JP2006086822A - Electronic watermark embedding apparatus and method thereof, and electronic watermark extracting apparatus and method thereof - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve the problem that photographing a printed image having an electronic watermark embedded therein causes distortion, resulting in the difficulty in correctly detecting the electronic watermark. <P>SOLUTION: A photographing unit 30 photographs an image obtained by sticking a frame around the periphery of an original image having an electronic watermark embedded therein and printing it. A lens distortion correcting unit 32 corrects distortion generated in the photographed image. A feature point extracting unit 34 extracts a plurality of feature points arranged with a predetermined interval in the frame of the photographed image after the lens distortion correction. An image shape/size determining unit 36 determines the shape and the actual size of an original image region on the basis of the arrangement of the plurality of feature points. A transparent distortion correction unit 38 uses a feature point template matching the shape and actual size of the original image region to detect positional deviation of each of the feature points in the frame of the photographed image, thereby correcting the transparent distortion generated in the original image region. A watermark extracting unit 40 extracts watermark information X from the corrected original image region. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a digital watermark technique, and more particularly to a digital watermark embedding apparatus and method for a printed image and a digital watermark extraction apparatus and method.

  There is a system for printing a digital image in which a digital watermark is embedded on a printing medium, photographing the printed image with a digital camera, a scanner, or the like and digitizing it again to detect the embedded digital watermark. For example, when a ticket or card is issued to a user, identification information about the issuer or the user is embedded in an image so that it cannot be visually detected as a digital watermark and printed on the ticket or card. By detecting the digital watermark when using a ticket or card, it is possible to prevent fraudulent acts such as forgery and unauthorized acquisition. In addition, when an image is printed with a copier or printer, it is possible to prevent unauthorized copying of copyrighted works, securities, etc. by embedding copyright information and device identification numbers as digital watermarks. .

  Generally, when a printed image is photographed and digitized using a digital camera or scanner, the photographed image can be seen through due to lens distortion that depends on the lens shape and focal length of the photographing device and the tilt of the optical axis at the time of photographing. Distortion occurs, and a pixel shift appears between the printed image and the captured image. Therefore, it is difficult to correctly extract the digital watermark embedded in the print image from the photographed image, and it is necessary to correct the distortion of the photographed image.

Patent Document 1 discloses an image processing apparatus that calculates the size of original image information from image information including distortion generated by optically reading a printed image and restores additional information.
Japanese Patent Laid-Open No. 2003-110846

  In the image processing apparatus of Patent Document 1, additional information is extracted from a captured image on the assumption that the shape of the image is a rectangle. However, an image of a product used for a printed matter such as a catalog for mail order is not limited to a quadrangle, and has various shapes such as a circle and a triangle, and the size of the image varies. When a watermark is embedded in such various types of images, the watermark cannot be extracted unless the shape and size of the image are specified. In the case of a printed image, it is necessary for the user to be able to distinguish between an image with a watermark and an image without a watermark. In addition, in order to correct image distortion due to shooting, information on distortion characteristics of the shooting device, information on the tilt of the optical axis at the time of shooting, and information on rotation around the optical axis are obtained, and the shot image is subjected to geometric conversion. There is a need.

  The present invention has been made in view of such a situation, and an object thereof is to embed a digital watermark embedding technique for generating a watermarked print image capable of distortion correction, and to correct distortion generated when a watermarked print image is photographed. An object of the present invention is to provide a digital watermark extraction technique capable of detecting a watermark.

  In order to solve the above problems, a digital watermark embedding device according to an aspect of the present invention matches a frame for arranging a plurality of feature points at regular intervals along the contour of an image without depending on the image size. A plurality of types of stored frame databases, a frame pasting unit that selects the frame that matches the shape of an original image to be embedded with a digital watermark from the frame database, and pastes the frame around the original image; A watermark embedding unit that embeds a watermark; and an output unit that outputs the original image in which the digital watermark is embedded and the plurality of feature points are arranged along an outline of the image on a print medium.

  The “shape of the original image” refers to the shape of the contour of the original image when the original image is formed in a predetermined shape such as a polygon or a circle. The “actual size of the original image” refers to the actual size of the original image, and is distinguished from the size of the captured image when the original image is captured.

  Another aspect of the present invention is a digital watermark extraction apparatus. The apparatus includes an image input unit that acquires a print image obtained by printing an original image in which a digital watermark is embedded as a digital image, and a constant interval that does not depend on the image size along the outline of the original image area of the digital image. For each of the plurality of feature points, a feature point extraction unit that extracts the plurality of feature points that has been performed, a determination unit that determines the shape and actual size of the original image based on the arrangement of the plurality of feature points, A perspective distortion correction unit that corrects a perspective distortion generated in the original image area by detecting a deviation from a desired position, and the original image corrected according to the determined shape and actual size of the original image A watermark extracting unit that extracts the digital watermark from the region.

  Yet another embodiment of the present invention is a digital watermark embedding method. This method is the same as the original image in which a plurality of feature points are arranged at regular intervals along the contour of the original image when an original image of a predetermined shape is given as an electronic watermark embedding target. A frame having a shape is pasted around the original image, an electronic watermark is embedded in the original image, and output for printing on a print medium.

