JP2010258627A - Image processor, image processing method, program, and storage medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To extract watermark information embedded by changing and using blank length between characters even from a document with a postscript line written therein by extracting a circumscribed rectangle of a character even from the document with the postscript line written therein. <P>SOLUTION: A document image with embedding information embedded therein is acquired by controlling a distance between characters. A first object composed of adjacent black pixel groups is detected within an area between characters in the document image. The line width of the first object is calculated as a parameter value for the first object. When the parameter value for the first object is equal to or less than a threshold, after the first object is regarded as an area composed of white pixel groups, each circumscribed rectangle of a character constituting the document image is calculated, and the embedding information embedded in the document image is extracted. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a technique for improving the reading accuracy of a digital watermark.

  As a technique for secretly embedding information in a printed document, for example, as shown in Patent Document 1, there is a digital watermark using a space length between characters in a document (hereinafter referred to as “character watermark digital watermark”). Inter-character digital watermarking is a technique for embedding information so that humans do not notice easily by changing the length of a space between characters. For this purpose, a character circumscribed rectangle extraction technique, which is an element technique of the character recognition technique, is used. Here, the term “inter-character digital watermark” includes both an electronic watermark embedded by changing the blank length between lines and an electronic watermark embedded by changing the blank length between characters.

JP 2002-232679 A

  When handwriting is added to a printed document in which an inter-character digital watermark is embedded, it is conceivable to add a line including an underline and a correction line. When a line is added, if no measures are taken, a result different from the circumscribed rectangle extraction result at the time of embedding occurs, and the embedded information may not be extracted reliably. For example, in the sample showing an example of the additional line as shown in FIG. 5A, the additional line penetrates the character. In such a case, it is difficult to extract the embedded watermark information by reliably cutting out each character.

  The present invention makes it possible to extract a character circumscribing rectangle from a document in which a writing line is written, and also from a document in which a writing line is written, by embedding the watermark information embedded by changing the blank length between characters. The purpose of this is to facilitate the extraction and to improve the extraction rate.

In order to achieve the object of the present invention, for example, an image processing apparatus of the present invention comprises the following arrangement. That is,
Means for acquiring a document image in which the embedded information is embedded in accordance with an embedding method for embedding the embedded information by controlling the distance between characters according to the embedded information;
Means for specifying a blank line area in the document image when the direction of arrangement of characters written in the document image is the row direction and the direction orthogonal to the row direction is the column direction;
Means for specifying an area other than the blank line area as a character string area in the document image;
Means for specifying an inter-character area in the character string area;
When a pixel having the same pixel value as a pixel constituting the character part in the document image is a black pixel, and a pixel other than the black pixel is a white pixel, the document image is composed of adjacent black pixels in the inter-character area. Means for detecting a first object that is present;
Means for determining a line width of the first object as a parameter value for the first object;
Calculating means for obtaining a projection for each of a row direction and a column direction of the document image, and obtaining a circumscribed rectangle of each character part constituting the document image based on the obtained projection;
Means for extracting embedded information embedded in the document image based on the distance between each circumscribed rectangle;
If the parameter value for the first object is less than or equal to a first threshold value, the calculation means regards the first object as an area composed of white pixel groups, and then calculates the document image Obtain a projection in the line direction, obtain a line according to the density obtained as a result, and then obtain a circumscribed rectangle of each character part constituting the document image by obtaining a projection in a direction perpendicular to the line for each line. And

In order to achieve the object of the present invention, for example, an image processing apparatus of the present invention comprises the following arrangement. That is,
An image processing method performed by an image processing apparatus,
Acquiring a document image in which the embedded information is embedded in accordance with an embedding method for embedding the embedded information by controlling the distance between characters according to the embedded information;
Specifying a blank line area in the document image when the line direction is a line direction of characters written in the document image and the column direction is a direction orthogonal to the line direction;
Identifying an area other than the blank line area as a character string area in the document image;
Identifying an inter-character area in the character string area;
When a pixel having the same pixel value as a pixel constituting the character part in the document image is a black pixel, and a pixel other than the black pixel is a white pixel, the document image is composed of adjacent black pixels in the inter-character area. Detecting a first object that is present;
Obtaining a line width of the first object as a parameter value for the first object;
A calculation step for obtaining a projection for each of the row direction and the column direction of the document image, and obtaining a circumscribed rectangle of each character part constituting the document image based on the obtained projection;
Extracting embedded information embedded in the document image based on a distance between each circumscribed rectangle,
In the calculation step, when the parameter value for the first object is equal to or less than a first threshold value, the first object is regarded as an area composed of a group of white pixels, and then the document image Obtaining a projection in the line direction, obtaining a line according to the density obtained as a result, and then obtaining a projection in a direction perpendicular to the line for each line to obtain a circumscribed rectangle of each character part constituting the document image And

  According to the present invention, a character circumscribed rectangle can be more reliably extracted from a document in which a write-once line is written. Therefore, it is possible to easily extract the watermark information embedded by changing the blank length between characters from the document in which the additional writing line is written, and contribute to the improvement of the extraction rate.

It is a block diagram which shows the structure of the electronic watermark extraction apparatus between characters. It is an operation | movement flowchart of the electronic watermark extraction apparatus between characters. It is a figure which shows the detail of (a) additional writing line deletion part 0102, (b) the identification part 0303. FIG. It is an operation | movement flowchart of the write-once line deletion part of step S0202. It is a figure explaining a write-once line and a write-once line erasing method. It is a figure explaining additional writing line erasure. It is a flowchart of the line-to-line additional line deletion process of step S0405. It is a flowchart of the additional writing line deletion process between characters of step S0408. It is a figure explaining the estimation of a character size, and a connection black pixel area | region. It is a block diagram which shows the electric constitution of a computer.

  Embodiments of the present invention will be described below with reference to the drawings. However, the scope of the invention is not limited to the following examples. In the following embodiments, “between lines” indicates between character string lines, “between characters” indicates between two characters in the character string line, and “between characters” indicates both between the lines and between characters. In addition, the arrangement direction of characters is a row direction, and a direction orthogonal to the row direction is a column direction.

[Example 1]
Hereinafter, an image processing apparatus that extracts a digital watermark between characters after deleting a line added between characters will be described. Hereinafter, a line added to a document is referred to as a postscript line. FIG. 1 shows a digital watermark extraction apparatus which is an image processing apparatus according to the present embodiment. Hereinafter, each unit of the digital watermark extracting apparatus will be described.

