JP5506588B2 - Image processing method and image processing apparatus using digital watermark - Google Patents

Description

  The present invention relates to a technique for embedding information by digital watermark in characters and figures included in an image.

  For the purpose of information leakage countermeasures, copyright management, copy control, etc. for information recorded on electronic media, digital watermark technology has been developed that inserts additional information so that it is not conspicuous in the content. In recent years, a digital watermark that can be applied to printed matter by applying a digital watermark technique to an image has been proposed.

  Since digital watermarking is a technique for inserting information by changing pixel values, degradation of image quality is inevitable. If the target image is an image composed of binary values such as white and black, the possible values of the pixel values are limited. For example, the shape of characters or figures in the image can be changed. In order to insert information inconspicuously, various techniques have been developed.

  In the case of a multi-value image in which the values that the pixels constituting the image can take are three or more values, information can be inserted inconspicuously by changing the brightness and color of the pixels minutely. However, in an image such as a document image where the area of the foreground part where the change of the pixel value is allowed is small and the area of the background part where the change of the pixel value is not allowed is large, only a slight change in the brightness and color of the foreground pixel Information embedding with sufficient tolerance may not be realized. In such a case, for example, tolerance cannot be ensured unless embedding using foreground shape change or the like is executed.

In the method of embedding information by changing the foreground shape,
Information embedding with minimal visual impact by giving the criteria and priority for determining the visibility of modification of pixels in an image based on human visual characteristics, and by modifying the pixels in order from the highest priority pixel. A method for realizing the above has been considered (see Non-Patent Document 1). Further, in information embedding, a method of suppressing image quality deterioration by detecting “cut” caused by pixel change and correcting the detected “cut” has been considered (see Patent Document 1).

JP 2008-219680 A

M. Wu and B. Liu: "Data Hiding in Binary Image for Authentication and Annotation", IEEE Trans. On Multimedia, vol. 6, no. 4, pp.528-538, August 2004

  Non-Patent Document 1 describes that determination of a pixel value modification criterion / priority order is determined based on a local arrangement of 3 × 3 pixels around a pixel of interest. Non-Patent Document 1 further describes that a pixel arrangement in a wider neighborhood is used in order to reduce the visual effect, but a specific method is not shown, and as the area to be used increases, There is a problem that the required memory capacity and calculation amount increase exponentially.

  Japanese Patent Application Laid-Open No. 2004-228561 corrects a cut caused by watermark embedding by using a labeling process for an image with embedded watermark and a pixel value changing process corresponding to the number of labels in the vicinity of the changed pixel by watermark embedding, thereby reducing image quality degradation. , Is described.

  However, Japanese Patent Application Laid-Open No. 2004-151561 discloses a cutout correction method suitable when the target image is a binary image. When the target image is a multivalued image, the foreground is processed prior to the cutout correction process. In addition, there are problems such as that it is necessary to determine pixel values for the background one by one in advance, and that the pixel values for the foreground and the background of the region where cutout can be corrected are limited. There is no specific disclosure regarding a cutout correction method in which a pixel value serving as a foreground or background is not determined in advance.

  In the present specification, a technique for reducing the degree of image quality degradation in information embedding in an image using digital watermarking is disclosed.

  In other words, a technique for suppressing image quality degradation by detecting “cut” caused by pixel change and correcting the detected “cut” in information embedding with an electronic watermark in an image is disclosed.

  One feature of the image quality deterioration suppression technique is that when the target image is a multi-valued image, it is not necessary to determine the foreground and background pixel values in advance prior to the cutout correction process.

One of the more specific modes disclosed is an image processing method for correcting a break that occurs due to a change in a pixel value in an information embedding process using a digital watermark,
Regarding the watermark image that is the result of the above information embedding process for the input image,
A pixel that has a pixel value that is greater than a preset threshold value for any adjacent pixel, or a pixel that is adjacent to each other and whose difference between pixel values is not greater than the threshold value When a set is composed of a plurality of pixels having values, each pixel belonging to the same set is assigned the same label value but different from the pixels belonging to the other set. A labeling step for outputting a label image identifiable by the label value;
A set pixel value setting step for setting one or a plurality of set pixel values representing pixel values of pixels belonging to the set to each of the sets;
A set representing the pixel values before the information embedding process of the target pixel included in a preset neighborhood area in the label image for the pixel whose pixel value has been changed by the information embedding process for the input image. A label determination processing step for identifying a set having pixel values;
In the label determination process, when there are two or more identified sets, a change cancellation process for performing a change cancellation process for returning the pixel value of the target pixel of the watermark image to the pixel value before the information embedding process Steps,
And a corrected image output step of outputting a corrected image subjected to the change canceling process.

  Furthermore, in the image processing method, the labeling step may include a pixel value quantization step for quantizing the pixel value of the watermark image with one or more preset threshold values.

  Further, in the image processing method, the set pixel value is one or a plurality of pixel values included in a pixel value of a pixel belonging to one set, or the plurality of pixel values Of these, the most frequently occurring pixel value or a pixel value within a certain range including the average value of the plurality of pixel values may be used.

  Furthermore, in the image processing method, the pixel targeted by the labeling step is a pixel having any one or a plurality of foreground pixel values set in advance, and the pixel targeted by the label determination processing step Is a pixel whose pixel value has been changed from one of the one or more foreground pixel values to a pixel value different from any of the one or more foreground pixel values by the information embedding process for the input image. There may be.

  Further, in the image processing method, the pixel targeted by the label determination processing step is a pixel whose pixel value has been changed to follow a preset increase / decrease in brightness by the information embedding process for the input image. May be.

  As described above, the disclosed image processing method applies the labeling process to the watermark image after the digital watermark is inserted, and observes the number of labels around the changed pixel, thereby eliminating the breaks caused by the watermark insertion. One main feature is to detect and correct it.

  The image in this specification includes a binary image and a multi-value image. The binary image in this specification refers to an image in which the pixel values of the pixels included in the image are one of two pixel values such as white and black or white and red. Point to. On the other hand, a multi-valued image in this specification refers to an image in which the pixel values that can be taken by the pixels included in the image are any of three or more pixel values. The pixel value may be, for example, a luminance value (256 steps from 0 to 255) represented by an 8-bit information amount, and the three elements of color are red (Red (R)) and green (Green (Green ( G)) and blue (Blue (B)) may be a combination of these three intensity values (R, G, B) in which the intensity of each element is represented by an 8-bit information amount. Alternatively, values and combinations of values in other color systems may be used, or other values other than these may be used.