  Yet another embodiment of the present invention is a digital watermark extraction method. In this method, a print image obtained by printing an original image in which a digital watermark is embedded is obtained as a digital image, and a plurality of images are arranged at regular intervals along the outline of the original image area of the digital image without depending on the image size. By extracting feature points and detecting deviations from the original positions of these feature points, the perspective distortion generated in the original image area is corrected and determined based on the arrangement of the plurality of feature points. The digital watermark is extracted from the corrected original image area according to the shape and actual size of the original image.

  It should be noted that any combination of the above-described constituent elements and a conversion of the expression of the present invention between a method, an apparatus, a system, a recording medium, a computer program, and the like are also effective as an aspect of the present invention.

  According to the present invention, a watermarked print image having various shapes can be generated, and distortion generated in an image obtained by photographing the watermarked print image is corrected to detect the watermark with high accuracy. Can do.

Embodiment 1
The digital watermark system according to Embodiment 1 of the present invention includes the digital watermark embedding apparatus 100 in FIG. 1 and the digital watermark extraction apparatus 200 in FIG. 7, and a print image in which the digital watermark is embedded by the digital watermark embedding apparatus 100. The digital watermark that has been generated is photographed by the digital watermark extraction apparatus 200, and the embedded digital watermark is extracted. The digital watermark embedding device 100 is used for, for example, issuing a watermarked ticket or card, generating a watermarked product image to be posted in a catalog such as mail order, and the digital watermark extracting device 200 detects a watermark. This is used to discover counterfeit tickets and cards, or to extract product information from product images posted in catalogs. Either device may be configured as a server accessed from a terminal on the network.

  FIG. 1 is a configuration diagram of a digital watermark embedding apparatus 100 according to the first embodiment. These configurations can be realized in hardware by any computer's CPU, memory, and other LSI, and in software, by programs with an image processing function and digital watermark embedding function loaded in the memory. However, here, functional blocks realized by their cooperation are depicted. Therefore, those skilled in the art will understand that these functional blocks can be realized in various forms by hardware only, software only, or a combination thereof.

  The image forming unit 10 changes the shape and size of the input digital image I, and generates an image suitable for the resolution at the time of printing. An image deformed by the image forming unit 10 is referred to as an original image. The original image is clearly marked as a watermarked image, and a frame is pasted along the outline of the original image in order to correct distortion at the time of shooting. Therefore, there are certain restrictions on the shape and size of the original image. For example, frames having a predetermined shape such as a rectangle, a circle, and a triangle are prepared in advance, and the image forming unit 10 shapes the original image into a shape that matches the shape of the prepared frame. In addition, as will be described later, the size of the original image is limited to an integral multiple of the interval between a plurality of feature points arranged in the frame, and the image forming unit 10 sets the size of the original image according to the limitation. adjust. The image forming unit 10 may include a user interface for cropping an image in accordance with restrictions on the shape and size of a frame prepared in advance.

  The image forming unit 10 provides the watermark embedding unit 12 with the original image transformed in accordance with the frame in this way. In addition, the image forming unit 10 gives information on the shape and size of the original image to the watermark embedding unit 12 and the frame pasting unit 16.

  The watermark embedding unit 12 embeds the watermark information X in the original image given from the image forming unit 10 with a block size suitable for the shape and size of the original image. For example, when the original image is a rectangle, the watermark embedding unit 12 divides the original image into blocks of an appropriate size and embeds the watermark information X for each block.

  The frame pasting unit 16 reads a frame matching the shape and size of the original image given from the image forming unit 10 from the frame database 18, pastes the frame around the original image, and creates an outline of the original image in the frame. A plurality of feature points are arranged along the predetermined intervals. A plurality of feature points in this frame are used to correct perspective distortion when extracting a watermark.

  The printing unit 14 prints the original image in which the watermark information X is embedded by the watermark embedding unit 12 and the frame is pasted by the frame pasting unit 16 on a printing medium such as paper or a card, thereby generating a framed print image P. . In the figure, the printing unit 14 is a component of the digital watermark embedding device 100. However, the printing unit 14 may be provided outside the digital watermark embedding device 100 and configured by a printer. The embedding device 100 and the printer are connected by a peripheral device connection cable or a network.

  FIG. 2 is a diagram for explaining the framed print image P output from the digital watermark extracting apparatus 200. FIG. An original image in which a digital watermark is embedded is printed on a print medium, and a frame 24 including a plurality of feature points 22 around an area 20 (hereinafter referred to as an original image area 20) where the original image is printed. Is printed. The frame 24 is provided on the outer periphery of the original image area 20 with a certain width, and may be white or colored in some color. The feature point 22 in the frame 24 is a mark having a size that can be detected when photographing with a camera, and is shown here as a circle, but may be another shape such as a square or a cross. The belt-like frame 24 is not arranged around the original image area 20, but the original image area 20 is overlaid on the rectangular frame 24 so that the framed print image P as shown in FIG. May be generated.

  Here, the interval L between adjacent feature points 22 is a fixed value set in advance, and does not depend on the size of the image, and the same interval L is applied to all images. If the interval L between the feature points 22 is set to be small, the number of feature points 22 increases and the accuracy of distortion correction increases. However, since this is annoying for the user, the interval L between the feature points 22 depends on the accuracy and design of the distortion correction. It is more preferable to set in consideration of both aspects.