  The input unit 0101 acquires a watermark embedded image as an input image. The additional line erasing unit 0102 removes a part that hinders the extraction of the inter-character digital watermark, that is, the additional line or a part of the additional line, from the additional line attached to the printed matter. The character rectangle extraction unit 0103 extracts a rectangle circumscribing the character and outputs the position coordinates of the circumscribed rectangle. The digital watermark extraction unit 0104 calculates a blank length between circumscribed rectangles, that is, a distance between characters, from the position coordinates acquired from the character rectangle extraction unit 0103, and extracts an embedded digital watermark.

  The processing of this digital watermark extracting apparatus will be described with reference to FIG. First, the watermark embedded image is input (step S0201). Then, the write-on line or the faulty part of the write-on line is deleted (step S0202). Next, character rectangle extraction (step S0203) is performed, and finally digital watermark extraction (step S0204) is performed.

  FIG. 3A is a diagram showing details of the write-once line erasing unit 0102 shown in FIG. The input unit 0101 acquires an input image (an image with embedded inter-character digital watermark). The input image is a binary image composed of black pixels and white pixels. In this embodiment, the character part in the input image is composed of black pixels, and pixels other than the black pixels are white pixels. The input unit 0101 provides an input image that is a binary image to a projection unit 0302 and a connected pixel region extraction unit 0304 described later. The input unit 0101 obtains an input image that is not a binary image, for example, a grayscale or full-color input image, and then binarizes the image, and then outputs the binarized image to the projection unit 0302 and the connected pixel region extraction unit 0304. May be provided.

  The projection unit 0302 detects a low-density area where the black pixel density in the document image is equal to or less than a threshold (low-density area detection unit). Specifically, the input image is projected in the row direction (horizontal for horizontal writing and vertical for vertical writing), and rows with black pixels greater than or equal to the threshold are displayed in high-density rows. A row in which the number of black pixels is equal to or less than the threshold value may be determined as a low density row. Here, projecting means totaling the number of black pixels on each line for each line in a pixel unit in the horizontal direction. (In the vertical direction, the number of black pixels on each column is counted for each column in pixel units.) And, depending on the density obtained as a result, which rows are high-density rows, Information indicating which rows are low-density rows is stored in the storage unit 0308 as row information. Further, the inter-row low density partial information indicating the position of the low density row region which is a continuous low density row is provided to the identification unit 0303. The low density line area corresponds to a blank line area in the document image. The projecting unit 0302 uses the row information stored in the storage unit 0308 to acquire continuous high-density rows as a high-density row region. The high-density line area is a part other than the blank line area in the document image, and corresponds to a character string area. Further, for each of the high-density row regions, that is, for each row of the character string, projection is performed in a direction orthogonal to each row. The part is determined as a low-density column. Then, low-density inter-character density information indicating the position of a low-density inter-character area that is a continuous low-density sequence is provided to the identification unit 0303. The low-density partial information is information that represents the position of a portion where the density is lower than a predetermined threshold based on the result of projection. That is, the inter-line low density partial information indicates a blank position between lines, and the inter-character low density partial information indicates a blank position between characters.

  The identification unit 0303 determines whether or not a black pixel in the area indicated by the low-density part information between lines or the low-density part information between characters, that is, a black pixel in the low-density area is a part of the additional line or the additional line. For the determination, a parameter value obtained from the connected black pixel group in the low density region is used. Then, the coordinates of the pixels constituting the additional writing line or a part of the additional recording line are provided as line information to the comparison unit 0305 and the erasing unit 0307.

  FIG. 3B is a diagram illustrating details of the identification unit 0303. The identification unit 0303 includes an integrated value calculation unit 1601, a line width calculation unit 1602, and a condition determination unit 1603. The integrated value calculation unit 1601 obtains a parameter value corresponding to the approximate length of a line composed of a group of connected black pixels in the low density row area indicated by the inter-row low density partial information. In addition, a parameter value corresponding to the approximate length of a line made up of connected black pixel groups in the low density inter-character area indicated by the inter-character low density partial information is obtained. This parameter value is called the length of the connected black pixel group. The connected black pixel group is a set of two or more black pixels connected to each other vertically, horizontally, or diagonally. However, it is possible to consider that two pixels that are adjacent obliquely, that is, corners and corners are basically not connected. In some cases it may be desirable to assume that the resolution is not connected when the resolution is low. (If the resolution is lowered, the original resolution may not be actually connected.) The line width calculation unit 1602 sets the parameter value corresponding to the approximate width of the line made up of connected black pixels. Ask for. This parameter value is called the width or line width of the connected black pixel group. The condition determination unit 1603 determines whether the obtained length and line width satisfy a predetermined condition, and outputs line information regarding a connected black pixel group that satisfies the condition. The line information is information indicating the coordinates in the input image of the black pixels included in the connected black pixel group. Specific processing of the identification unit 0303 will be described later.

  The connected pixel area extraction unit 0304 extracts a connected black pixel group (third object) in the document image (connected black pixel group detection means). The comparison unit 0305 acquires line information from the identification unit 0303, character size information from the character size estimation unit 0306, and connected black pixel group information from the connected pixel region extraction unit 0304. Then, it is determined by comparing the size of the connected black pixel group and the character size information whether the line indicated by the line information is an added line, that is, an additional line or a part of the additional line. As a result, when it is determined that it is a write-once line or a part of the write-online, the line information acquired from the identifying unit 0303 is provided to the erasing unit 0307 as write-once line information. For example, the line in the portion where the density shown in FIG. 9E is determined to be low is a write-on line or a part of the write-on line, and information representing these is the write-on line information. Specific processing will be described later.

  The character size estimation unit 0306 estimates the character size using the connected black pixel group information acquired from the connected pixel region extraction unit 0304. In estimating the character size, the character size is estimated for each block divided by a predetermined method. For example, the input image may be divided into half in the vertical direction and the horizontal direction to form four blocks, and the character size may be estimated for each block. The number of blocks may be larger, and may be divided for each character object by using a block selection technique. Then, character size information including information on the character size estimated for each block is generated and provided to the comparison unit 0305.

