horizontal direction is obtained by cutting out a block of the histogram. Next, a block image encircled by left and right cutout coordinates is scanned from the top to the bottom and the point that first hits the black pixel is the upper end, and the point that scans from the bottom to the top and hits the black pixel first is the bottom. As a result of character cutout, as shown in FIG. 26, the character cutout result 261 of the first character corresponding to preprint (1), the character cutout result 262 of the second character corresponding to preprint (1), preprint ( The character cutout result 263 of the first character corresponding to 2) and the character cutout result 264 of the second character corresponding to preprint (2) are divided.
[0054]
In character recognition, the character recognition means 8 recognizes individual cutout results shown in FIG. At this time, the recognition results of the cutout results 261 and 262 are recognition results corresponding to preprint (1), that is, “year”. Similarly, the recognition results of the cutout results 263 and 264 are recognition results corresponding to preprint (2), that is, “month”.
[0055]
In this embodiment, an example in which preprints are written in characters is shown, but the present invention is not limited to this, and symbols, pictures, and the like may be used.
In the present embodiment, an example of one type of preprint color designation is shown. However, the present invention is not limited to this, and a plurality of types of colors can be designated by designating a preprint color for each preprint. Can do.
In this embodiment, the preprint information and the preprint image stored in the form reading format are associated with each other by the position information. However, the present invention is not limited to this, and correspondence can be made by using color information, shape information, and the like.
In this embodiment, extraction of the character entry area is extracted from the adjacent position of the preprint image. However, the present invention is not limited to this, and a position away from the preprint image by a certain amount can be freely set.
As described above, according to the present embodiment, characters entered in a multi-color form with preprints can be accurately cut out in association with the preprints, so that the characters can be accurately recognized.
[0056]
Example 6
In this embodiment, the operation of the image extraction means will be described.
FIG. 27 is a flowchart for explaining the operation of the image extracting means 6.
FIG. 28 is a diagram illustrating an example of a form image corresponding to the area obtained from the form type determination unit 5. In the figure, 281 is a field frame, 282 is an entry character, 283 is an imprint written in a color that does not correspond to any of the entry character color, field frame color, and background color.
FIG. 29 shows image information corresponding to the field frame color in the area shown in FIG. In the figure, 281 is the same image as the frame in FIG.
FIG. 30 is a diagram showing image information corresponding to the entered character color in the area shown in FIG. In the figure, 282 is the same image as the entered characters in FIG.
FIG. 31 is a diagram showing image information corresponding to colors not registered in the multicolor form reading format in the area shown in FIG. In the figure, 283 is the same as the seal in FIG.
FIG. 32 is an explanatory image diagram for explaining the superimposition of image information. In the figure, 282 and 283 are the same as the corresponding images in FIG.
[0057]
Based on the flowchart of FIG. 27, the operation of the image extracting means 6 will be described with reference to specific image examples of FIGS. The multi-color form reading format shown in FIG. 3 is applied.
Image extraction is performed on the image information corresponding to the area obtained from the form type determination means 5. The image information corresponding to the area obtained from the form type determination means 5 is classified by the color registered in the multicolor form reading format. For example, the image shown in FIG. 29 for the color B, the image shown in FIG. 30 for the color E, and the image shown in FIG. 31 for the color not registered in the multicolor form reading format are the color integration means 4. And the form type determination means 5.
[0058]
First, in step S271, an image composed of the colors of the constituent elements specified by the image clipping conditions stored in the multicolor form reading format is extracted. In this case, in the multi-color form reading format of FIG. 3, field frame compliant 308 is set as the image cut-out condition, and color B 309 is set as the field frame color, so image information for color B shown in FIG. 29 is extracted.
In step S272, the image information included in the image information of FIG. 29 is extracted including the field frame itself.
In step S273, a written character image is extracted. In this case, the entry character color 310 is selected as the image cutout target color (1) in the multicolor form reading format. Further, a color E307 is set as the entry character color. Therefore, the color corresponding to the color E is extracted from the included image information. The extracted image information is the same image information as FIG.
In step S273, an image other than the entered characters is extracted. In this case, 311 (other than the entry character color, other than the background color, and other than the field frame color) is selected as the image clipping target color (2) in the multicolor form reading format. Further, color E307 is set as the entry character color, color F39 is set as the background color, and color B309 is set as the field frame color. Therefore, colors corresponding to (other than color E, color F, and color B) are extracted from the included image. The extracted image information is the same image information as FIG.
[0059]
In step S275, the extracted images obtained in steps S273 and S274 are superimposed. Specifically, the extraction of the image obtained by superimposing the extracted image obtained in step S273 and the extracted image obtained in step S274 is extracted according to the instruction of the multicolor form reading format.
In the multi-color form reading format, since two image cutout target colors shown in 310 and 312 are selected, the image of FIG. 32 in which FIGS. 30 and 31 are superimposed is output.
In the present embodiment, an example in which two image cutout target colors are set is shown, but the present invention is not limited to this, and three or more types and all colors may be set.
[0060]
In the present embodiment, the field frame is an example of a rectangle, but the present invention is not limited to this, and other shapes such as a circle and a rhombus can be set.