  In the example of FIG. 2, seven feature points 22 are arranged at equal intervals L in the horizontal direction of the original image region 20, and five feature points 22 are arranged at equal intervals L in the vertical direction. Therefore, it can be seen that the original image region 20 has a horizontal length of 6L and a vertical length of 4L.

  When the original image area 20 on the print medium is photographed, the size of the photographed image changes depending on the zoom magnification at the time of photographing. However, since the interval L between the feature points 22 is a constant value regardless of the actual size of the original image, the actual size of the original image region 20 is obtained by counting the number of feature points 22 in the captured original image region 20. Can be requested.

  Even if the original image area 20 on the print medium is rectangular, the photographed image may be trapezoidal, for example, due to perspective distortion when photographed from an oblique direction, but whether the original image area 20 of the trapezoid was originally photographed or due to perspective distortion. Since it is impossible to determine whether the image has become trapezoidal, the actual size of each side of the original image region 20 is generally not known from the captured image. However, even in this case, the actual size of each side can be obtained by counting the number of feature points 22 on each side of the captured original image region 20, and whether the shape of the original image region 20 is a rectangle. Whether it is a trapezoid can be determined.

  To be precise, the actual size of the original image area 20 is the length of the original image area 20 in the horizontal direction in the example of FIG. 2 in consideration of the distance d from the feature point 22 to the edge of the original image area 20. Is calculated as 6L-2d and the length in the vertical direction is 4L-2d. However, in the following, for the sake of simplicity, the distance d from the feature point 22 to the edge of the original image area 20 will not be mentioned. .

  FIGS. 3A and 3B are diagrams for explaining another example of the framed print image P. FIG. In FIG. 3A, five feature points 22 in the horizontal direction of the original image area 20 and four feature points 22 in the vertical direction are arranged at equal intervals L, and the length of the original image area 20 in the horizontal direction is shown. It can be seen that the length is 4L and the length in the vertical direction is 3L. In FIG. 3B, four feature points 22 in the horizontal direction of the original image area 20 and five feature points 22 in the vertical direction are arranged at equal intervals L, and the length of the original image area 20 in the horizontal direction is shown. It can be seen that the length is 3L and the length in the vertical direction is 4L. Even when the framed print image P is photographed, the actual size of the original image region 20 can be determined by counting the number of feature points 22 in the frame 24 pasted on the original image region 20.

  4A and 4B are diagrams for explaining still another example of the framed print image P. FIG. In this example, the original image region 20 is circular, and the frame 24 attached to the outer periphery of the original image region 20 is also circular. 4A and 4B, the size of the original image region 20 is different, but the arc length L indicated by the dotted line formed by the two adjacent feature points 22 is common. Therefore, by counting the number of feature points 22 in the frame 24, the circumference length of the original image area 20 can be obtained. In the framed print image P of FIG. 4A, since there are 24 feature points 22 in the frame 24, it can be seen that the circumference of the original image area 20 is 24L. On the other hand, in the framed print image P of FIG. 4B, since there are 16 feature points 22 in the frame 24, it can be seen that the circumference of the original image area 20 is 16L.

  Even if the framed print image P of FIG. 4B is zoomed and photographed at the time of photographing, and the photographed image is larger than the framed print image P of FIG. 4A, the number of feature points 22 in the frame 24 Therefore, regardless of the zoom magnification, the actual size of the original image area 20 can be obtained based on the number of feature points 22 in the frame 24 pasted to the original image area 20.

  FIGS. 5A and 5B are diagrams illustrating still another example of the framed print image P. FIG. In this example, the original image area 20 is an equilateral triangle, and the frame 24 attached to the outer periphery of the original image area 20 is also an equilateral triangle. 5A and 5B, the size of the original image region 20 is different, but the interval L between two adjacent feature points 22 is the same, and the number of feature points 22 in the frame 24 is counted. The length of each side of the original image area 20 can be obtained. In the framed print image P in FIG. 5A, since there are six feature points 22 in each frame of each equilateral triangle, it can be seen that the length of each side is 5L. On the other hand, in the printed image P with a frame in FIG. 5B, since there are five feature points 22 in each frame of the equilateral triangle, it can be seen that the length of each side is 4L.

  In the frame database 18, frames 24 of various shapes such as arbitrary polygons such as rectangles and triangles and circles are registered in advance, and the frame pasting unit 16 stores frames 24 matching the shape of the original image in the frame database. A selection is made from 18 and pasted around the original image, and a plurality of feature points 22 are arranged in the frame 24 at predetermined intervals L along the outline of the original image. If the actual size of the original image is not an integral multiple of L, the feature points 22 cannot be arranged at equal intervals around the original image. In this case, the image forming unit 10 does not provide a part of the original image. Or the original image is reduced or enlarged so that the actual size of the original image becomes an integral multiple of the predetermined interval L.

  FIG. 6 is a flowchart for explaining a digital watermark embedding procedure by the digital watermark embedding apparatus 100.

  The image forming unit 10 performs a deformation process on the digital image I (S10). The digital image I is cropped into a shape such as a rectangle, a circle, or a triangle in accordance with the shape of the frame 24 registered in the frame database 18.