  An image in this specification refers to a set of a plurality of pixels. Therefore, the disclosed contents of the present specification can be applied to a part of images included in larger contents.

  The point of the labeling process is that any pixel that is adjacent (may include diagonal) is adjacent to each other by one pixel whose difference between pixel values is larger than a preset threshold value, or that is adjacent (may also include diagonal). In addition, when one set is constituted by a plurality of pixels having pixel values whose difference between pixel values is not greater than a threshold value, a label image that can identify the set to which each pixel belongs is output.

  More specifically, the same pixel value is assigned to each pixel belonging to the same set, and a label value different from the label value assigned to pixels belonging to another set is assigned.

  Prior to the labeling process, the pixel value of the labeling target image (watermark image) may be quantized using one or more preset threshold values.

  There are a number of documents on specific algorithms for labeling. For example, from page 173 to page 180 of the following document, there is a description relating to labeling processing for binary images.

"Binary Digital Image Processing: A Discrete Approach", Stephane Marchand-Maille, Yazid M. Sharaiha, Academic Pr (1999/12)
The disclosed correction processing can be applied to any digital watermark algorithm in which image modification for information embedding has redundancy, and digital watermark embedding with less visual influence can be realized by application.

  It becomes possible to further reduce the degree of image quality degradation due to digital watermark embedding.

  The purpose of preventing cuts, which is the main cause of image quality degradation accompanying digital watermark embedding, is realized without depending on the digital watermark embedding algorithm.

  In this embodiment, the image is a general multi-valued image, and no limitation is imposed on the pixel values of the foreground and background pixels.

  FIG. 1 is a block diagram showing a configuration of a digital watermark embedding system 10 according to an embodiment of the present invention.

  In the present embodiment, the digital watermark embedding system 10 accepts the image content 21 and the correction parameter 31 as inputs, and outputs the watermarked image content 22 that has undergone the watermark embedding and cutout correction processing. In this specification, a watermarked image may be simply referred to as a watermark image.

  The digital watermark embedding system 10 includes a digital watermark insertion device 20 and a cut correction device 30.

  The digital watermark insertion apparatus 20 includes an input unit 200, a watermark image synthesis unit 210, and an output unit 220. The input unit 200 receives the image content 21 to be embedded with the digital watermark and a user input operation via a keyboard, a mouse, or the like, and sets parameters related to the watermark image synthesis in the watermark image synthesis unit 210 according to the operation. The watermark image composition unit 210 is instructed to start the watermark image composition.

  The watermark image synthesis unit 210 receives the image content 21 and parameters related to the watermark image synthesis from the input unit 200, embeds a digital watermark in the image content 21, and synthesizes the watermark image. The output unit 220 receives an image that has undergone the digital watermark embedding process and the cutout correction process, and outputs the image as the watermark image content 22 from the watermark insertion device 20.

  The cut correction device 30 includes a correction parameter input unit 300, an input image input unit 310, a watermark image input unit 320, a labeling unit 330, a watermark image correction unit 340, and a corrected image output unit 350. The correction parameter input unit 300 receives the correction parameter 31 recorded in advance in an external file or the like, and sets parameters related to processing in the labeling unit 330 and the watermark image correction unit 340.

  The input image input unit 310 and the watermark image input unit 320 receive the input image content 21 and the watermark image from the digital watermark insertion device 20, respectively. The labeling unit 330 receives parameters related to the labeling process from the correction parameter input unit 300 and the watermark image from the watermark image input unit 320, and performs a labeling process to be described later on the watermark image. The watermark image correction unit 340 receives parameters related to the image correction processing from the correction parameter input unit 300, the input image content 21 from the input image input unit 310, the watermark image from the watermark image input unit 320, and the label image from the labeling unit 330. Each is received, and a cutout correction process (change cancellation process) described later is applied to the watermark image to create a corrected image. The corrected image output unit 350 receives the corrected image from the watermark image correcting unit 340 and transmits the corrected image to the output unit 220 of the digital watermark inserting apparatus 20.

  The digital watermark insertion device 20 is configured to complete the digital watermark embedding process with only the input unit 200, the watermark image synthesis unit 210, and the output unit 220 regardless of the presence of the cutoff correction device 30. Is added, and the digital watermark embedding system 10 is configured to realize digital watermark embedding with less visual influence.

  Each of the digital watermark embedding system 10, the digital watermark insertion device 20, and the cut correction device 30 according to the present embodiment can be realized on a general computer, and FIG. 2 shows an example of the hardware configuration thereof. .

  The computer 40 includes a CPU 400, a RAM 410, a storage 420, a user interface 430, a communication interface 440, and a media interface 450. The CPU 400 operates based on programs stored in the RAM 410 and the storage 420, and implements each processing unit shown in FIG. In the following embodiments, each processing unit will be described as an execution subject.

  The RAM 410 temporarily stores programs executed by the CPU 400 and data such as images used by the CPU 400. The storage 420 stores a boot program executed by the CPU 400 at startup, a program depending on hardware, a program for realizing each processing unit shown in FIG. 1, image content, and the like. The user interface 430 is an input device such as a mouse or keyboard and a display device such as a display. The user operates the digital watermark embedding system 10 via the user interface 430.

  The communication interface 440 is used to communicate with an external device via the network 460 or the like, receives the image content 21 to be embedded with the digital watermark, stores it in the storage 420, and further stores the watermark image content 22 in the external device or network. Used to transmit to 460. The media interface 450 reads the image content 21 to be embedded with the digital watermark from the recording medium 470 and stores it in the storage 420, and further records the watermark image content 22 on the recording medium 470.

  When the digital watermark insertion device 20 and the cut correction device 30 are realized on different computers 40, communication between the devices is performed via the communication interface 440.

  A program for realizing the digital watermark insertion device 20 and the cutout correction device 30 in the computer 40 is received from an external device via a communication medium by the communication interface 440 or from the recording medium 470 by the media interface 450, The data is provided to the CPU 400 via the storage 420. The communication medium refers to a network 460 or a carrier wave or a digital signal that propagates through the network 460.

  FIG. 3 is a conceptual diagram showing an outline of labeling processing executed by the labeling unit 330.