  FIG. 9A is a diagram illustrating a method of integrating character rectangles overlapping connected black pixel groups. FIGS. 9B and 9C are diagrams showing histograms used for character size estimation. There are the following methods for estimating the character size. First, when the circumscribed rectangles overlap each other for a plurality of connected black pixel groups, they are integrated (see FIG. 9A). Then, for the circumscribed rectangle after the integration process in the block, a histogram is taken for each area, width, or height (see FIGS. 9B and 9C), and a value corresponding to the mode is taken. Thus, the size of the character is estimated. As a value corresponding to the mode value, a peak value extracted from the result of convolution integration of a normal distribution pattern and a histogram graph may be used. This is because it is not realistic that the mode value is set to one for the character size. In the case of a value corresponding to the mode value regarding the area, a value obtained by taking the square root thereof is set as the estimated character size, and in the case of the width, the value corresponding to the mode value is set as the estimated character size. The unit of the estimated character size is pixel.

  As an alternative character size estimation method, there is a method of using a table indicating a relationship between a character size obtained from a general document and a histogram pattern. In this case, the pattern in the table closest to the histogram for the connected black pixel group after performing the integration processing in the block may be determined, and the corresponding character size may be used as the estimated character size. When half-width characters and full-width characters are mixed, the histogram is considered to have bimodality. Corresponding to this case, the character size can also be estimated by detecting the case where the histogram for the connected black pixel group after the integration processing in the block partially includes the pattern in the table. It is. In the case of Japanese, full-width characters can be regarded as almost square, but in the case of half-width characters and alphanumeric characters, the difference between width and height is large. The length may be estimated, and each result may be treated as a character size.

  The erasing unit 0307 corrects the input image acquired from the input unit 0101 by converting the black pixel indicated by the additional line information acquired from the comparison unit 0305 into a white pixel. A storage unit 0308 is used as a buffer, and a corrected image in the middle of correction is input / output to / from the storage unit 0308. Further, the corrected image is also stored in the storage unit 0308.

  The storage unit 0308 inputs / outputs line information to / from the projecting unit 0302 and holds the line information. Further, a corrected image is input / output to / from the erasing unit 0307, and the corrected image is held. The output unit 0309 acquires a corrected image that has been all corrected from the storage unit 0308, and outputs it as an output image.

  Next, the details of the additional line erasing process in step S0202, that is, the process of the additional line erasing unit 0102 will be described with reference to FIG. First, the connected pixel area extracting unit 0304 extracts a connected black pixel group from the input image (step S0401). Then, the character size estimation unit 0306 performs character size estimation (step S0402) based on the extraction result. Next, the projecting unit 0302 performs horizontal projection on the input image (step S0403), and specifies a low-density row area (rectangular area) (step S0404). Then, the identification unit 0303 specifies a write-on line in the low-density row area, and the erase unit 0307 erases the write-on line (step S0405). Details of step S0405 will be described later.

  Next, the projecting unit 0302 performs projection in the direction orthogonal to the row for each of the high-density row regions (step S0406), and specifies the low-density character region (rectangular region) (step S0407). These two processes are performed by the projection unit 0302 in FIG. Then, the identification unit 0303 specifies a part of the additional line and the additional line in the low-density character space area (an additional line specifying unit), and the erasing unit 0307 erases the additional line and a part of the additional line (step S0408). . Details of step S0405 will also be described later. Finally, the output unit 0309 outputs the corrected image (step S0409).

  Next, details of the process of erasing the line-to-line additional line or a part of the additional line in step S0405 will be described with reference to FIG. First, in step S1000-1, the identification unit 0303 selects one of the row-to-row low density regions. For example, the selection may be made in order from the first line to the last line. In step 1000-2, the identification unit 0303 performs a connected black pixel group extraction process on the low-interline density area selected in step S1000-1. That is, a black pixel group (second object) in the blank row area is detected. The extracted connected black pixel group is referred to as an inter-row connected black pixel group. The connected black pixel group extraction process in step S1000-2 may be performed in the same manner as in step S0401. However, it should be noted that the process in step S0401 is performed on the entire input image, whereas the process in step S1000-2 is performed on the low-density area between lines. The identification unit 0303 provides the coordinates of the extracted black pixels constituting the extracted connected black pixel group to the comparison unit 0305 as line information. In step S1000-3, the integrated value calculation unit 1601 selects one of the connected black pixel groups extracted in step S1000-2.

  In steps S1001 to S1006, the length of the connected black pixel group selected in step S1000-3 is obtained. The processing from step S1001 to step S1006 is an example of processing for obtaining a length. The integrated value calculation unit 1601 performs the processing from step S1001 to step S1006.

  First, the variable i is initialized with 1, and the variable L is initialized with 0 (step S1001). In the case of a horizontally written document, i indicates the y coordinate where the upper end of the connected black pixel group is 1. Subsequently, one of the black runs in the i-th row of the connected black pixel group is selected (step S1002). A black run indicates a group of black pixels continuous in the row direction. There may be a case where a plurality of black runs exist in one row. In this case, the black runs may be selected one by one in order, for example, from the left.

  In steps S1003a to S1003e, a parameter value called an integrated value is assigned to the black run. The integrated value is calculated as follows. First, it is examined whether or not the black run in the (i-1) th row is adjacent to the black run selected in step S1002 (step S1003a). In step S1003a, it is determined that the black runs are adjacent to each other even if they are in contact with each other at an angle. When the black run of the (i-1) th row exists adjacently, the process proceeds to step S1003b, and when it does not exist adjacently, the process proceeds to step S1003c. When the process proceeds to step S1003b, the integration of the black run adjacent to the (i-1) th row is added to the distance between the center of gravity of the black run selected at step S1002 and the barycenter of the black run adjacent to the (i-1) th row. The sum of the values is taken as the integrated value of the black run selected in step S1002. If two or more black runs in the i-1th row are in contact with the black run selected in step S1002, an integrated value corresponding to each black run in the i-1th row is obtained and the total value is obtained. The integrated value of black run A selected in step S1002 is used. For example, consider a case where the black runs A and i-1 are in contact with the black runs A and i. In this case, black is added to the distance between the center of gravity of black run A and the center of gravity of black run B plus the integrated value of black run B and the distance between the center of gravity of black run A and the center of black run C. The sum of the integrated value of run C is added to obtain the integrated value of black run A. When the process proceeds to step S1003c, the integrated value of the black run selected in step S1002 is set to 1.

