JP3848150B2 - Image processing apparatus and method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an image processing apparatus that embeds a digital watermark into a document image including a character composed of a plurality of character pieces, and is embedded into a document image including a character composed of a plurality of character pieces. The present invention relates to an image processing apparatus and an image processing method for extracting a digital watermark.
[0002]
[Prior art]
In recent years, in digital image forming apparatuses such as printers and copying machines, the image quality has been remarkably improved, and it has become possible to easily obtain high-quality printed materials. That is, anyone can obtain the required printed matter by image processing using a high-performance scanner, printer, copier, and computer. For this reason, problems such as illegal copying and falsification of documents have occurred, and in recent years, there has been an active movement to embed access control information in the printed matter itself in order to prevent or prevent them (digital watermark).
[0003]
Such functions include embedding access control information in an invisible form (invisible type), embedding a bitmap pattern corresponding to the access control information in the margin of the document, and applying scramble encryption to the document image. Things are now common. Of these, those that embed access control information invisibly are generally implemented as a method of embedding information by controlling the amount of space in the English character string, a type of rotating characters, and scaling characters A type that embeds information by deforming characters or the like has been proposed.
[0004]
FIG. 8 shows a type of printed matter that controls the amount of space in the English character string. Here, 801 is called a space. If the watermark bit to be embedded is 0, this space is set to p ← (1 + p) (p + s) / 2s ← (1-p) (p + s) / 2, and if the watermark bit is 1, p ← (1-p ) (P + s) / 2s ← (1 + p) (p + s) / 2.
[0005]
FIG. 9 shows a printed matter of a type that rotates characters. Here, (a) shows a state before rotation, and (b) shows a state after rotation. Reference numeral 901 denotes the rotation angle of the character.
[0006]
FIG. 10 shows an example of printing that enlarges or reduces characters. 1001 indicates the original size. Reference numeral 1002 denotes a size after enlargement / reduction.
[0007]
[Problems to be solved by the invention]
However, the above-mentioned access control information that is embedded invisible is embedded in the text itself (usually a binary image) with less redundancy, particularly in a text image, so that characters and spaces are uncomfortable. And the deterioration of the document quality becomes conspicuous. In general, such an image has low read resistance.
[0008]
The present invention is more has been made in view of the problems, while minimizing the deterioration of the typeface, an object of ensuring to Rukoto certain level of information embedding accuracy and amount.
[0009]
[Means for Solving the Problems]
The present invention can solve the above problems by providing the following configuration.
(1) An image processing apparatus that embeds watermark information in a document image that includes characters each consisting of a plurality of character pieces, the extraction unit extracting characters from the document image, and the extraction unit extracting the characters Among the characters, a selection unit that selects a character having a predetermined character fragment configuration, and a watermark on the selected character by changing a relative shape of the character fragment selected by the selection unit. An image processing apparatus comprising an embedding unit for embedding information.
[0010]
(2) An image processing apparatus for extracting watermark information embedded in a document image including a character in which one character is composed of a plurality of character pieces, the character extracting means for extracting a character from the document image, and the character The watermark information is selected based on the selection means for selecting a character whose character fragment configuration has a predetermined configuration among the characters extracted by the extraction means, and the relative shape of the character fragment of the character selected by the selection means. An image processing apparatus comprising watermark information extracting means for extracting.
[0011]
(3) An image processing method for embedding watermark information in a document image including a character consisting of a plurality of character pieces, wherein the character is extracted from the document image, and extracted in the extraction step Among the characters, a selection step of selecting a character whose character fragment configuration has a predetermined configuration, and changing the relative shape of the character fragment of the character selected by the selection means, thereby watermarking the selected character. An image processing method comprising an embedding step of embedding information.