  The labeling process executed by the labeling unit 330 is a process of allocating a target pixel among pixels included in an input image to the labeling unit 330 to one of a plurality of sets based on a parameter. This is a process of assigning one label each indicating the set to which it belongs.

  As a result of the labeling process, one or a plurality of pixel values can be associated with each label. That is, for each set, a set pixel value representing a pixel value of a pixel belonging to the set is given.

  When a plurality of pixels assigned with one and the same label all have the same pixel value, the pixel values that the plurality of pixels have in common may be associated with the label as one aggregate pixel value . For example, in a multi-value image in which the pixel value is a luminance value represented by an 8-bit information amount, when the luminance values of a plurality of pixels to which one identical label is assigned are all “120”, “120” can be associated with the collective pixel value of the label. Alternatively, a plurality of pixel values including pixel values shared by a plurality of pixels may be used as the aggregate pixel value. This is because, for example, when the luminance values of a plurality of pixels assigned with the same label are all “120”, a plurality of luminance values such as “brightness values from 110 to 130” are set to the collective pixel value. And so on.

  On the other hand, when a plurality of pixels assigned with the same label have a plurality of types of pixel values, a plurality of pixel values obtained from the plurality of types of pixel values may be used as the aggregate pixel value. For example, all of a plurality of types of pixel values of a plurality of pixels having the same label may be simply set as a set pixel value, or a wider range of pixel values including all of the plurality of types of pixel values may be set. It may be a pixel value. In addition, a pixel value within a certain range including the most frequently occurring pixel value or the average value of the plurality of types of pixel values among the plurality of types of pixel values may be used as the collective pixel value.

  Various methods can be considered as a method for obtaining the aggregate pixel value, and the present invention is not limited to these specific examples, and any method can be applied.

  The labeling process includes a pixel value quantization process and a label value giving process. As parameters of the labeling process, there are a threshold value used in a pixel value quantization process executed as a pre-process, and a threshold value for determining whether or not adjacent pixels are allocated to the same set in a label value adding process. Note that the quantization process may be executed or may not be executed.

  3A shows the image data 500 received by the labeling unit 330, FIG. 3B shows the image data 510 after the pixel value quantization processing by the labeling unit 330, and FIG. 3C shows the label value by the labeling unit 330. The image data 520 after the application process is represented.

  In FIG. 3, the size of one minimum square represents the size of one pixel, and for the purpose of facilitating the understanding of the positional relationship between the pixels, each pixel is separated by a grid line. 3A and 3B, pixels having the same pixel value are represented by squares having the same pattern.

  Pixel value quantization processing is processing that reduces the types of pixel values included in an image. Various methods can be applied to specific quantization processing. For example, when the pixel value is represented by a single value called a luminance value, there is a method of quantizing the luminance value according to a preset quantization threshold. Also, for example, the three elements of color are red (Red (R)), green (Green (G)), and blue (Blue (B)), and depending on the combination of intensity values (R, G, B) of each element When a pixel value is represented, a method of calculating a single value such as a luminance value from the set of values and performing quantization according to a preset quantization threshold for the calculated single value There is also a method of performing quantization according to a preset quantization threshold for each of a plurality of values constituting a pixel value. The quantization method is not limited to these specific examples, and any method can be applied.

  In the example of FIG. 3, it is assumed that the pixel value is a luminance value in 256 levels (“0” to “255”). Assume that the quantization threshold is set every luminance value “32” when the pixel values A, B, C, and D are luminance values “10”, “40”, “200”, and “20”, respectively. As a result of the quantization, the pixel value A (luminance value “10”) and the pixel value D (luminance value “20”) are changed to the luminance value “0”, and the pixel value B (luminance value “40”) is the luminance value “32”. The pixel value C (luminance value “200”) is changed to the luminance value “192”. The image data 500 before the pixel value quantization process includes four types of pixel values, whereas the image data 510 after the process includes only three types of pixel values.

  The label value assigning process is a process in which the difference between pixel values (pixel values after quantization when the pixel values are quantized) is larger than a preset threshold value for any adjacent (which may include diagonal) pixels. When one set is composed of one pixel or a plurality of pixels that are adjacent to each other (may include diagonal) and whose difference between pixel values is not greater than the threshold value, This is a process for assigning a label value that is the same to a pixel and different from that of a pixel belonging to another set. The set to which each pixel belongs can be identified by the output label image.

  What is the definition of the difference between pixel values as long as it can be interpreted as “both pixel values are equal when the difference is 0, and the greater the difference, the greater the difference between the two pixel values”? May be. For example, when a pixel value is represented by a single value called a luminance value, the absolute value of the difference in luminance value may be adopted as the difference between the two pixel values. Also, for example, when the pixel value is represented by a combination of the intensity of the three-element RGB of the color, the absolute value of each RGB difference between the two pixels is calculated, and the largest value among the two pixel values is calculated. Can be adopted as the difference between. Alternatively, a value called “color difference” calculated by converting two pixel values into the L * a * b * color system can be adopted. The values adopted as the differences between the pixel values are not limited to these specific examples, and any value can be applied.

  In the example of FIG. 3, when the absolute value of the luminance value difference is adopted as the difference between the pixel values and the threshold value is “64”, each pixel of the image data 510 has four types of 1, 2, 3, 4 Any one of the labels is given, and image data 520 is output. As can be seen by comparing FIG. 3B and FIG. 3C, in this example, a plurality of pixels assigned the same label for three types (1, 3, 4) of the four types of labels are: All have the same pixel value. Therefore, for a set identified by each of these labels, a pixel value shared by a plurality of pixels belonging to the set can be used as a set pixel value of the set. That is, the collective pixel value of label 1 and label 4 can be the luminance value “192”, and the collective pixel value of label 3 can be the luminance value “0”. On the other hand, since the pixel given the label 2 has the luminance value “0” or “32”, as the collective pixel value of the set identified by the label 2, for example, both of these luminance values “0” and “ 32 "can be set.

  FIG. 4 is a flowchart showing an example of the operation of the digital watermark embedding system 10.

  The operation of the digital watermark embedding system 10 shown in this flowchart starts at a predetermined timing such as when an image is input to the input unit 200. First, the watermark image synthesis unit 210 embeds a digital watermark in the input image to create a watermark image (S100). When creating a watermark image, the pixel value before being changed by the digital watermark embedding process may be stored. Next, the labeling unit 330 performs a labeling process on the watermark image and outputs a label image (S101).