  The image forming unit 10 checks whether the size of the original image after the deformation process matches the size of the frame 24 (S12). For example, when the shape of the original image is a rectangle, it is checked whether the vertical and horizontal sizes are an integral multiple of the interval L between the feature points 22 in the frame 24. When the shape of the original image is circular, it is checked whether or not the length of the circumference is an integral multiple of the length L of the arc formed by two adjacent feature points 22 in the frame 24. When the shape of the original image is a triangle, it is checked whether the length of each side is an integral multiple of the interval L between the feature points 22 in the frame 24.

  When the original image does not match the size of the frame 24 (N in S12), the image forming unit 10 adjusts the image size to match the size of the frame 24 (S14). The adjustment of the image size is performed by, for example, enlarging or reducing the original image or cutting out the periphery of the original image. When the original image matches the size of the frame 24 (Y in S12), the image size adjustment process is not necessary.

  The watermark embedding unit 12 embeds the watermark information X in the original image transformed by the image forming unit 10 (S16). The frame pasting unit 16 reads a frame 24 matching the shape and size of the original image from the frame database 18 and arranges a plurality of feature points 22 at predetermined intervals along the outline of the original image in which the watermark information X is embedded. The frame 24 is pasted (S18). The printing unit 14 prints the original image with the watermark information X embedded by the watermark embedding unit 12 and the frame 24 pasted by the frame pasting unit 16 on a print medium, and generates a framed print image P (S20).

  In the above procedure, the order of embedding the watermark information X by the watermark embedding unit 12 and the pasting of the frame 24 by the frame pasting unit 16 may be reversed. That is, after the frame pasting unit 16 pastes the frame 24 to the original image, the procedure may be changed so that the watermark embedding unit 12 embeds the watermark information X in the original image.

  FIG. 7 is a configuration diagram of the digital watermark extracting apparatus 200 according to the first embodiment. These configurations can also be realized in various forms by any combination of hardware such as a CPU and memory, and software having an image processing function and a digital watermark extraction function.

  The photographing unit 30 photographs the framed print image P generated by the digital watermark embedding apparatus 100 and digitizes the framed print image P. In the figure, the photographing unit 30 is a component of the digital watermark extracting apparatus 200. However, the photographing unit 30 may be provided outside the digital watermark extracting apparatus 200 and configured by a digital camera or a scanner. The watermark extraction apparatus 200 and the digital camera or scanner are connected by a connection cable or network of peripheral devices. In particular, when the digital camera has a wireless communication function, a captured image captured by the digital camera is transmitted to the digital watermark extraction apparatus 200 wirelessly.

  When the framed print image P is captured by the photographing unit 30, lens distortion and perspective distortion occur in the photographed image. Since the digital watermark is not robust against such geometric distortion, the digital watermark embedded in the original image can be accurately extracted unless the geometric distortion generated in the photographed image is corrected. I can't. As will be described below, the geometric correction of the captured image is performed in two stages of lens distortion correction and perspective distortion correction.

Assuming that the coordinates of the point in the image obtained by photographing the framed print image P are (X _d , Y _d ), the lens distortion correction unit 32 uses the following lens distortion correction function f to convert the point (X By mapping _{d 1} , Y _d ) to the point (X _c , Y _c ) after the lens distortion correction, the lens distortion generated in the captured image is corrected.

(X _c , Y _c ) = f (X _d , Y _d )

  The lens distortion correction function f is a function for mapping a point in a captured image in which lens distortion has occurred to a point in an image without lens distortion, and is specifically represented by the following polynomial.

^{_{X c = a 0 X d 2}} + a 1 X d Y d + a 2 Y d 2 + a 3 X d + a 4 Y d + a 5
^{_{Y c = a 6 X d 2}} + a 7 X d Y d + a 8 Y d 2 + a 9 X d + a 10 Y d + a 11

Here, parameters a ₀ , a ₁ ,..., A ₁₁ are lens distortion coefficients. The lens distortion coefficient can be obtained by photographing a calibration image such as a lattice pattern image in advance. The lens distortion depends on the type of lens used for photographing and the focal length. Note that the focal length changes depending on the zoom magnification when the zoom lens is used for photographing. In order to measure lens distortion, a grid pattern image for calibration is photographed by the photographing unit 30, and a positional shift due to lens distortion of lattice points of the grid pattern image is detected. Based on the positional deviation of the lattice points, a lens distortion coefficient of the lens distortion function f is obtained. When shooting with a zoom lens, change the zoom magnification and shoot a grid pattern image for calibration at different focal lengths to detect misalignment of the grid points, and distort the lens for each focal length. Find the coefficient.

More specifically, the coordinate values (X _dk , Y _dk ) (k = 0, 1,..., N−1) of N lattice points on the image obtained by capturing the lattice pattern image and the original lattice pattern The true coordinate values (X _ck , Y _ck ) (k = 0, 1,..., N−1) of the N lattice points on the image are associated one-to-one and applied to the lens distortion correction function f. Lens distortion coefficients a ₀ , a ₁ ,..., A ₁₁ are obtained by the following least square method.