  In step S1003d, it is checked whether the black run selected in step S1002 is adjacent to the black run in the (i + 1) th row (step S1003d). Even in step S1003d, it is determined that the black runs are adjacent to each other even if they are in contact with each other at an angle. If there is an adjacent black run on the (i + 1) th row, the process proceeds to step S1004. If the black run in the (i + 1) th row does not exist adjacently, the process proceeds to step S1003e, and the integrated value of the black run selected in step S1002 is added to the variable L, and then the process proceeds to step S1004.

  In step S1004, it is determined whether a black run that has not been selected in step S1002 still remains in the i-th row. If it still remains, the process returns to step S1002 to select the next black run. If no more remains, the process proceeds to step S1005. Next, i is incremented (step S1005), and it is determined whether i is equal to or smaller than the number of pixel rows in the connected black pixel group selected in step S1000-3 (step S1006). If i is less than or equal to the number of pixel rows in the connected black pixel group, the process returns to step S1003, and if i is not less than or equal to the number of pixel rows in the connected black pixel group, the process proceeds to step S1007. The value of the variable L when proceeding to step S1007 is a parameter value corresponding to the length of the connected black pixel group.

  In step S1007, the line width calculation unit 1602 obtains a parameter value (line width) corresponding to the width of the connected black pixel group selected in step S1000-3. The line width is the number obtained by dividing the number of black pixels in the connected black pixel group by the length of the connected black pixel group obtained in steps S1001 to S1006. Here, a table as shown in FIG. 9D is obtained. FIG. 9D is a diagram showing a table used in line segment integrated value calculation. Based on the diagram shown in FIG. 5B, the black run integrated value of each pixel row and each pixel row The number of black pixels is calculated. The total number of black pixels in each pixel row matches the number of black pixels included in the line. FIG. 5 (b) is a diagram showing a method for erasing additional lines between rows.

  Instead of steps S1001 to S1007, other methods for obtaining the line width can be employed. For example, as a simpler method, there is a method of estimating using the circumscribed rectangle of the connected black pixel group selected in step S1000-3. The width in the column direction of the circumscribed rectangle of the connected black pixel group selected in step S1000-3 is used as a parameter value corresponding to the length of the connected black pixel group. A value obtained by dividing the number of black pixels of the connected black pixel group selected in step S1000-3 by the width in the column direction of the circumscribed rectangle may be used as the line width of the connected black pixel group.

  In step S1008, the condition determination unit 1603 checks whether or not the width and length of the connected black pixel group selected in step S1000-3 match the conditions. Here, the conditions are that the width of the connected black pixel group is equal to or smaller than the line width (second threshold value) of characters in the input image, and that the length of the connected black pixel group is equal to or greater than a predetermined value (fourth). 2) or more. If the condition is met, it is determined that the connected black pixel group selected in step S1000-3 is the additional writing line to be erased or a part of the additional recording line. Then, the position information of the black pixels constituting the connected black pixel group is provided to the erasing unit 0307 as additional writing line information, and the process proceeds to step S1009. If the condition is not met, it is determined that the connected black pixel group selected in step S1000-3 is not an erasure target, and the process proceeds to step S1010-1.

  By setting the width of the connected black pixel group to be equal to or smaller than the line width of the character in the input image, the erasure condition is that only a part of the additional writing line or a part of the additional writing line is smaller than the line width of the character in the input image. Erase and prevent accidental erasure of characters and character parts that are printed from the beginning. If the length of the connected black pixel group is too short, it can be considered that the input image has a thin portion at the end of the character. However, as shown in FIG. 5D, it is preferable that an erasure target can be achieved even when the connected black pixel group ends in the middle of the low-density row region. For this reason, when the length of the connected black pixel group is equal to or smaller than a predetermined value, it is not considered as an erasure target. The line width of the character in the input image can be determined by known detection means or based on the character size estimated by the character size estimation unit 0306. However, whether or not the group of connected black pixels is to be erased depends on the custom of the printed document that differs depending on the country, language, culture, business customs, etc., and is not limited to the conditions shown here. It suffices if at least the width of the connected black pixel group is equal to or smaller than the threshold (second threshold or less). 5 (c) and 5 (d) are diagrams showing a method for erasing additional lines between characters.

  In step S <b> 1009, the erasing unit 0307 acquires the position information of the black pixel to be erased, which is indicated by the additional line information acquired from the comparison unit 0305. An input image is acquired from the input unit 0101. Then, the pixel value of the black pixel at the position indicated by the additional line information in the input image is manipulated and converted to a white pixel.

  In step S1010-1, the identification unit 0303 determines whether all the connected black pixel groups extracted in step S1000-2 have been selected. If all of them are selected, the process proceeds to step S1010-2. If there are any unselected connected black pixel groups, the process returns to step S1000-3 to select an unselected connected black pixel group. In step S1010-2, the identification unit 0303 determines whether all the low-density regions between rows have been selected. If all are selected, the process of step S0405 is terminated. If an unselected low-line area between lines remains, the process returns to step S1000-1 to select the next low-line area between lines. By these processes, as shown in FIGS. 6A and 6B, the write-once line or a part of the write-online between the rows is erased. FIGS. 6A and 6B are diagrams showing an example of line segment erasing between rows. FIG. 6A shows an example before processing, and FIG. 6B shows an example after processing.

  FIG. 8 is a flowchart of the inter-character writing line erasing process in step S0408. Basically, the processing in step S0408 is substantially the same as that in step S0405. However, the process in step S0405 aims to erase the additional lines between lines and a part of the additional lines, whereas the process in step S0408 erases the additional lines and parts of the additional lines between characters. There is a difference.

  First, in step S1100-1, the identification unit 0303 selects one of the low density inter-character areas. In step 1100-2, the identification unit 0303 performs a connected black pixel group extraction process on the low-density character space selected in step S1100-1 in the same manner as in step S1000-2. The extracted connected black pixel group is called an inter-character connected black pixel group (first object). The identification unit 0303 provides the coordinates of the extracted black pixels constituting the extracted connected black pixel group to the comparison unit 0305 as line information. In step S1100-3, the integrated value calculation unit 1601 selects one of the connected black pixel groups extracted in step S1100-2.