  Next, the watermark image correcting unit 340 starts pixel scanning of the received input image, watermark image, and label image (S102). The watermark image correcting unit 340 determines, for each scanning pixel, whether the pixel is a pixel whose pixel value has been changed by the watermark image combining unit 210 (S103). Whether or not the pixel value has been changed may be determined by comparing a difference between the pixel values before and after the processing of the watermark image composition unit 210 with a predetermined threshold.

  When the scanning pixel is a pixel whose pixel value has been changed by the watermark image synthesis unit 210 (S103: Y), the watermark image correction unit 340 is located in the vicinity of the scanning pixel, and is a pixel before the change of the scanning pixel. A label value having a collective pixel value representing the value is identified, and a label determination process is performed to determine whether there are two or more types of label values (S104). The shape and size of the neighboring region observed at the time of this determination depend on the parameters set by the correction parameter input unit 300. Next, when the watermark image correcting unit 340 determines that there are two or more types of label values identified in the vicinity region of the scanning pixel (S104: Y), the watermark image correcting unit 340 uses the pixel value of the scanning pixel as the watermark value. A change cancellation process (cut correction process) for returning to the pixel value before being changed by the image composition unit 210 is performed (S105), and the process proceeds to scanning of the next pixel (S106).

  On the other hand, when the watermark image correcting unit 340 determines that the scanned pixel is not a pixel whose pixel value has been changed by the watermark image synthesizing unit 210 (S103: N), scanning before the change included in the vicinity region of the scanned pixel is performed. When it is determined that the label value having the collective pixel value representing the pixel value of the pixel is not more than one type (S104: N), the watermark image correcting unit 340 immediately moves to the next pixel scan (S106).

  Next, the watermark image correction unit 340 determines whether or not scanning of all pixels has been completed (S107). If the watermark image correction unit 340 determines that all pixels have been scanned (S107: Y), the watermark correction image is output from the output unit 220 (S108), and the watermark embedding by the watermark embedding system 10 shown in this flowchart is performed. End the process. If the watermark image correction unit 340 determines that scanning of all pixels has not been completed (S107: N), the pixel value has been changed by the watermark image composition unit 210 for the next unscanned pixel. The process returns to the step of determining whether or not it is a pixel (S103).

  FIG. 5 is a conceptual diagram for explaining the effect of the image correction processing of the watermark image correction unit 340. Here, when the neighborhood region determined by the correction parameter received by the watermark image correction unit 340 from the correction parameter input unit 300 is a square composed of pixels centered on the target pixel and having a distance of 1 from the target pixel. A case where the pixel is a square having a pixel centered on the pixel of interest and having a distance of 2 from the pixel of interest will be described as an example. The shape and size of the neighborhood region are not limited to these specific examples, and any shape can be applied.

  FIG. 5A shows an image 600 included in an input image to the digital watermark embedding system 10. The pixels included in the image 600 have one of two types of pixel values. Here, these pixel values are referred to as a pixel value A and a pixel value B. In FIG. 5, the pixels having the pixel value A and the pixel value B are represented by a black filled region and a halftone dot region, respectively. FIG. 5B shows an image 610 as a result of the watermark embedding process performed on the image 600 by the watermark image synthesis unit 210. In the image 610, the B pixel 612 whose pixel value has been changed from A to B by the watermark image synthesis unit 210 is expressed by being surrounded by a dotted line.

  Pixels included in the image 610 are labeled by a labeling unit 330 and include an A pixel block 614 including one or more A pixels, an A pixel block 616 including one or more A pixels, and a B pixel including one or more B pixels. They are classified into three sets of pixel blocks 618 and labeled.

  All of the pixels included in each set have the same pixel value, and the pixel value that they have in common is set as the set pixel value. The watermark image correction unit 340 receives the image 600, the watermarked image 610, and an image labeled with the image 610, and starts scanning pixels. The watermark image correction unit 340 does not correct pixels that have not been changed by the watermark image synthesis unit 210. For the B pixel 612 whose pixel value has been changed from A to B by the watermark image synthesizing unit 210, the watermark image correcting unit 340 observes the number of label values having a collective pixel value representing the pixel value A in the neighboring area. .

  FIG. 5C is a conceptual diagram 620 showing a neighborhood region 624 of the distance 1 of a certain target pixel 622 in the B pixel 612 whose pixel value has been changed from A to B by the watermark image synthesis unit 210. FIG. 5D shows a conceptual diagram 630 showing the vicinity region 634 of the distance 2 of the target pixel 632 at the same position.

  In the conceptual diagram 620, a region 624 near the distance 1 of the target pixel 622 (displayed with hatched hatching) belongs to a set having the pixel value (pixel value A) of the target pixel 622 before the change as a set pixel value. Pixels are not included. Therefore, when the neighborhood distance that is the correction parameter received by the watermark image correction unit 340 from the correction parameter input unit 300 is 1, the watermark image correction unit 340 does not change the target pixel 622. Similarly, for all the B pixels 612 whose pixel values have been changed from A to B by the watermark image composition unit 210 included in the image 610, the pixel values A included in the neighborhood region 624 of the distance 1 are set to the aggregate pixel value. Since there are no more than one type of label value, when the neighborhood distance is 1, the image output from the watermark image correction unit 340 is equal to the input watermark image to the watermark image correction unit 340 and becomes an image 610.

  In the conceptual diagram 630, the vicinity region 634 at the distance 2 of the target pixel 632 (displayed by hatching) includes a part of the A pixel block 614 and a part of the A pixel block 616. The number of label values in which the pixel value (pixel value A) of the target pixel 632 before the change included in the neighborhood region 634 is the aggregate pixel value is two. Therefore, when the neighborhood distance that is the correction parameter received by the watermark image correction unit 340 from the correction parameter input unit 300 is 2, the watermark image correction unit 340 changes the target pixel 632 to the A pixel. Similarly, the watermark image correcting unit 340 includes, in the image 610, an observation of the number of label values having the pixel value of the target pixel before the change in the neighboring region 634 at the distance 2 as the aggregated pixel value. In order to change all the B pixels 612 whose pixel values have been changed from A to B by 210, and to change the pixel values of the four B pixels included in the B pixel 612 to A, the watermark image An image output from the correction unit 340 is an image 640 represented in FIG.