J = Σ _{k = 0} ^N−1 [(X _ck − (a ₀ X _dk ² + a ₁ X _dk Y _dk + a ₂ Y _dk ² + a ₃ X _dk + a ₄ Y _dk + a ₅ )) ² + (Y _Ck − ( ^{_{a 6 X dk 2 + a 7}} X dk Y dk + a 8 Y dk 2 + a 9 X dk + a 10 Y dk + a 11)) 2] → min

By solving 式 J / ∂a ₀ = 0, ∂J / ∂a ₁ = 0,..., ∂J / ∂a ₁₁ = 0, the lens distortion coefficients a ₀ and a ₁ that minimize J are obtained. , _..., it can be obtained _{a 11.}

  A database in which the lens distortion coefficient is associated with the lens type and the focal length is provided, and the lens distortion correction unit 32 reads out the lens distortion coefficient corresponding to the lens type and the focal distance at the time of shooting from the database and refers to them. If the lens type is determined by the camera model, it may be registered in the database using the camera model information instead of the lens type. The model information and focal length of the camera can be acquired from information added to the header of the photographed digital image. For example, when the photographed image is given in EXIF (Exchangeable Image File Format), the camera model information and the focal length at the time of photographing can be acquired from the EXIF information included in the image data.

  The feature point extraction unit 34 extracts a plurality of feature points 22 in the frame 24 from the captured image whose lens distortion has been corrected by the lens distortion correction unit 32. Since the feature point 22 is arranged in the frame 24 around the original image area 20, it can be easily extracted with high accuracy. Further, the lens distortion tends to increase as the distance from the center of the photographed image increases, and the lens distortion may be the largest in the vicinity of the frame 24 around the original image area 20, but the lens distortion correction unit 32. Thus, the feature point 22 is extracted from the photographed image after the lens distortion is corrected, so that the feature point 22 can be extracted without being affected by the lens distortion.

  The feature point extraction unit 34 provides the position information of the extracted feature points 22 to the image shape / size determination unit 36 and the perspective distortion correction unit 38.

  The image shape / size determination unit 36 determines the shape and actual size of the original image region 20 of the photographed image based on the arrangement of the plurality of feature points 22 extracted by the feature point extraction unit 34. For example, if the arrangement of the feature points 22 is a rectangle, it can be estimated that the shape of the original image area 20 is at least a rectangle. Since it is known that the interval between adjacent feature points 22 is a fixed value L that does not depend on the actual size of the original image, the actual size of the original image region 20 is determined based on the number of extracted feature points 22. Can do. For example, when the shape of the original image area 20 is estimated to be a quadrangle, the number of feature points 22 arranged along each side of the original image area 20 is counted to determine the actual number of each side of the original image area 20. The size can be determined, and if the actual size of two opposite sides of the four sides is equal, the shape of the original image area 20 can be determined to be a rectangle. If the actual size of the four sides is equal, It can be determined that the shape of the image region 20 is a square.

  The image shape / size determination unit 36 provides the perspective distortion correction unit 38 with the determined shape and actual size information of the original image region 20.

  The perspective distortion correction unit 38 reads the feature point template 54 that matches the shape and actual size of the original image region 20 from the feature point template storage unit 50. The feature point template storage unit 50 stores the frame 24 registered in the frame database 18 used by the digital watermark embedding apparatus 100 of FIG. The feature point template 54 has position information of the reference feature point 52 as information indicating the position coordinates of the feature point 22 to be originally arranged in the frame 24.

  When a new frame 24 is registered in the frame database 18 referred to by the digital watermark embedding apparatus 100 in FIG. 1, the feature point template 54 of the feature point template storage unit 50 referred to by the digital watermark extraction apparatus 200 is also updated accordingly. It is configured as follows. Alternatively, the digital watermark extracting apparatus 200 may be configured to directly refer to the frame database 18 of the digital watermark embedding apparatus 100 via the network and use the frame 24 read from the frame database 18 as the feature point template 54. Good.

  The perspective distortion correction unit 38 uses the coordinate values of the reference feature points 52 in the feature point template 54 read from the feature point template storage unit 50 and the frame 24 extracted from the photographed image by the feature point extraction unit 34. The next perspective distortion correction function g is obtained by associating the coordinate values of the plurality of feature points 22 on a one-to-one basis.

(X _u , Y _u ) = g (X _c , Y _c )

Here, (X _c , Y _c ) is the coordinate value of the point on the projection plane by the camera, and (X _u , Y _u ) is on the original image plane and corresponds to the point (X _c , Y _c ). The coordinate value of the point. The perspective distortion correction function g is a function for mapping a point in the captured image in which the perspective distortion has occurred to a point in the original image having no perspective distortion, and is specifically expressed by the following equation.

_{X u = (b 0 X c} + b 1 Y c + b 2) / (b 6 X c + b 7 Y c +1)
_{Y u = (b 3 X c} + b 4 Y c + b 5) / (b 6 X c + b 7 Y c +1)

The eight parameters b ₀ , b ₁ ,..., B ₇ of the perspective distortion correction function g are perspective distortion coefficients. The perspective distortion coefficient can be obtained in principle by giving the coordinate values on the projection plane and the coordinate values on the original image plane for the four corner points of the original image area 20. By using a plurality of feature points 22 arranged around, the accuracy of the perspective distortion coefficient to be obtained can be increased.

The perspective distortion correction unit 38 stores the coordinate values (X _ck , Y _ck ) (k = 0, 1,..., M−1) of the M feature points 22 in the frame 24 on the captured image and the feature point template storage. The coordinate values (X _uk , Y _uk ) (k = 0, 1,..., M−1) of the M reference feature points 52 in the feature point template 54 read from the unit 50 are associated one-to-one. And the perspective distortion coefficients b ₀ , b ₁ ,..., B ₇ are obtained by the following least square method.