In the present embodiment, an example has been shown for the entry character color and the color that is not registered in the multicolor form reading format, but other elements can also be extracted by specifying the character frame color, field frame color, etc. it can.
In this embodiment, the image extraction means is applied only once to one area. However, a plurality of fields having the same conditions except for the color to be cut out of the image are defined, and a plurality of fields are defined for the same area. By applying the time image extracting means, it is also possible to extract each of the image of only the seal imprint and the image of only the written character color.
As described above, according to the present embodiment, it is possible to freely extract components distinguished by color from an image set with a frame in a multicolor form, and to efficiently extract a necessary image. it can.
[0061]
Example 7
In this embodiment, the operation of the result output means will be described.
FIG. 33 is a flowchart for explaining the operation of the result output means 9.
Next, the operation of the result output means 9 will be described with reference to FIGS.
First, in step S 331, an image acquired by the image extraction unit 6 and the character cutout unit 7 and a character recognition result obtained by the character recognition unit 8 are input to the result output unit 9.
In step S332, it is determined whether or not to display the character recognition result and the acquired image. Whether to display or not is specified by the multi-color form reading format.
In step S333, one of display based on the multicolor form reading format and display based on the image color is selected. The display color selection is specified by the multi-color form reading format.
If display based on the multicolor form reading format is selected here, the character recognition result and the acquired image are displayed on the display device in the character color described in the multicolor form reading format and the color of the acquired image in S334. To do.
If display based on the image color is selected, the character recognition result and the acquired image are displayed using the color of the character obtained from the multicolor form image 1 and the color of the acquired image in S335. For example, in the display based on the image color, when a part of the character string written in the multicolor form image 1 is written in red and the remaining characters are written in black, the character recognition result corresponding to the red character part is displayed in red. Thus, the character recognition result corresponding to the black character portion is displayed in black.
Thereafter, in S336, the operator performs a correction operation for misread characters on the displayed character recognition result or image, and the processed result is converted into a data file.
In S337, the created data file is stored in the storage device, and the process ends.
On the contrary, when the character recognition result and the acquired image are not displayed, the acquired image and the character recognition result are directly converted into a data file, stored in the storage device, and the process is terminated.
[0062]
In this embodiment, an example in which the presence / absence of display and the selection of display color are specified by the multi-color form reading format is shown, but the operator may specify interactively.
In the present embodiment, the result output means 9 is configured to output the image color or the color specified in the multi-color form reading format. However, when the image color and the color described in the color form reading format are different, the color is different. It is good also as a structure which displays an area | region on a screen.
As described above, according to the present embodiment, it is possible to display the input image and the same color as the multi-color form reading format, and it is possible to smoothly perform the correction work without error.
[0063]
【The invention's effect】
As described above, according to the present invention, the multi-color form reading format creation / changing means for setting or changing the standard form, the form type determining means for determining the form type of the character image input, and the character based on the determined type Image extracting means for extracting the target area from the image input, character cutout means for cutting out the target character with the setting value of the corresponding field of the standard form of the type determined as the extracted reading area, and character recognition means for the read target character Therefore, there is an effect that character images can be recognized with accurate cutout positions even for character image inputs having a plurality of types of fields with different colors.
[0064]
Furthermore, since a color integration unit that integrates similar color components into a designated color is added, an effect of reducing the types of colors to be recognized and reducing the comparison time is added.
[0065]
Furthermore, since the extracted image and the result output means for outputting the character recognition result are added, the effect of facilitating confirmation and correction is added.
[0066]
Furthermore, since only a specific color image is extracted and output, and the setting of the range position and color can be changed, the operation of the apparatus can be confirmed, and the operation can be easily corrected.
[0067]
Furthermore, since the form type is determined based on the standard multi-color form and the color information of the character image input, there is an effect that the detailed form can be distinguished.
[0068]
Furthermore, since the form type is determined by associating the data for each field for each color, it is possible to distinguish even forms with similar colors, and there is an effect that information on fields of other arbitrary standard forms can be used. .
[0069]
Further, since the character image input is cut out using the relative position information between the color information of the field and the character image input, there is an effect of preventing erroneous recognition by reading the characters together with the position of the entry frame.
[0070]
Furthermore, since a plurality of pieces of color information are extracted from the target area and superimposed and output, necessary images can be extracted efficiently, and character image information can be recognized accurately.
[Brief description of the drawings]
FIG. 1 is a configuration diagram of a multicolor character image reading apparatus according to an embodiment of the present invention.
FIG. 2 is a configuration block diagram of a multi-color form reading format creation / change unit shown in the first and third embodiments.
FIG. 3 is a diagram illustrating an example of standard form data stored in a multicolor form reading format group memory;
FIG. 4 is a diagram illustrating an example of a multicolor form image obtained by reading an input character image.
FIG. 5 is a diagram illustrating an example of a multi-color form image in the second embodiment.
FIG. 6 is an enlarged frame diagram for explaining the operation of the color integration unit 4 in the second embodiment.
7 is a diagram for explaining an operation of a form type determination unit 5 in Embodiment 2. FIG.