[0012]
(4) An image processing method for extracting watermark information embedded in a document image including a character in which one character is composed of a plurality of character pieces, the character extracting step for extracting a character from the document image, and the character Based on the selection step of selecting a character whose character fragment configuration has a predetermined configuration from among the characters extracted in the extraction step, and the watermark information based on the relative shape of the character fragment of the character selected by the selection means An image processing method comprising a watermark information extracting step of extracting.
[0026]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
[0027]
Example 1
FIG. 1 is a block diagram showing a schematic configuration of a digital watermark apparatus according to an embodiment of the present invention. (A) is a block diagram of a digital watermark embedding unit, and (b) is a block diagram of a digital watermark reading unit.
[0028]
In (a), 2 is an image input means (input unit) such as a scanner, a camera, or a file reading device for inputting a target document for embedding a digital watermark, 4 is a processor for performing various processes, and 6 is a command to the processor 4. Keyboard for input, 8 is a disk for storing embedded information or a read document image, 10 is a temporary data storage for various processes performed in the processor 4, or a document image read by the image input means (input unit) 2 is stored. A memory 12, a command input to the processor 4, and a display showing the processing status, and 14 an output means (output unit) such as a printer for outputting a document image in which access control information is embedded.
[0029]
In (b), 22 is an image input means (input unit) such as a scanner, camera, or file reading device for inputting a document in which a digital watermark is embedded, 24 is a processor for performing various processes, and 26 is a command to the processor 24. , 28 is a disk for storing the read document image or retrieving the original file of the read document, 30 is a temporary data storage for various processing performed in the processor 24, or an image input means (input unit) ) A memory for storing the document image read in 22, 32 is a command input to the processor 24, a display showing the processing status, and 34 and 36 are a network interface and a printer for utilizing the read document access control information, respectively. is there.
[0030]
Next, the operation will be described. When embedding a digital watermark, according to a command input from the keyboard 6, first, an embedded document image digitized by the image input means (input unit) 2 is acquired and developed in the memory 10.
[0031]
Further, embedded information (document access control information) is input from the keyboard 6 or the disk 8, and the processor 4 embeds the information in a document image developed on the memory 10. The document image in which the predetermined document access control information is embedded is output as an electronic watermark embedded document by the output means (output unit) 14.
[0032]
On the other hand, at the time of reading the digital watermark, the digital watermark embedded document is first digitized by the image input means (input unit) 22 in accordance with a command input from the keyboard 26 and developed in the memory 30. Next, the document access control information embedded from the document image developed on the memory 30 is read by the processor 24, and predetermined processing is performed based on the instruction. The predetermined processing is, for example, reporting to the outside when an illegal reading is detected, searching the original document to the internal disk 28 or the outside, or printing out the attribute information. Therefore, the network I / F 34 and the printer 36 are used.
[0033]
The operations of various control processes executed by the digital watermarking apparatus of this embodiment, particularly the processors 4 and 24, will be described below with reference to FIGS. 2, 3, 4, 5, and 6. FIG.
[0034]
FIG. 2 is an overall flowchart when the processor 4 embeds a digital watermark in one document.
[0035]
S200 is a routine for performing control for taking a document from the text image input means (input unit) 2 and transferring it to the memory 10 as electronic image data. This routine also includes pre-processing such as orientation of the read document and skew correction.
[0036]
S202 is a routine for performing region identification on the document image developed in the memory 10 in S200 and extracting all character blocks in the image. This can be realized by applying, for example, a block selection technique described in JP-A-6-068301.
[0037]
S204 is a routine for performing character recognition on all the character blocks extracted in S202.
[0038]
S206 is a routine for extracting a target character to be embedded with the document access control information from each character converted into the character code in S204.
[0039]
S208 is a routine for inputting document access control information to be embedded in the characters extracted in S206. Here, the document access control information is, for example, copy restriction information, falsification prevention information, original document management information, and the like.
[0040]
S210 is a routine for embedding the document access control information input in S208 in the characters extracted in S206.
[0041]
S206 and S210 are routines that are the main subject of the present invention, and will be described in detail later.