  In steps S1101 to S1106, the length of the connected black pixel group selected in step S1100-3 is obtained. The processing from step S1101 to step S1106 is an example of processing for obtaining the length. The integrated value calculation unit 1601 performs the processing from step S1101 to step S1106.

  First, the variable i is initialized with 1, and the variable L is initialized with 0 (step S1101). In the case of a horizontally written document, i indicates an x coordinate where the left end of the connected black pixel group is 1. Subsequently, one of the black runs in the i-th column of the connected black pixel group is selected (step S1102). A black run indicates a group of black pixels continuous in the column direction. There may be a case where a plurality of black runs exist in one row. In this case, black runs may be selected one by one in order, for example, from the top.

  In steps S1103a to S1103e, a parameter value called an integrated value is assigned to the black run. The integrated value is calculated as follows. First, it is checked whether or not the black run selected in step S1102 is adjacent to the black run in the (i-1) th column (step S1103a). In step S1103a, it is determined that the black runs are adjacent to each other even if they are in contact with each other at an angle. If the black run in the (i-1) th column exists adjacently, the process proceeds to step S1103b. If the black run does not exist adjacently, the process proceeds to step S1103c. When the process proceeds to step S1103b, the integration of the black run adjacent to the (i-1) th column is added to the distance between the center of gravity of the black run selected at step S1102 and the barycenter of the black run adjacent to the (i-1) th column. The sum of the values is taken as the integrated value of the black run selected in step S1102. When two or more black runs in the (i-1) th column are in contact with the black run selected in step S1102, an integrated value corresponding to each black run in the (i-1) th column is obtained and the total value is obtained. The integrated value of the black run A selected in step S1102 is used. For example, consider a case where black runs A and i-1 are in contact with black runs A and i. In this case, black is added to the distance between the center of gravity of black run A and the center of gravity of black run B plus the integrated value of black run B and the distance between the center of gravity of black run A and the center of black run C. The sum of the integrated value of run C is added to obtain the integrated value of black run A. When the process proceeds to step S1103c, the integrated value of the black run selected in step S1102 is set to 1.

  Next, in step S1103d, it is checked whether or not the black run selected in step S1102 is adjacent to the black run in the (i + 1) th column (step S1103d). Even in step S1103d, it is determined that the black runs are adjacent to each other even if they are in contact with each other at an angle. If there are adjacent black runs in the (i + 1) th column, the process advances to step S1104. If the black run in the (i + 1) th column does not exist adjacently, the process proceeds to step S1103e, and the integrated value of the black run selected in step S1102 is added to the variable L, and then the process proceeds to step S1104.

  In step S1104, it is determined whether there is still a black run that has not been selected in step S1102 in the i-th column. If it remains, the process returns to step S1102 to select the next black run. If there is no more left, the process proceeds to step S1105. Next, i is incremented (step S1105), and it is determined whether i is equal to or smaller than the number of pixel columns in the connected black pixel group selected in step S1100-3 (step S1106). If i is less than or equal to the number of pixel columns in the connected black pixel group, the process returns to step S1103. If i is not less than or equal to the number of pixel columns in the connected black pixel group, the process proceeds to step S1107. The value of the variable L when proceeding to step S1107 indicates a parameter value corresponding to the length of the connected black pixel group.

  In step S1107, the line width calculation unit 1602 obtains a parameter value (line width) corresponding to the width of the connected black pixel group selected in step S1100-3. The line width is a number obtained by dividing the number of black pixels in the connected black pixel group by the length of the connected black pixel group obtained in steps S1101 to S1106.

  As in the case of the low-density row region, another method for obtaining the line width can be employed instead of steps S1101 to S1107. For example, there is a method of estimating using the circumscribed rectangle of the connected black pixel group selected in step S1100-3. The width in the row direction of the circumscribed rectangle of the connected black pixel group selected in step S1100-3 is used as a parameter value corresponding to the length of the connected black pixel group. Then, a value obtained by dividing the number of black pixels of the connected black pixel group selected in step S1100-3 by the width in the row direction of the circumscribed rectangle may be used as a parameter value corresponding to the width of the connected black pixel group.

  In step S1108, the condition determination unit 1603 checks whether or not the width and length of the connected black pixel group selected in step S1100-3 match the conditions. Here, the conditions are that the width of the connected black pixel group is equal to or smaller than the line width (first threshold) of the characters in the input image, and that the length of the connected black pixel group is equal to or greater than a predetermined value (third 2) or more. If the condition is met, the condition determining unit 1603 sets the connected black pixel group selected in step S1100-3 as a candidate for additional writing line to be erased (step S1109). If the condition is not met, it is determined that the connected black pixel group selected in step S1100-3 is not an erasure target. However, as described above, the condition is not limited to this, and it is only necessary that at least the width of the connected black pixel group is equal to or smaller than the threshold value (less than the first threshold value).

  Next, the condition determination unit 1603 determines whether or not the width of the connected black pixel group including the candidate for the additional line or a part of the additional line is equal to or greater than the calculated value based on the estimated character size, which is different for each region ( Step S1110). At this time, the connected black pixel group including the candidate for the additional line or a part of the additional line is not the connected black pixel group extracted in step S1100-2 but the connected black pixel group detected in step S0401. is there.

  FIG. 9E shows the width of the connected black pixel region. In the case of FIG. 9 (e), the connected black pixel group including the candidate for the additional writing line or the portion of the additional writing line of “part determined to be low in density” is the entire black pixel of FIG. 9 (e). Note that it is not just the black pixels that exist in the “part where the density is judged to be low”. A line in a portion (between characters) determined to be low in density cannot be denied the possibility of being a character portion simply based on the density and line width. However, the possibility of being a character part can be excluded by examining the width of the connected black pixel group to which the candidate for the additional line or a part of the additional line belongs in step S1110. For example, when characters are connected by additional writing lines, the connected black pixel group becomes larger, and the width of the connected black pixel group to which the candidate for the additional writing line or a part of the additional writing line belongs becomes larger than the estimated character size. On the other hand, if the part of the character that is not added with the additional line is a candidate for the additional line or part of the additional line, the connected black pixel group to which the candidate for the additional line or part of the additional line belongs is one character. It is considered to be about the estimated character size. Therefore, by comparing the width of the connected black pixel group that includes the candidate for the additional line or a part of the additional line with the calculated value based on the estimated character size, the connected black that includes the candidate for the additional line or a part of the additional line. It can be confirmed whether or not the pixel group is a portion of a character not provided with an additional writing line.