J = Σ _{k = 0} ^M−1 [(X _u (b ₆ X _c + b ₇ Y _c +1) − (b ₀ X _c + b ₁ Y _c + b ₂ )) ² + (Y _u (b ₆ X _c + b ₇ Y _c +1)-(b ₃ X _c + b ₄ Y _c + b ₅ )) ² ] → min

By solving 式 J / ∂b ₀ = 0, ∂J / ∂b ₁ = 0,..., ∂J / ∂b ₇ = 0, the perspective distortion coefficient b ₀ , b ₁ that minimizes J is obtained. ,..., B ₇ can be obtained.

  The perspective distortion correction unit 38 corrects the perspective distortion of the original image area 20 by mapping each point in the original image area 20 using the calculated perspective distortion correction function g, and the watermark extraction unit extracts the corrected original image area 20. Give to 40. The watermark extraction unit 40 identifies the original image area 20 according to the determined shape and actual size of the original image area 20, and based on the block size suitable for the shape and actual size of the original image area 20, the corrected original image area 20 is identified. The watermark information X is extracted from the image area 20 for each block and output.

  When the watermark bit is detected, if the original image area 20 is distorted, it is difficult to detect the watermark. However, since the geometric distortion is corrected by the lens distortion correction unit 32 and the perspective distortion correction unit 38, Watermark detection accuracy is guaranteed. In addition, since the shape and actual size of the original image area 20 can be determined from the arrangement of the plurality of feature points 22 in the frame 24, the original image area 20 is accurately cut out from the captured image, and the watermark is extracted from the original image area 20. Information X can be extracted.

FIGS. 8A to 8C are diagrams illustrating how the geometric distortion generated in the photographed image is corrected by the lens distortion correction unit 32 and the perspective distortion correction unit 38. FIG. FIG. 8A shows a captured image in which lens distortion and perspective distortion occur, and a point (X _d , Y _d ) in the image. FIG. 8B shows an image in which lens distortion is corrected by the lens distortion correction unit 32 and a point (X _c , Y _c ) in the image. A point (X _d , Y _d ) in the photographed image before correction shown in FIG. 8A is converted into a point (( _{D d} , Y _d ) in the image after lens distortion correction shown in FIG. _Xc , _Yc ). Figure 8 (c) shows further point in the image and image perspective distortion is corrected by perspective distortion correcting unit 38 (X _{u, Y u)} a. A point (X _c , Y _c ) in the image after the lens distortion correction shown in FIG. 8B is a point in the image after the perspective distortion correction shown in FIG. 8C by the perspective distortion correction function g. _{Maps to} (X _u , Y _u ).

  FIGS. 9A and 9B are diagrams for explaining how perspective distortion is corrected based on a plurality of feature points 22 extracted from a captured image. As shown in FIG. 9A, the original image area 20 is geometrically distorted due to perspective distortion, and the coordinate positions of a plurality of feature points 22 arranged around the original image area 20 are shifted accordingly. ing. FIG. 9B shows the positions of the reference feature points 52 of the original image area 20 and the feature point template 54 whose perspective distortion has been corrected by the perspective distortion correction function. The perspective distortion correction function g maps each feature point 22 in FIG. 9A to each corresponding reference feature point 52 in FIG. 9B, and this perspective distortion correction function g causes the FIG. The original image area 20 in a) is converted to the original image area 20 without perspective distortion in FIG. 9B, and the perspective distortion is corrected.

  FIG. 10 is a flowchart for explaining a digital watermark extraction procedure by the digital watermark extraction apparatus 200.

  The imaging unit 30 captures the framed print image P (S30). The lens distortion correction unit 32 corrects lens distortion generated in the captured image (S32). The feature point extraction unit 34 extracts a plurality of feature points 22 from the frame 24 of the captured image in which the lens distortion is corrected (S34). The image shape / size determination unit 36 determines the shape and actual size of the original image area 20 from the arrangement of the extracted feature points 22 (S36).

  The perspective distortion correction unit 38 reads the feature point template 54 that matches the shape and actual size of the original image area 20 from the feature point template storage unit 50 (S38). The perspective distortion correction unit 38 associates each reference feature point 52 in the feature point template 54 with each feature point 22 in the frame 24 of the photographed image on a one-to-one basis, and applies it to the perspective distortion correction function g. A coefficient is calculated, and the perspective distortion generated in the original image area 20 is corrected by the calculated perspective distortion correction function g (S40). The watermark extraction unit 40 extracts the watermark information X from the original image area 20 in which the lens distortion and the perspective distortion are corrected (S42).

  As described above, the digital watermark embedding apparatus 100 according to the present embodiment pastes the frame 24 that matches the shape of the original image in which the digital watermark is embedded, and includes a plurality of feature points 22 for geometric correction in the frame 24. Arrange at predetermined intervals. Since the frame 24 is prepared according to various shapes of the original image, a watermarked print image can be created in various shapes such as a polygon and a circle, and convenience is improved.