[0042]
S212 is a routine for outputting the character image in which the document access control information is embedded in S210.
[0043]
FIG. 3 is an overall flowchart when the processor 24 extracts desired information from one digital watermark embedded document.
[0044]
S300 is a routine for performing control to take a digital watermark embedded document from the text image input means (input unit) 22 and transfer it to the memory 30 as electronic image data. This routine also includes pre-processing such as orientation correction and skew correction of the read document as in S200.
[0045]
S302 is a routine for performing region identification on the digital watermark embedded document image developed in the memory 30 in S300 and extracting all character blocks in the document image.
[0046]
S304 is a routine for performing character recognition on all the character blocks extracted in S302.
[0047]
S306 is a routine for extracting only the character in which the document access control information is embedded from each character converted into the character code in S304.
[0048]
S302, S304, and S306 are the same processing routines as S202, S204, and S206, respectively.
[0049]
S308 is a routine for reading the document access control information from the characters extracted in S306.
[0050]
Similar to S206 and S210, S306 and S308 are routines that focus on the present invention, and will be described in detail later.
[0051]
S310 is a routine for performing predetermined control according to the document access control information read in S308, and instructs, for example, copy prohibition processing, document search processing, and the like.
[0052]
FIG. 4 is a detailed flowchart of a character extraction routine (S206, S306) for embedding / reading a digital watermark which is the main object of the present invention.
[0053]
S400 is a routine for performing control to transfer the character code to the character extraction work memory based on the character recognition result.
[0054]
S402 is a routine for determining whether or not all character codes included in the document have been transferred to the character extraction work memory. If all the transfers have been completed, the process proceeds to S404. If not, the process proceeds to S400.
[0055]
S404 is a routine for counting, for each character, target characters to be embedded with document access control information set in advance. Here, “preliminarily set” means that, for example, a Chinese character having a somewhat complicated radical configuration, such as a Chinese character having three or more constituent radicals, is set in advance. By performing such a setting, it becomes possible to embed information of a certain level without making it conspicuous. This will be described in detail later.
[0056]
S406 is a routine for sorting the target characters counted in S404 by the number of counts.
[0057]
S408 is a routine for determining whether or not the number of counts exceeds a certain level, that is, whether or not there are a certain number of embedding target characters frequently appearing in the document. This is a measure for taking a predetermined number or more of characters to ensure embedding / reading accuracy, that is, repeatedly embedding the same information for the same character. Although the accuracy increases as the number is more than a certain number here, for example, as illustrated in FIG. By doing so, the reading accuracy is greatly improved.
[0058]
Further, if the number of such target characters exceeds a certain number, in order to guarantee the amount of embedded information, for example, if it is possible to embed 3 bits of information per character as shown in FIG. Thus, if it is 11 characters, it becomes possible to embed 30-bit information, but this is to limit the number of characters.
[0059]
Here, if it is determined that there is no target character number above a certain level, it is determined that embedding of a predetermined amount of information is impossible, and control is shifted to S414, and if not, control is shifted to S410.
[0060]
S410 is a routine for selecting a character having the highest appearance frequency in the document from among the embedding target characters and calculating a reference value for embedding / reading operation (corresponding to (4) in FIG. 6). This reference value will be described later.
[0061]
S412 is a routine for performing information embedding / reading operations by obtaining the result of the sorting in S406 of characters to be embedded other than the character for which the reference value has been obtained, that is, the second and subsequent characters, for example, up to 11th in the case of 30 bits. A specific method will be described in detail later with reference to FIGS.
[0062]
S414 is a routine for performing a predetermined process when it is determined in S408 that there are few characters to be embedded and it is impossible to embed. The predetermined process is, for example, issuing a warning.
[0063]
FIG. 5 is a diagram for explaining the principle of the information embedding / reading method performed in S412, and FIG. 6 is an explanatory diagram in the case of embedding a 30-bit information amount in an actual document.