  Specifically, a value obtained by multiplying the estimated character size estimated by the character size estimation unit 0306 by a predetermined value is used as a calculated value. For example, if the estimated character size is 60 pixels and the predetermined value is 2, 60 × 2 = 120 pixels is a calculated value based on the estimated character size. Since the estimated character size is calculated for each block, the calculated value may naturally be a different value for each block.

  In step S1110, if the width of the connected black pixel group to which the candidate of the additional line or a part of the additional line belongs is equal to or greater than the calculated value, the condition determining unit 1603 determines that the candidate of the additional line or the additional line is the additional line or the additional line. Judge as part. Then, the additional line or candidate line information of the additional line is provided to the erasing unit 0307 as additional line or additional line information. Subsequently, in step S <b> 1111, the erasing unit 0307 acquires the position information of the black pixel to be erased, which is indicated by the additional line information acquired from the comparison unit 0305. An input image is acquired from the input unit 0101. Then, the pixel value of the black pixel at the position indicated by the additional line information in the input image is manipulated to convert it to a white pixel, and the process proceeds to step S1112-1. In step S1110, if the width of the connected black pixel group to which the additional line candidate belongs is not equal to or greater than the calculated value, it is determined that the additional line candidate is not the additional line or a part of the additional line, and the process proceeds to step S1112-1.

  In step S1112-1, the identification unit 0303 determines whether all the connected black pixel groups extracted in step S1100-2 have been selected. If all have been selected, the process proceeds to step S1112-2, and if there remains a non-selected connected black pixel group, the process returns to step S1100-3 to select an unselected connected black pixel group. In step S <b> 1112-2, the identification unit 0303 determines whether all the low density inter-character areas have been selected. If all are selected, the process of step S0408 ends. If there is still a low-density character space area that has not yet been selected, the process returns to step S1100-1 to select the next low-density character space area. By these processes, as shown in FIGS. 6C and 6D, the additional writing line between characters and a part of the additional writing line are deleted. 6C and 6D are diagrams showing an example of line segment erasing between characters. FIG. 6C is an example before processing, and FIG. 6D is an example after processing.

  FIG. 6E is a diagram showing a relationship between line segment erasure between characters and a connected pixel region. FIG. 6E shows an example of the relationship between a line in a low density portion obtained by vertical projection and a connected black pixel group to which the line belongs. The additional lines and part of the additional lines, which are also the connected black pixel areas shown in this figure, are deleted, and the information between characters is appropriately extracted.

  With the above processing, it is possible to erase a write-once line and a part of the write-online, which is a pre-process for improving the inter-character digital watermark detection performance. The method of the present embodiment is particularly effective when two or more additional write lines or a part of the additional write line is drawn without crossing between characters or lines. Judging from the result of the projection alone, there is a high possibility that the postscript line is judged to be a character part, but by distinguishing the thickness of the line, it is possible to more accurately distinguish the character and the postscript line. it can. The threshold for detecting the low density region is preferably higher than the threshold for extracting the character rectangle after erasing the postscript line or part of the postscript line. By increasing the threshold value, it is possible to set a character or a postscript line, or a portion where it is difficult to determine whether or not to be a low-density region as a thickness determination target. Since the write-once line that is difficult to determine can be erased by the process of deleting the write-online, if the threshold value is lowered when extracting the character rectangle, it is possible to recognize a character that is difficult to determine while excluding the write-online. Although it is common to embed a digital watermark using an error correction code, the processing of this embodiment does not conflict with the use of this error correction code technology, but rather functions in a mutually complementary manner. To do.

  In the present embodiment, the method for improving the reading accuracy of the inter-character digital watermark by erasing the additional lines attached to the blank portion between the characters has been described. However, in order to read a digital watermark, it is often sufficient to delete only the additional line or part of the additional line provided between characters. In this case, even if the processing of step S0404 and step S0405 related to the processing between lines is omitted, it is possible to delete the additional writing line between characters or a part of the additional writing line and improve the reading accuracy of the inter-character digital watermark.

  Further, although it is preferable that there is a process of step S1110 to determine whether or not the width of the connected black pixel group including the additional writing line candidate is equal to or larger than a certain value, it is possible to delete the additional writing line that is sufficiently narrower than the line width of the character. It can be omitted if desired. In this case, the process of step S1109 and step S1110 can be omitted. Further, the connected pixel region extraction unit 0304 and the character size estimation unit 0306 can be omitted. On the contrary, in this embodiment, the process of determining whether or not the width of the connected black pixel group including the additional line candidate is greater than a certain value is performed only for the additional line candidate between characters. You may perform similarly when detecting a part.

  In the present embodiment, an apparatus has been described in which an image obtained by erasing an additional line or a part of an additional line attached to a blank portion between characters from an input image and reading an electronic watermark from the image is described. However, in view of the purpose of improving the reading accuracy of the intercharacter digital watermark, it is not always necessary to output an image from which the additional writing line has been deleted. That is, the coordinates of a pixel determined to be a write-on line or a part of the write-on line are stored. When detecting the space length between characters, that is, when detecting a circumscribed rectangle, it is sufficient to take a configuration in which a pixel group determined to be a write-on line or a part of the write-on line is regarded as a white pixel group. It is. In this configuration, the output unit 0309 is not necessary, the additional line information provided by the comparison unit 0305 is stored in the storage unit 0308, and the additional line information stored in the storage unit 0308 is used in the character rectangle extraction process in step S0203. That's fine.

  Also, in this embodiment, it is assumed that an image consisting only of a text area is input, but as a preceding stage, an area dividing means for dividing the image into areas having attributes such as an image, a table, a graph, and text. Can also be used. Naturally, it is possible to input only the text area as an image to the system having the configuration shown in the present embodiment after using the area dividing means. In addition, a substantially horizontal line between lines can be removed by looking at the height of the line, and a substantially vertical line between characters can be distinguished by comparing the width with other character widths. Therefore, it goes without saying that the object of the line to be removed in the above-described embodiment can be removed by using the parameters (width and height) of the circumscribed rectangle of the connecting element, if necessary.