  When the neighborhood distance that is the correction parameter received by the watermark image correction unit 340 from the correction parameter input unit 300 is 1, the watermark image correction unit 340 does not correct the image 610, but the neighborhood distance is 2. In this case, as can be seen by comparing the image 610 and the image 640, the watermark image combining unit 210 changes the A pixel to the B pixel, and the watermark image correcting unit 340 converts a part of the B pixel to the A pixel. It is corrected by returning to. As described above, the effect of the cut correction differs depending on the shape and size of the neighboring area.

  FIG. 6 is a conceptual diagram illustrating how the appearance of characters included in the image content input in the process of the digital watermark embedding process using the digital watermark embedding system 10 changes. A pixel included in the image shown in the conceptual diagram has one of two types of pixel values. Here, the pixels having these two types of pixel values are represented by white and black areas. In the correction parameter input unit 300, the proximity distance which is a correction parameter is set to 2.

  FIG. 6A shows an outline drawing 700 of the character “y” included in the input image. FIG. 6B shows an outline drawing 710 of the character “y” included in the same place as “y” in FIG. 6A of the watermark image created by the watermark image composition unit 210. FIG. 6C shows an outline drawing 720 of the letter “y” included in the same place as “y” in FIGS. 6A and 6B of the label image created by the labeling unit 330. Represent. In FIG. 6C, all the pixels are labeled by the labeling unit 330 as one of a pixel block 721, a pixel block 722, a pixel block 723, a pixel block 724, a pixel block 725, and a pixel block 726. The pixel block 726 is a set of pixels having pixel values represented by white, and is the set having the largest area in this figure. 6D is included in the same place as “y” in FIGS. 6A, 6B, and 6C in the output image created by the watermark image correcting unit 340. FIG. The outline drawing 730 of the character “y” is shown.

  By comparing FIG. 6B and FIG. 6D, it can be seen that the watermark image correcting unit 340 corrects the cuts generated by the watermark combining processing of the watermark image combining unit 210. Further, by comparing FIG. 6A and FIG. 6D, the image correction processing by the watermark image correcting unit 340 does not correct all the pixels changed by the watermark image combining unit 210. It can be confirmed that the change of the portion that does not cut is left.

  In this example, the processing for characters is illustrated, but the same effect can be obtained even when the input image content includes graphics other than characters.

  As is apparent from the above description, the digital watermark embedding system 10 according to the present embodiment can implement the digital watermark embedding in which the image quality degradation is not noticeable with the cut-off corrected while adopting the existing digital watermark embedding algorithm.

  An embodiment will be described in which a general multi-valued image is targeted and the image quality degradation due to digital watermark embedding is further reduced when the pixel values of foreground pixels are limited in advance.

  The limitation on the pixel value of the foreground pixel is, for example, that “the foreground is composed of pixels whose pixel value is black”. Further, for example, a plurality of pixel values may be cited as candidates for the foreground pixel value, such as “the foreground is composed of pixels whose pixel values are black or red”.

  The digital watermark embedding system 10 in this embodiment has the same configuration as that of the first embodiment. That is, the digital watermark embedding system 10 in this embodiment includes a digital watermark insertion device 20 and a cut correction device 30. The difference between the present embodiment and the first embodiment resides in an image processing method for correcting a cut that occurs due to a change in pixel value by embedding a watermark.

  FIG. 7 is a flowchart showing an example of the operation of the digital watermark embedding system 10 in this embodiment. In FIG. 7, two processes of a labeling process (S201) to be performed on the watermark image and a process of determining whether or not each scanning pixel has been changed by the watermark embedding (S203) are shown in FIG. It is different from the flowchart of

  In the labeling process (S201) executed on the watermark image, the labeling unit 330 assigns labels only to pixels having the foreground pixel values. More specifically, the embodiment is different from the first embodiment in that the target pixel to be labeled is limited to pixels having one or a plurality of foreground pixel values set in advance.

  The watermark image correcting unit 340 determines, for each scanning pixel, whether the pixel is a pixel whose pixel value has been changed from the foreground pixel value to a pixel value different from the foreground pixel value by the watermark image combining unit 210 (S203). . Specifically, the subsequent determination target pixel is changed from one of the one or more foreground pixel values set in advance to a pixel value different from any of the one or more foreground pixel values. This is different from the first embodiment in that it is limited to pixels.

  The other steps are the same as those in the first embodiment, so that it is possible to correct a break that occurs with a pixel having one or more preset foreground pixel values as a foreground.

  When the foreground of an area that may be cut off due to watermark embedding can be limited in advance, as shown in this embodiment, by limiting the foreground to be cut off, the processing amount is reduced and the cut is corrected at high speed. It becomes possible. For example, when the image content 21 is a printed document image and the character whose correction should be corrected is often composed of pixels having a specific pixel value, such foreground limitation is appropriate. .

  When the foreground pixel value set in advance matches the foreground pixel value in the actual image, it is possible to correct the cut and reduce the degree of image quality degradation as in the first embodiment. When the foreground pixel value thus set does not match the pixel value of the foreground in the actual image, the cut that has occurred in the foreground is output without being corrected.

  Next, a description will be given of an embodiment in which a general multi-value image is targeted, and the degree of image quality deterioration due to digital watermark embedding is further reduced when the pixel values of the background pixels are limited in advance.

  Limiting the pixel value of the background pixel is, for example, “the background is composed of pixels whose pixel value is white”. Further, for example, a plurality of pixel values may be cited as candidates for the background pixel value, such as “the background is composed of pixels whose pixel values are white or red”.

  The digital watermark embedding system 10 in this embodiment has the same configuration as that of the first embodiment. The difference between the present embodiment and the first embodiment resides in an image processing method for correcting a cut that occurs due to a change in pixel value by embedding a watermark.

  FIG. 8 is a flowchart showing an example of the operation of the digital watermark embedding system 10 in this embodiment. In FIG. 8, two processes of a labeling process (S301) executed on the watermark image and a process of determining whether or not each scanning pixel has been changed by watermark embedding (S303) are shown in FIG. It is different from the flowchart of

  In the labeling process (S301) executed on the watermark image, the labeling unit 330 assigns labels only to pixels having pixel values other than the background pixel values. More specifically, the embodiment is different from the first embodiment in that the target pixel to be labeled is limited to a pixel having a pixel value different from any one or a plurality of preset background pixel values.