FIG. 8 is a diagram for explaining a collation operation of the form type determination unit 5 in the second embodiment.
FIG. 9 is another image diagram for explaining the operation of the form type determination unit 5 in the second embodiment.
FIG. 10 is a diagram illustrating operation buttons for changing color information according to the third embodiment.
FIG. 11 is a flowchart showing the processing operation of the character cutout means 7 according to the fourth embodiment.
FIG. 12 is a diagram illustrating color classification element data classified into field frame colors according to the fourth embodiment.
FIG. 13 is a diagram illustrating color classification element data classified into character frame colors according to the fourth embodiment.
14 is a diagram showing color classification element data classified into colors not registered in a standard form in Example 4. FIG.
FIG. 15 is a diagram illustrating a character frame cutout result according to the fourth embodiment.
FIG. 16 is a diagram illustrating a relationship between a character cutout target and a character cutout position according to the fourth embodiment.
FIG. 17 is a diagram illustrating an example of character cutout results according to the fourth embodiment.
FIG. 18 is a flowchart showing the processing operation of the character cutout means 7 according to the fifth embodiment.
FIG. 19 is a diagram illustrating an example of a multicolor form image according to the fifth embodiment.
FIG. 20 is a diagram illustrating color classification element data classified into field frame colors according to the fifth embodiment.
FIG. 21 is a diagram illustrating color classification element data classified into preprint colors according to the fifth embodiment.
22 is a diagram showing color classification element data classified into entry character colors in Example 5. FIG.
FIG. 23 is an image diagram of a frame and the like for explaining the operation of extracting a character string area in the fifth embodiment.
FIG. 24 is an image diagram for explaining a character string cut-out operation according to the fifth embodiment.
FIG. 25 is a diagram illustrating a character string cutout result according to the fifth embodiment.
FIG. 26 is a diagram showing a result of character segmentation in Example 5.
27 is a flowchart showing the processing operation of the image extracting means in Embodiment 6. FIG.
FIG. 28 is a diagram illustrating an example of a multicolor form image according to the sixth embodiment.
FIG. 29 is a diagram illustrating color classification element data classified into field frame colors according to the sixth embodiment.
30 is a diagram showing color classification element data classified into entry character colors in Example 6. FIG.
FIG. 31 is a diagram showing color classification element data classified into colors not registered in a standard form in Example 6.
FIG. 32 is a diagram illustrating an image in which an entry character color and a non-registered color are superimposed on a standard form in Example 6.
FIG. 33 is a flowchart showing the processing operation of the result output means in the seventh embodiment.
FIG. 34 is a configuration block diagram of a conventional multicolor form processing apparatus.
FIG. 35 is a diagram illustrating an example of a multicolor form image having a character frame.
FIG. 36 is a diagram for explaining a character cut-out operation by a conventional reading device.
FIG. 37 is a diagram showing a conventional character cutout result for the input shown in FIG. 35.
FIG. 38 is a diagram illustrating an example of a multicolor form image having a preprint.
FIG. 39 is a diagram showing a conventional character cutout result for the input shown in FIG. 38;
[Explanation of symbols]
1 Multicolor form image 2 Multicolor form reading format (standard form) group memory 3 Multicolor form reading format creation / change means 4 Color integration means 5 Form type determination means 6 Image extraction means 7 Character extraction means 8 character recognition means 9 result output means 24 image range designation means 25 color information setting / changing means 26 reading format structure setting means 27 color information setting means for reading format

Claims (8)

A multi-color form reading format creation change means for creating a standard form in which a range is set for each designated color of character image input and the range position and color are grouped for each field;
A standard form created by the multicolor form reading format creation change means, and a form type determination means for comparing the character image input to determine the type of the input form;
Image extracting means for extracting a target area from the character image input based on the type determined by the form type determining means;
A character cutout means for cutting out a read target character from the reading area extracted by the image extraction means and the setting value of the corresponding field of the standard form of the type determined by the form type determination means;
A multicolor character image reading apparatus comprising character recognition means for recognizing a character to be read for each region cut out by the character cutout means. 2. The multicolor character image reading apparatus according to claim 1, further comprising color integration means for integrating pixels having similar color components with respect to character image input into a designated color. 2. The multicolor character image reading apparatus according to claim 1, further comprising a result output means for outputting the image extracted by the image extraction means and the recognition result of the character recognition means output to the basic configuration. Furthermore, the multi-color form reading format creation changing means extracts and outputs only an image corresponding to a specific color so that the range position and color setting can be changed. Multicolor character image reader. 2. The multicolor character image reading apparatus according to claim 1, wherein the form type determining means determines the form type based on the color information of each of the standard form and the character image input. Further, the form type determining means determines the form type by associating data for each color field in addition to the color information of each of the standard form and character image input. Color character image reader. 2. The multi-color character image reading apparatus according to claim 1, wherein the character cutout means cuts out the character image input by using the relative color information of the field color information and the character image input. 2. The multicolor character image reading apparatus according to claim 1, wherein the image extracting unit extracts a plurality of pieces of color information from the target region and outputs the extracted information.

Images