  The digital watermark extraction apparatus 200 according to the present embodiment extracts a plurality of feature points 22 from the frame 24 of the photographed image and corrects perspective distortion. Since the feature point 22 exists in the frame 24 provided around the captured image, it can be easily distinguished from the pixels in the original image region 20 and can be accurately extracted. By correcting the perspective distortion using the plurality of extracted feature points 22, distortion generated in the image can be corrected with high accuracy, and the detection frequency of the digital watermark can be increased. Moreover, even if the shape of the original image is a complicated shape, since the shape and actual size of the original image can be determined from the arrangement of the plurality of feature points 22, the original image region 20 can be accurately identified, A digital watermark can be extracted.

Embodiment 2
In the digital watermark system according to the second embodiment, a frame 24 including the feature point 22 is pasted on the original image in which the digital watermark is embedded, and a symbol mark such as a logo mark indicating that the image is a watermarked image is added to the original image. The images are superimposed and printed on a print medium, and the symbol mark is used for geometric correction of the captured image when extracting the watermark.

  FIG. 11 is a configuration diagram of the digital watermark embedding apparatus 100 according to the second embodiment. The same components as those in the digital watermark embedding apparatus 100 according to the first embodiment are denoted by the same reference numerals and description thereof will be omitted as appropriate, and the configuration and operation different from those of the digital watermark embedding apparatus 100 according to the first embodiment will be described.

  The digital watermark embedding apparatus 100 according to the present embodiment includes a logo mark superimposing unit 17 and superimposes a logo mark indicating that it is a watermarked image on the original image in which the watermark is embedded. This logo mark serves as a certification mark that proves that the image is a watermarked image.Other than that, when extracting the watermark, the shape or actual size of the original image is determined and the optical axis around the image is taken. It is also used to estimate the rotation angle.

  The logo mark superimposing unit 17 superimposes a logo mark having a predetermined shape on a predetermined position on the original image where the frame 24 is pasted by the frame pasting unit 16. The logo mark registered in advance in the frame database 18 together with the frame 24 is used.

  The printing unit 14 uses the original image in which the watermark information X is embedded by the watermark embedding unit 12, the frame 24 is pasted by the frame pasting unit 16, and the logo mark is superposed by the logo mark superimposing unit 17. Printing is performed to generate a printed image Q with a logo with a frame.

  In this embodiment, after embedding the watermark information X by the watermark embedding unit 12, the frame 24 is pasted by the frame pasting unit 16, and the logo mark superimposing unit 17 then superimposes the logo mark. However, the order of these processes may be changed. For example, processing may be performed in the order of frame pasting, logo mark superimposition, and watermark embedding, or logo mark superposition, watermark embedding, and frame pasting. There are six orders of these processes in total, but any order may be used.

  FIG. 12 is a diagram for explaining a print image Q with a logo with a frame output by the digital watermark extracting apparatus 200 of the present embodiment. A frame 24 including a plurality of feature points 22 is printed around the original image area 20 as in the first embodiment. A star-shaped logo mark 26 is printed in the lower left corner of the original image area 20. The position where the logo mark 26 is superimposed on the original image is determined at a predetermined position such as the lower left corner of the original image. Since the overlapping position of the logo mark 26 is determined, when the watermark is extracted, the relative positional relationship between the logo mark 26 and the feature point 22 is used to capture the printed image Q with the logo mark with a frame. It becomes possible to accurately determine the rotation angle around the optical axis.

  FIGS. 13A and 13B are diagrams illustrating another example of the printed image Q with a logo with a frame. FIG. 13A shows a rhombus logo mark 26 superimposed on the circular original image region 20, and FIG. 13B shows a hexagonal logo mark 26 superimposed on the triangular original image region 20. As in this example, by making the shape of the logo mark 26 different depending on the shape of the original image area 20, the shape of the original image area 20 can be determined from the shape of the logo mark 26 when extracting the watermark. .

  As another use example of the logo mark 26, various attributes such as the shape and color of the logo mark 26 are changed according to the actual size of the original image area 20, and the actual size of the original image area 20 is changed according to the type of the logo mark 26. May be determined. Further, the shape and the actual size of the original image area 20 are represented by the type of the logo mark 26, such as the actual size of the original image area 20 is represented by the color of the logo mark 26, and the shape of the original image area 20 is represented by the shape of the logo mark 26. May be determined simultaneously.

  FIG. 14 is a configuration diagram of the digital watermark extraction apparatus 200 according to the second embodiment. The same components as those in the digital watermark extracting apparatus 200 of the first embodiment are denoted by the same reference numerals, and the description thereof will be omitted as appropriate. The configuration and operation different from those of the digital watermark extracting apparatus 200 of the first embodiment will be described.

  The feature point extraction unit 34 extracts a plurality of feature points 22 in the frame 24 from the captured image whose lens distortion has been corrected by the lens distortion correction unit 32, and uses the positional information of the plurality of feature points 22 as an image shape / size determination unit. 36, the perspective distortion correction unit 38, and the rotation angle estimation unit 46.

  The logo mark extraction unit 44 extracts the logo mark 26 from the original image region 20 of the captured image whose lens distortion has been corrected by the lens distortion correction unit 32, and uses information about the logo mark 26 as the image shape / size determination unit 36 and the rotation angle. This is given to the estimation unit 46.

  The image shape / size determination unit 36 determines the shape and actual size of the original image region 20 of the captured image based on the arrangement of the plurality of feature points 22 extracted by the feature point extraction unit 34. Information on the shape and the actual size is given to the rotation angle estimation unit 46 and the template rotation unit 48.