  The watermark image correcting unit 340 determines, for each scanning pixel, whether the pixel is a pixel whose pixel value has been changed from a pixel value other than the background to a background pixel value by the watermark image combining unit 210 (S303). Specifically, the subsequent pixel to be judged was changed from a pixel value different from any of the preset one or more background pixel values to one of the one or more background pixel values. This is different from the first embodiment in that it is limited to pixels.

  The other steps are the same as those in the first embodiment, so that it is possible to correct a break that occurs with a pixel having one or more preset background pixel values as a background.

  If the background of an area that may be cut off due to watermark embedding can be limited in advance, as shown in this embodiment, by limiting the background to be cut off, the processing amount is reduced and the cut is corrected at high speed. It becomes possible. For example, when the image content 21 is a printed document image and the character whose cut-off is to be corrected is often arranged with a pixel having a specific pixel value as a background, such background limitation is appropriate. It is.

  When the preset background pixel value matches the pixel value of the background in the actual image, it is possible to correct the cut and reduce the degree of image quality degradation as in the first embodiment. When the background pixel value thus set does not match the pixel value of the background in the actual image, the cut that has occurred in the area having the background is output without correction.

  In addition, the limitation regarding the pixel value of the pixel serving as the background in the present embodiment is equivalent to the limitation regarding the pixel value of the pixel serving as the foreground in the second embodiment. This means that the above two limitations are not different in the sense that all possible pixel values are classified as either foreground or background. Classifying into the foreground and other areas, and classifying into the background and other areas have a reverse relationship. In fact, the limitation that “the background is composed of pixels whose pixel values are white” is equivalent to the limitation that “the foreground is composed of pixels whose pixel values are other than white”. The difference between the second embodiment and the third embodiment is that when the limitation to one or more preset pixel values is limited to a small number of pixel values, the classification of the pixel values into the foreground and background It is only in that it is biased to.

  FIG. 9 is a conceptual diagram for explaining the difference between the cut corrected by the cut correction device 30 of the first to third embodiments and the cut not corrected.

  In FIG. 9, (a), (b), and (c) all represent image areas in which the foreground is cut from the upper right to the lower left due to the change of the pixel value by the watermark image synthesis unit 210. In this figure, the portion corresponding to the cut is represented by a dotted line. In addition, pixels displayed in the same pattern have the same pixel value, and pixels displayed in the different pattern have different pixel values. That is, the foreground pixel values in FIGS. 9A and 9B are equal, which is different from the foreground pixel value in FIG. Also, the background pixel values in FIGS. 9A and 9C are the same, which is different from the background pixel value in FIG. 9B. In FIG. 9A, the foreground pixel value is A and the background pixel value is B.

  In the cut correction process executed by the cut correction device 30 according to the first embodiment, no limitation is imposed on the pixel values of the foreground and background pixels. In this case, all the pixels shown in FIGS. 9 (a), 9 (b), and 9 (c) are corrected, although it is necessary to perform the labeling process for all the pixels and the label determination process for all changed pixels. Is done.

  In the cut correction process executed by the cut correction device 30 of the second embodiment, the pixel values of the foreground pixels are limited in advance. When “the pixel value of the foreground pixel is A” is set, the processing amount is reduced due to the limitation of the processing target pixel, and it is possible to correct the cut at high speed. a) The cut shown in (b), and the cut shown in FIG. 9 (c) is not corrected.

  In the cut correction process executed by the cut correction device 30 according to the third embodiment, the pixel values of the background pixels are limited in advance. When “the pixel value of the background pixel is B” is set, the processing amount is reduced due to the limitation of the processing target pixel, and it is possible to correct the cut at high speed. a) The cut shown in (c), and the cut shown in FIG. 9B is not corrected.

  In the cutout correction methods of the second and third embodiments, all cutouts shown in FIGS. 9A, 9B, and 9C are corrected by setting a plurality of pixel values as the foreground or background. be able to. In the cutout correction method of the second embodiment, the pixel value A and the foreground pixel value in FIG. 9C are used. In the cutout correction method of the third embodiment, the pixel value B and the background pixel value of FIG. And the background pixel value.

  Next, an embodiment will be described in which a general multi-valued image is the target, the size relationship between the brightness of the foreground pixels and the background pixels is limited, and the degree of image quality degradation due to digital watermark embedding is further reduced.

  The limitation on the brightness relationship is either “the foreground pixels are darker than the background pixels” or “the foreground pixels are brighter than the background pixels”. Yes, the foreground and background pixel values are not limited to specific values.

  In this embodiment, the brightness of a pixel is a value calculated from the pixel value of each pixel. Specifically, various values can be adopted. For example, when the pixel value is represented by a combination of the intensity of the three elements RGB of the color, a luminance value calculated as follows from the R, G, and B values may be adopted.

Luminance value = 0.29891 × R + 0.58661 × G + 0.11448 × B (Formula 1)
However, the value adopted as the brightness is not limited to the above specific example, and any value can be adopted.

  The digital watermark embedding system 10 in the present embodiment has the same configuration as in the first to third embodiments. The difference between the present embodiment and each of the above embodiments lies in an image processing method that corrects a break that occurs due to a change in pixel value due to watermark embedding.

  FIG. 10 is a flowchart showing an example of the operation of the digital watermark embedding system 10 in this embodiment. In FIG. 10, the process of determining whether or not each scanning pixel has been changed by watermark embedding (S403) is different from the flowchart of FIG.

  The watermark image correcting unit 340 determines, for each scanning pixel, whether or not the pixel has been changed by the watermark image combining unit 210 to be bright (S403). This is different from the first embodiment in that subsequent determination target pixels are limited to pixels that are changed to be brighter.

  The other steps are the same as in the first embodiment, so that it is possible to correct a break that occurs in a region where the foreground is darker than the background.

  FIG. 10 is a flowchart in a case where the restriction that “the foreground pixel is darker than the background pixel” is provided. When the restriction that “the foreground pixel is brighter than the background pixel” is set, step S403 is “determining whether or not the pixel has been changed to be dark by the watermark image composition unit 210”. By doing so, similarly, the effect of reducing the degree of image quality degradation due to the digital watermark embedding can be obtained.