[0064]
In FIG. 5 and FIG. 6, for example, a pattern with three small radicals at the top and two large radicals at the bottom, such as a pattern of three letters, for example, “type”. Take up and explain. Note that the drawings are exaggerated for the sake of explanation.
[0065]
First, the extracted character image is decomposed into individual radicals to obtain the reference value. The reference value is the most important value when embedding document access information in an invisible form in the present invention.
[0066]
More specifically, as described in FIG. 5A, the absolute height and width of the character image are represented by the ratio of the width and height of each decomposed radical. To explain further, although it is referred to as each radical here, only the upper two radicals are used, and the proportion of the size of the lower radical is not defined. If this is defined for all radicals, the relative reference value will fluctuate, and further, the deformation of the character will become large and the deterioration will be conspicuous. One is used and only the other two radicals (in this case) are used.
[0067]
As a specific information embedding method, consider embedding information of 3 bits per character using the four reference values k, m, n, and p defined earlier. In FIG. 4, S410, the reference values k, m, n, and p of the character having the highest reliability, that is, the appearance frequency are obtained (corresponding to FIG. 6 (4)). On the other hand, document access control to be embedded, for example, 30 bits are prepared, and the second and subsequent characters selected in FIG. 4 and S412 are set in order according to the rule shown in the right half of FIG. K ′, m ′, n ′, and p ′ are obtained for each bit, and the aspect ratio of each radical of the target character is changed to reflect the result (corresponding to FIG. 6 (5)). Arbitrary information can be embedded by such a procedure.
[0068]
At the time of reading, character recognition is performed, the characters are extracted, sorted, and until the first reference value is obtained, it is exactly the same as when embedding, and finally the aspect ratio of each radical after the second is the reference value And the allocated bit array is reproduced. Usually, this is performed a plurality of times (here, at least two times or more), and it is possible to make a judgment based on the majority vote or the like to further improve accuracy.
[0069]
FIG. 6 shows a procedure for actually embedding 30-bit information such as “001 100 111 000 010 011 111 101 111 000”.
[0070]
As described above, according to the present invention, after performing region identification and character recognition, skillful use of the relative shape change for each radical of a complex Chinese character having a radical structure makes it possible to minimize deterioration of the font. However, it is possible to achieve a digital watermarking device with high noise resistance even at the time of reading by ensuring the accuracy and amount of information embedding beyond a certain level by sorting target characters according to appearance frequency. Further, since there is no dependency on the font size in principle, it is clear that this is an effective method even for an original with a small number of characters.
[0071]
(Example 2)
In the first embodiment, the radical configuration adopted as the embedding target was a single pattern as shown in FIG. 5, but nothing is limited to this, as shown in FIG. A plurality of Chinese character patterns having radicals may be set simultaneously. In this case, the amount of embedded information further increases.
[0072]
【The invention's effect】
As described above, according to the present invention, by utilizing the relative shape change of a character piece with respect to a document image including a character composed of a plurality of character pieces, the deterioration of the font is minimized. However, it is possible to secure a certain degree of information embedding accuracy and amount by sorting the target characters based on the appearance frequency, and to realize a digital watermark device having high noise resistance even at the time of reading. Further, since there is no dependency on the font size in principle, it is clear that this is an effective method even for an original with a small number of characters.
[Brief description of the drawings]
FIG. 1 (a) is a block diagram showing a schematic configuration of a digital watermark embedding unit of a digital watermark apparatus according to an embodiment of the present invention. FIG. 1 (b) is an electronic diagram of a digital watermark apparatus according to an embodiment of the present invention. FIG. 2 is a block diagram showing a schematic configuration of a watermark reading unit. FIG. 2 is an overall flow chart when the processor 4 embeds a digital watermark in one document. FIG. 3 shows that the processor 24 receives desired information from one digital watermark embedded document. Overall Flowchart for Extraction [FIG. 4] Detailed Flowchart of Target Character Extraction Routine (S206, S306) for Embedding / Reading Digital Watermarking Which is the Subject of the Present Invention [FIG. 5] FIG. Explanatory diagram of the principle [FIG. 6] (1), (2), (3), (4), (5) Explanatory diagram when embedding 30 bits of information in an actual document [FIG. 7] Second Fruit Examples of illustration (other embedded / read radical pattern)
FIG. 8 is an explanatory diagram of a conventional technique. FIG. 9 is an explanatory diagram of a conventional technique. FIG. 10 is an explanatory diagram of a conventional technique.