In the above-described embodiment, the following calculation procedures (1 to 7) can be cited as a method for calculating the line width of a character. The more accurate it is, the more computational complexity is expected. Which procedure is used depends on the application target and the non-functional requirements (extraction speed, etc.) of the actually operated system.
1. A raster scan (or rotation in units of 90 degrees) of the circumscribed rectangle of the character is performed to obtain a black run, a histogram thereof is obtained, and the length of the frequent black run is set as the character width.
2. The character size and character density are calculated, and an approximate value is calculated from the table. (The result may be used in connection with the character size estimation unit 0306.)
3. Calculate character recognition, character size and character density, and calculate approximate values from the table.
4). After thinning, the number of pixels is counted and (number of character pixels before thinning / number of pixels) is calculated.
5). After thinning, the distance between pixels is measured and (number of character pixels before thinning / total distance between pixels) is calculated.
6). After thinning, a second-order Spline curve is formed, and distance calculation is performed to calculate (number of character pixels before thinning or character black area / character skeleton length).
7). After thinning, a cubic Spline curve is formed, and distance calculation is performed to calculate (number of character pixels before thinning or character black area / character skeleton length).
8). After character vectorization, calculate the length of the outline of the character (result is a), calculate the area of the character (result is b), and (b / a) × 2 is the line width of the character Calculate as

・ 2. And 3. About The table used here is a table for determining the corresponding character line width from the character density, size, or character recognition result. This table may be created in advance by a human or statistically created using a machine learning method. Further, the part of the table may be a function created by machine learning or by a human. In addition, the character line width is obtained by dividing the area of the character by the length of the character skeleton representing the character as a line having no width. In creating the table, it is also possible to use the calculation of the character line width described later. Such a method is effective when a character font used in advance is known and limited, and particularly when both speeding up and reliability of the result are required.

In addition, as a method for calculating the line width of a write-on line or a part of the write-on line, the following procedure (1-5) may be used instead of the method described above. (The upper one is an approximation, and the lower it is, the more precise it is, but the more it will be, the more computation is required.) Also for this, which procedure is actually used depends on the application target and the non-functional requirements of the system actually operated Speed).
1. After thinning, the number of pixels is counted and (number of character pixels before thinning / number of pixels) is calculated.
2. After thinning, measure the distance between pixels and calculate (number of character pixels before thinning / total distance between pixels).
3. After thinning, a second-order Spline curve is formed and the length of the line is calculated to calculate (number of character pixels before thinning or area of black area of candidate line / length of thinned line).
4). After thinning, a third-order Spline curve is formed, and the length of the line is calculated to calculate (number of character pixels before thinning or area of black area of candidate line / length of thinned line).
5). After vectorization, the length of the outline is calculated (result is a), the area of the line is calculated (result is b), and (b / a) × 2 is calculated as the line width.

  The thinning process used above is described in “Basic Techniques of Image Processing” <Introduction to Techniques> Hasegawa, Usui, Nakayama, and Yokoi Co. A thinning algorithm may be used. Alternatively, Hildith's thinning method described in “Introduction to Image Processing in C Language” (Shokodo) may be used. Also, some other thinning process may be used.

[Example 2]
In the first embodiment, each unit illustrated in FIG. 1 is implemented by hardware, but each unit other than the storage unit 0308 may be implemented by software. In this case, the software is stored in various storage devices included in the computer, and the computer implements the functions of each unit illustrated in FIG. 1 by the CPU executing the software. The storage unit 0308 can be realized using a storage device such as a main storage device or a hard disk.

  FIG. 10 is a block diagram showing a hardware configuration of a computer applicable to this embodiment. The computer 1801 is a general-purpose information processing apparatus such as a personal computer that is generally used. Inside the computer 1801, blocks to be described later are connected by a bus 1807, and various data can be transferred.

Depending on the device to which the computer 1801 is applied, all the components shown in FIG. 10 are not essential, and may be omitted as appropriate. Further, hardware having the same type of function may be replaced, or the computer 1801 may be configured by a plurality of computer bodies.
In FIG. 10, a CPU 1802 controls the entire computer 1801 using programs and data loaded in the main storage device 1803 and executes the above-described processes performed by the image processing apparatus to which the computer 1801 is applied. To do.

  Reference numeral 1803 denotes a main memory represented by RAM. The main storage device 1803 has an area for temporarily storing programs and data read from various storage devices. The storage device includes an HDD (hard disk drive) 1804, a CD drive 1809, a DVD drive 1810, an FDD (floppy (registered trademark) disk drive) 1811, and the like. Further, the main storage device 1803 also has an area for temporarily storing image data acquired from the scanner 1817 via the I / F (interface) 215. Further, the main storage device 1803 also has a work area that is used when the CPU 1802 executes various processes. Including the above, the main storage device 1803 can provide various information recording locations as appropriate.

  An HDD 1804 holds an OS (operating system) and various images (including document images). Further, the HDD 1804 also stores programs and data for causing the CPU 1802 to control the functions of the units illustrated in FIG. 10 or for causing the CPU 1802 to execute the above-described processes performed by an apparatus to which the computer 1801 is applied. Programs and data stored in the HDD 1804 are appropriately loaded into the main storage device 1803 according to control by the CPU 1802 and are processed by the CPU 1802. Note that some of the information described as being stored in the main storage device 1803 may be stored in the HDD 1804.

  Reference numeral 1805 denotes a video controller which sends display data such as image data and character data received from the main storage device 1803 and the HDD 1804 to the monitor 1806 as a signal. A monitor 1806 includes a CRT, a liquid crystal screen, and the like, and displays images, characters, and the like based on signals received from the video controller 1805. Reference numeral 1808 denotes an I / F for connecting the printer 1816 to the computer 1801. The computer 1801 can transmit print data to the printer 1816 via the I / F 1808 and receive status information of the printer 1816 transmitted by the printer 1816.

  A CD drive 1809 reads programs and data recorded on a CD as a storage medium, and sends the read programs and data to the HDD 1804, the main storage device 1803, and the like. A DVD drive 1810 reads programs and data recorded on a DVD as a storage medium, and sends the read programs and data to the HDD 1804 and the main storage device 1803. Reference numeral 1811 denotes an FDD, which reads out programs and data recorded on a floppy (registered trademark) disk as a storage medium, and sends the read programs and data to the HDD 1804, the main storage device 1803, and the like.