  In the cutoff correction methods of the second and third embodiments, all the pixel values are associated with either the foreground or the background, whereas in the present embodiment, there is no such association. In this embodiment, it is possible to perform cut correction even when a pixel having a certain pixel value forms a foreground in one area and a background in another area. That is, when pixels having the same pixel value constitute both the foreground and the background in one image, the method of the second and third embodiments assumes that a plurality of pixel values are set as the foreground or the background. However, the cut correction method according to the present embodiment can handle all cuts as long as the relationship between the foreground and background brightness is consistent.

  The method of using the restriction on the brightness of the foreground and the background in this embodiment is determined only by selecting one of the two kinds of restrictions. Foreground or background as in the second and third embodiments. This method is more convenient than the method of enumerating the pixel value candidates, and can handle more types of cuts. In FIG. 9, if the brightness of the display pattern directly represents the brightness of the pixel, the foreground is composed of pixels darker than the background in all of FIGS. 9 (a), 9 (b), and 9 (c). Become. In this case, the cut correction processing executed by the cut correction device 30 in this embodiment is shown in FIG. 9A with a restriction that “the foreground pixels are darker than the background pixels”. (B) (c) Correction can be made for all cuts. Such a limitation is appropriate when it is known that the image content 21 is a printed document image. This is because many characters that should be corrected for cuts in a printed document image are often composed of pixels that are darker than the pixels that make up the background. However, the opposite restriction that “the foreground pixels are brighter than the background pixels” may be employed depending on the nature of the processing target image.

  FIG. 11 is a conceptual diagram illustrating the difference in image processing effect for correcting cuts between the first embodiment and the present embodiment. Here, it is assumed that the neighboring area that is a range for identifying the label is a square having a distance of 2 from the target pixel.

  FIGS. 11A and 11B both show an image area including a pixel having one of two types of pixel values. Here, the pixel values are referred to as a pixel value A and a pixel value B, and the pixel value A is darker than the pixel value B. In FIG. 11, the pixel value A and the pixel value B are represented by a black filled area and a halftone dot area, respectively.

  The difference between FIG. 11A and FIG. 11B is only the pixel values of four pixels near the center. It is assumed that either one of (a) and (b) is subjected to watermark embedding processing by the watermark image synthesizing unit 210, and changes to the other of (a) and (b). The effect of the correction process will be described.

  The change from (a) to (b) due to the watermark embedding process is interpreted as occurrence of cut when the pixel value A is considered as the foreground, but no break occurs when the pixel value B is assumed as the foreground. This is interpreted as a cluster of foreground pixels stuck together (here called fusion). On the other hand, the change from (b) to (a) is interpreted as occurrence of fusion when the pixel value A is regarded as the foreground, and is interpreted as occurrence of breakage when the pixel value B is regarded as the foreground. Thus, the interpretation of occurrence of a cut differs depending on which pixel value is regarded as the foreground despite the same image change.

  In the cut correction process according to the first embodiment, the pixel change in the vicinity of the center is corrected both in the case of the change from (a) to (b) and the change from (b) to (a) due to the watermark embedding process. That is, both cut and fusion are targeted for correction. This property is advantageous in that it can correct the cut-off regardless of which pixel value is the foreground, but it also has the disadvantage of correcting a pixel change that is not cut-out. Fragment correction is a process that cancels pixel changes due to watermark embedding, and is a process that weakens the redundancy of the embedded digital watermark. Therefore, it is desirable to minimize the correction, and it is better not to correct parts that do not affect image quality. Good.

  In the cut correction process of the fourth embodiment, when the premise that “the foreground is darker than the background” is provided, the change from (a) to (b) due to the watermark embedding process is corrected as the occurrence of a cut, No correction is made in the change from (b) to (a). As described above, the cut correction process according to the fourth embodiment has a feature that, when a premise can be appropriately set in relation to the brightness of the foreground and the background, only the pixel change that causes the cut can be accurately corrected. Convenient.

  As will be described below, the cut correction method described in each of the above embodiments and the cut correction method described in Patent Document 1 may be switched and applied for each image portion.

  The cut correction device 30 divides the inputted input image content 21 and watermark image into a plurality of partial images, determines and executes a cut correction method to be applied to each divided partial image, and performs the cut corrected portion. Combine and output images.

  In the division into partial images, the size and shape of the divided image may be determined in advance or may be determined according to the characteristics of the image to be divided.

  There are various methods for determining the cutting correction method. Hereinafter, a method for determining the cut correction method will be exemplified, but the technical scope of the present invention is not limited to these specific examples, and any method can be applied.

  The cut correction device 30 totals the pixel values of the pixels included in the partial image, and determines a cut correction method to be executed according to the total result. Specifically, among the pixel values in the partial image, a pixel value within a certain range may be set as the foreground pixel value, and the cut correction method described in the second embodiment may be executed. Alternatively, the most frequently occurring pixel value in the partial image may be set as the background pixel value, and the cutoff correction method described in the third embodiment may be executed. Alternatively, when the brightness distribution of the pixel values in the partial image is biased toward the bright side, the cutoff correction method described in the fourth embodiment is executed by assuming that “the foreground pixels are darker than the background pixels”. May be. Alternatively, when there is only one type of pixel value of a pixel included in the partial image, the cut correction process may not be performed on the partial area.

  As another method for determining the cutout correction method, for example, the determination of the user who is presented with the partial image of the image content 21 via the user interface 430 may be used. The user presented with the partial image determines the background pixel value of the partial image, and sets the background pixel value via the user interface 430. The cut correction device 30 executes the cut correction method described in the third embodiment based on the set background pixel value.

  The technical scope of the present invention is not limited to the scope described in each of the above embodiments. For example, the image content 21 in the present embodiment may be a still image or a moving image.

  Further, the input image, watermark image, and corrected image used in each of the above embodiments may be a part of the image content 21 and its processing result, that is, a partial area in one screen or one frame, or at least of a plurality of frames. One frame may be used.