2 Image input means (input unit)
4 processor 6 keyboard 8 disk 10 memory 12 display 14 output means (output unit)
22 Image input means (input unit)
24 processor 26 keyboard 28 disk 30 memory 32 display 34 network interface 36 printer

Claims (11)

An image processing apparatus that embeds watermark information in a document image that includes a character consisting of a plurality of character pieces.
  Extraction means for extracting characters from the document image;
  A selection means for selecting a character whose character fragment configuration has a predetermined configuration among the characters extracted by the extraction means;
  An image processing apparatus comprising: an embedding unit that embeds watermark information in the selected character by changing a relative shape of a character piece selected by the selecting unit. The extraction means further includes
  A character block extracting means for extracting a character block from the document image;
  Character recognition is performed for characters included in the character block extracted by the character block extraction means, a character code is generated as a recognition result, and character recognition means for extracting an image of the character from the character block is provided. The image processing apparatus according to claim 1, wherein: The selection unit includes a counting unit that counts, for each character, the number of characters in which the character piece configuration of the character piece has a predetermined configuration among the characters extracted by the extraction unit,
  The embedding unit embeds the watermark information in the character selected by the selecting unit when the number of characters whose count by the counting unit is equal to or greater than a predetermined number. Item 3. The image processing apparatus according to Item 1 or 2. The image processing apparatus according to claim 1, further comprising a calculation unit configured to calculate a reference value used when embedding the watermark information. 5. The calculation unit according to claim 4, wherein a ratio of the width and height of each character piece of the character to the width and height of the character selected for calculating the reference value is calculated as the reference value. An image processing apparatus according to 1. An image processing apparatus for extracting watermark information embedded in a document image including a character including a character consisting of a plurality of character pieces,
  Character extraction means for extracting characters from the document image;
  A selection means for selecting a character whose character fragment configuration has a predetermined configuration among the characters extracted by the character extraction means;
  An image processing apparatus comprising: watermark information extraction means for extracting the watermark information based on a relative shape of a character piece selected by the selection means. The character extraction means further includes
A character block extracting means for extracting a character block from the document image;
Character recognition is performed for characters included in the character block extracted by the character block extraction means, a character code is generated as a recognition result, and character recognition means for extracting an image of the character from the character block is provided. The image processing apparatus according to claim 6. 8. The image processing apparatus according to claim 1, wherein the watermark information includes at least one of document access control information, copy restriction information, falsification prevention information, and original document management information. The image processing apparatus according to claim 1, wherein the character piece includes a radical of a Chinese character. An image processing method for embedding watermark information in a document image that includes characters each consisting of a plurality of character pieces,
  An extraction step of extracting characters from the document image;
  Among the characters extracted in the extraction step, a selection step of selecting a character whose character fragment configuration has a predetermined configuration;
  An image processing method comprising: an embedding step of embedding watermark information in the selected character by changing a relative shape of a character piece of the character selected by the selection means. Embed a document image that contains a character consisting of multiple characters. An image processing method for extracting rare watermark information,
  A character extraction step of extracting characters from the document image;
  Of the characters extracted in the character extraction step, a selection step of selecting a character whose character fragment configuration has a predetermined configuration;
  An image processing method comprising: a watermark information extraction step for extracting the watermark information based on a relative shape of a character piece selected by the selection means.

Images