  Reference numerals 1813 and 1814 denote a mouse and a keyboard as operation input devices, respectively. A user of the computer 1801 can input various instructions to the CPU 1802 by operating the mouse 1813 and the keyboard 1814. Reference numeral 1812 denotes an I / F for connecting the keyboard 1814 and the mouse 1813 to the bus. An operation instruction input by the user from the mouse 1813 or the keyboard 1814 is sent as a signal to the CPU 1802 via the I / F 1812.

  Reference numeral 1815 is for connecting a scanner 1817 that generates image data by reading a document, a film, or the like to the computer 1801. Image data generated by the scanner 1817 is sent to the HDD 1804, the main storage device 1803, and the like via the I / F 1815.

  Reference numeral 1818 denotes an I / F for exchanging information with an electronic device such as another computer. Information including image data acquired from the network 1819 according to an instruction from the CPU 1802 is sent to the HDD 1804, the main storage device 1803, and the like via the I / F 1818.

[Other Embodiments]
Needless to say, the object of the present invention can also be achieved as follows. That is, a recording medium (or storage medium) that records a computer program code that implements the functions of the above-described embodiments is supplied to a system or apparatus. Then, the computer (or CPU or MPU) of the system or apparatus reads and executes the program code stored in the recording medium. In this case, the program code read from the recording medium itself realizes the functions of the above-described embodiment, and the program code and the recording medium on which the program code is recorded constitute the present invention.

  Further, when the computer program code read by the computer is executed, an operating system (OS) or the like running on the computer may perform part or all of the actual processing based on the instruction of the program code. is there. Needless to say, the present invention includes the case where the functions of the above-described embodiments are realized by the processing.

  Further, it is assumed that the computer program code read from the recording medium is written in a memory provided in a function expansion card inserted into the computer or a function expansion unit connected to the computer. Then, based on the instruction of the program code, the CPU or the like provided in the function expansion card or function expansion unit performs part or all of the actual processing, and the functions of the above-described embodiments are realized by the processing. Needless to say, it is included in the invention.

Claims (9)

Means for acquiring a document image in which the embedded information is embedded in accordance with an embedding method for embedding the embedded information by controlling the distance between characters according to the embedded information;
Means for specifying a blank line area in the document image when the direction of arrangement of characters written in the document image is the row direction and the direction orthogonal to the row direction is the column direction;
Means for specifying an area other than the blank line area in the document image as a character string area; means for specifying an inter-character area in the character string area;
When a pixel having the same pixel value as a pixel constituting the character part in the document image is a black pixel, and a pixel other than the black pixel is a white pixel, the document image is composed of adjacent black pixels in the inter-character area. Means for detecting a first object that is present;
Means for determining a line width of the first object as a parameter value for the first object;
Calculating means for obtaining a projection for each of a row direction and a column direction of the document image, and obtaining a circumscribed rectangle of each character part constituting the document image based on the obtained projection;
Means for extracting embedded information embedded in the document image based on the distance between each circumscribed rectangle;
If the parameter value for the first object is less than or equal to a first threshold value, the calculation means regards the first object as an area composed of white pixel groups, and then calculates the document image Obtaining a projection in the line direction, obtaining a line according to a density obtained as a result, and then obtaining a projection in a direction orthogonal to the line for each line to obtain a circumscribed rectangle of each character part constituting the document image An image processing apparatus. Means for detecting a second object composed of a group of adjacent black pixels in the blank row region;
Means for determining a line width of the second object as a parameter value for the second object;
When the parameter value for the first object is less than or equal to a first threshold, the calculation means regards the first object as an area composed of white pixel groups, and sets the parameter for the second object. If the value is less than or equal to the second threshold value, the second object is regarded as an area composed of white pixel groups, and then a projection in the row direction of the document image is obtained, and the resulting density The image processing apparatus according to claim 1, wherein a circumscribing rectangle of each character portion constituting the document image is obtained by obtaining a line by the step and then obtaining a projection in a direction orthogonal to the line for each line.   The calculation means regards the first object as an area configured by a group of white pixels when the length of the line made of the first object is equal to or greater than a third threshold value, and determines from the second object 3. The image processing apparatus according to claim 2, wherein the second object is regarded as an area composed of a group of white pixels when the length of the line is equal to or greater than a fourth threshold value. Means for detecting a third object comprised of adjacent black pixels in the document image;
The calculating means regards the first object as an area composed of a group of white pixels when the size of the third object including the first object is equal to or smaller than a threshold value. The image processing apparatus according to any one of claims 2 and 3. A value obtained by dividing the number of black pixels constituting the first object by the length in the row direction of the first object is obtained as a line width of the first object,
3. The value obtained by dividing the number of black pixels constituting the second object by the length in the column direction of the second object is obtained as a line width of the second object. 5. The image processing apparatus according to any one of items 4 to 4.   The image processing apparatus according to claim 2, wherein the first threshold value and the second threshold value are determined based on a line width of characters in the document image. An image processing method performed by an image processing apparatus,
Acquiring a document image in which the embedded information is embedded in accordance with an embedding method for embedding the embedded information by controlling the distance between characters according to the embedded information;
Specifying a blank line area in the document image when the line direction is a line direction of characters written in the document image and the column direction is a direction orthogonal to the line direction;
Identifying an area other than the blank line area as a character string area in the document image;
Identifying an inter-character area in the character string area;
When a pixel having the same pixel value as a pixel constituting the character part in the document image is a black pixel, and a pixel other than the black pixel is a white pixel, the document image is composed of adjacent black pixels in the inter-character area. Detecting a first object that is present;
Obtaining a line width of the first object as a parameter value for the first object;
A calculation step for obtaining a projection for each of the row direction and the column direction of the document image, and obtaining a circumscribed rectangle of each character part constituting the document image based on the obtained projection;
Extracting embedded information embedded in the document image based on a distance between each circumscribed rectangle,
In the calculation step, when the parameter value for the first object is equal to or less than a first threshold value, the first object is regarded as an area composed of a group of white pixels, and then the document image Obtaining a projection in the line direction, obtaining a line according to a density obtained as a result, and then obtaining a projection in a direction orthogonal to the line for each line to obtain a circumscribed rectangle of each character part constituting the document image An image processing method.   The computer program for functioning a computer as each means which the image processing apparatus of any one of Claims 1 thru | or 6 has.   A computer-readable storage medium storing the computer program according to claim 8.

Images