1 is a block diagram illustrating a configuration of a digital watermark embedding system 10. FIG. 2 is a diagram illustrating an example of a hardware configuration of a digital watermark embedding system 10. FIG. It is a conceptual diagram which illustrates the outline | summary of the labeling process which the labeling part 330 performs. 4 is a flowchart illustrating an operation of the digital watermark embedding system 10. It is a conceptual diagram explaining the effect of the image correction process of the watermark image correction unit 340. It is a conceptual diagram explaining the character external appearance change in the process of the digital watermark embedding process using the digital watermark embedding system. 10 is a flowchart illustrating an operation of the digital watermark embedding system 10 according to the second embodiment. 10 is a flowchart illustrating an operation of the digital watermark embedding system 10 according to the third embodiment. It is a conceptual diagram explaining the difference between the cut corrected by the cut correction device 30 of the first to third embodiments and the cut not corrected. 10 is a flowchart illustrating an operation of the digital watermark embedding system 10 according to the fourth embodiment. It is a conceptual diagram explaining the difference of the effect of the image processing which corrects a cutting | disconnection of Example 1 and Example 4. FIG.

  10: digital watermark embedding system, 20: digital watermark insertion device, 21: image content, 22: watermark image content, 200: input unit, 210: watermark image synthesis unit, 220: output unit, 30: cut correction device, 31: Correction parameter, 300: correction parameter input unit, 310: input image input unit, 320: watermark image input unit, 330: labeling unit, 340: watermark image correction unit, 350: corrected image output unit, 40: computer, 400: CPU , 410: RAM, 420: storage, 430: user interface, 440: communication interface, 450: media interface, 460: external device or network, 470: recording medium, 500, 510, 520: image data, 600, 610: Image, 612: B pixel, 614, 6 6: A pixel block, 618: B pixel block, 620: conceptual diagram, 622: attention pixel, 624: neighboring region, 630: conceptual diagram, 632: attention pixel, 634: neighboring region, 640: image, 700, 710, 720: Outline drawing of character “y”, 721, 722, 723, 724, 725, 726: Pixel block, 730: Outline drawing of character “y”.

Claims (10)

An image processing method for correcting cuts that occur due to pixel value changes in information embedding processing using digital watermarks,
About the watermark image that is the result of the information embedding process for the input image,
A pixel in which any adjacent pixel has a pixel value that has a pixel value greater than a preset threshold value, or a pixel that is adjacent to each other and whose difference between pixel values is not greater than the threshold value When a set is composed of a plurality of pixels having values, each pixel belonging to the same set is assigned the same label value but different from the pixels belonging to the other set. A labeling step for outputting a label image identifiable by the label value;
A set pixel value setting step for setting one or a plurality of set pixel values representing pixel values of pixels belonging to the set for each of the sets;
A set representing a pixel value before the information embedding process of the target pixel included in a preset neighborhood area in the label image for a pixel whose pixel value has been changed by the information embedding process for the input image A label determination processing step for identifying a set having pixel values;
In the label determination process, when there are two or more identified sets, a change cancellation process for performing a change cancellation process for returning the pixel value of the target pixel of the watermark image to the pixel value before the information embedding process Steps,
And a corrected image output step of outputting a corrected image subjected to the change canceling process. The image processing method according to claim 1,
The labeling step includes
An image processing method comprising: a pixel value quantization step for quantizing a pixel value of the watermark image with one or a plurality of preset threshold values. The image processing method according to claim 1 or 2,
The set pixel value is one or a plurality of pixel values included in a pixel value of a pixel belonging to one set, or the most frequently used pixel value among the plurality of pixel values, or An image processing method characterized in that the pixel values are within a certain range including an average value of the plurality of pixel values. An image processing method according to any one of claims 1 to 3, wherein
The pixel targeted by the labeling step is a pixel having either one or a plurality of foreground pixel values set in advance,
The target pixel of the label determination processing step is a pixel value of any one of the one or more foreground pixel values from the one or more foreground pixel values by the information embedding process for the input image. An image processing method characterized in that the pixels are changed to different pixel values. An image processing method according to any one of claims 1 to 4,
The pixel targeted by the label determination processing step is a pixel whose pixel value has been changed by the information embedding processing for the input image so as to follow a preset brightness increase / decrease. . In an information embedding process using a digital watermark, an image processing apparatus that corrects a break that occurs due to a modification of a pixel value,
An input image input unit for receiving an input image to be the information embedding process;
A watermark image input unit for receiving a watermark image as an output of the information embedding process;
For the watermark image, any adjacent pixel has a pixel value having a pixel value greater than a preset threshold value, or is adjacent to each other, and the difference between pixel values is the threshold value. When one set is composed of a plurality of pixels having pixel values not larger than the same, each pixel belonging to the same set is assigned the same label value but different from the pixels belonging to the other set. A labeling unit that outputs a label image that can identify a set to which a pixel belongs by the label value, and sets one or a plurality of set pixel values representing pixel values of pixels belonging to the set to each of the sets; ,
The input image is input from the input image input unit, the watermark image is input from the watermark image input unit, and the label image is received from the labeling unit. Pixels whose pixel values are changed by the information embedding process for the input image are received. As a target, a set having a set pixel value representing a pixel value before the information embedding processing of the target pixel included in a preset neighboring region in the label image is identified, and two or more of the identified sets are identified. A watermark image correction unit that performs a change cancellation process for returning the pixel value of the target pixel of the watermark image to the pixel value before the information embedding process;
An image processing apparatus comprising: a correction image output unit that receives a result of change cancellation processing by the watermark image correction unit and outputs the result as a corrected image. The image processing apparatus according to claim 6, wherein
The labeling part is
An image processing apparatus comprising: a pixel value quantization unit that quantizes pixel values of the watermark image using one or more preset threshold values. The image processing apparatus according to claim 6 or 7,
The set pixel value is one or a plurality of pixel values included in a pixel value of a pixel belonging to one set, or the most frequently used pixel value among the plurality of pixel values, or An image processing apparatus comprising pixel values within a certain range including an average value of the plurality of pixel values. An image processing apparatus according to any one of claims 6 to 8, comprising:
The pixel targeted by the labeling unit is a pixel having either one or a plurality of foreground pixel values set in advance,
The pixel targeted by the watermark image correction unit is a pixel value of any one of the one or more foreground pixel values from the one or more foreground pixel values by the information embedding process for the input image. An image processing apparatus characterized in that the pixel is changed to a different pixel value. An image processing apparatus according to any one of claims 6 to 9,
The pixel targeted by the watermark image correction unit is a pixel whose pixel value has been changed by the information embedding process with respect to the input image so as to follow a preset increase or decrease in brightness. .

Images