  The image shape / size determination unit 36 may determine the shape or actual size of the original image region 20 based on the type of the logo mark 26 extracted by the logo mark extraction unit 44. For example, in the examples of FIGS. 12, 13A, and 13B, if the extracted logo mark 26 is a star shape, it is determined that the original image area 20 is a rectangle, and the logo mark 26 is a rhombus. For example, the original image area 20 is determined to be a circle, and if the logo mark 26 is a hexagon, the original image area 20 is determined to be a triangle. When the shape of the original image region 20 can be determined by the logo mark 26, information on the determined shape is given to the feature point extraction unit 34 to prevent erroneous detection when the feature point extraction unit 34 detects the feature point 22. You can also. For example, if the shape of the original image region 20 is known to be circular due to the shape of the logo mark 26, the feature point extraction unit 34 searches the feature points 22 along the circumference, thereby And the point outside the frame 24 can be prevented from being erroneously detected as feature points, and the detection accuracy of the feature points 22 can be increased.

  The rotation angle estimation unit 46 reads out from the feature point template storage unit 50 the feature point template 54 that matches the shape and actual size of the original image area 20 given from the image shape / size determination unit 36. The feature point template 54 includes position information of a plurality of reference feature points 52 and a reference logo mark 56. The rotation angle estimation unit 46 compares the position information of the feature points 22 and the logo marks 26 in the frame 24 of the captured image with the position information of the reference feature points 52 and the reference logo marks 56 in the feature point template 54. Thus, the rotation angle θ around the optical axis at the time of photographing is estimated.

  FIGS. 15A and 15B are diagrams illustrating a method of estimating the rotation angle θ by the rotation angle estimation unit 46. FIG. FIG. 15A shows the feature point template 54 read from the feature point template storage unit 50. The feature point template 54 includes 18 reference feature points 52 a, 52 b,..., 52 r and a reference logo mark 56. Corresponding to the decision to place the logo mark 26 at the lower left corner of the original image area 20 when embedding the watermark in the digital watermark embedding apparatus 100, the reference logo mark 56 is also arranged at the same position at the lower left corner.

The coordinates of the lower left corner of the reference feature points 52a of the feature point template 54 as _(x 0, _{y 0),} see below for feature point 52b in order counterclockwise, ..., a coordinate value of 52r _(x 1, y ₁ ) to (x ₁₇ , y ₁₇ ). A slope a ₀ of a straight line connecting the coordinate values (x ₀ , y ₀ ) and (x ₁ , y ₁ ) of two adjacent reference feature points 52a and 52b is given by the following equation.

a ₀ = tan (y ₁ −y ₀ ) / (x ₁ −x ₀ ) + π · f (x ₀ −x ₁ )

  However, f (x) is a step function that takes a value of 1 when x> 0 and 0 when x ≦ 0.

In general, a slope a _i of a straight line connecting the coordinate values (x _i , y _i ) and (x _{i + 1} , y _{i + 1} ) of two adjacent reference feature points is given by the following equation, and this slope a _i is expressed as a feature point template. It is used as an index indicating the inclination of 54.
a _i = tan (y _{i + 1} −y _i ) / (x _{i + 1} −x _i ) + π · f (x _i −x _{i + 1} )

  FIG. 15B shows the frame 24 and the logo mark 26 in the captured image. From the relative positional relationship between the logo mark 26 on the original image area 20 and the frame 24, each feature point 22 in the frame 24 and each reference feature point 52 in the feature point template 54 are associated one-to-one. In this example, the reference feature point 52a in the lower left corner of the feature point template 54 in FIG. 15A is associated with the feature point 22a in the upper right corner of the frame 24 of the captured image in FIG. 15B.

Similarly, the reference feature point 52b immediately adjacent to the reference feature point 52a in the lower left corner of the feature point template 54 in FIG. 15A is the feature point 22a in the upper right corner of the frame 24 in FIG. 15B. The reference feature point 52r adjacent to the feature point 22b adjacent to the lower left corner of the feature point template 54 shown in FIG. ) In the upper right corner of the frame 24 is associated with the feature point 22r adjacent to the lower right corner of the feature point 22a. The coordinate value of the feature point 22a at the upper right corner in the frame 24 is (v ₀ , w ₀ ), and the coordinate values of the feature point 22 are sequentially (v ₁ , w ₁ ) to (v ₁₇ , w ₁₇ ) counterclockwise. )far. The slope b ₀ of a straight line connecting the coordinate values (v ₀ , w ₀ ) and (v ₁ , w ₁ ) of two adjacent feature points 22a and 22b is given by the following equation.

b ₀ = tan (w ₁ −w ₀ ) / (v ₁ −v ₀ ) + π · f (v ₀ −v ₁ )

In general, the inclination b _i of a straight line connecting the coordinate values (v _i , w _i ) and (v _{i + 1} , w _{i + 1} ) of two adjacent feature points is given by the following equation, and this inclination b _i is expressed as the inclination of the frame 24. Used as an indicator.
b _i = tan (w _{i + 1} −w _i ) / (v _{i + 1} −v _i ) + π · f (v _i −v _{i + 1} )

A difference c _i = b _i −a _i between the feature point template 54 and the frame 24 gives an angle at which the frame 24 is rotated around the optical axis with the feature point template 54 as a reference position. The rotational angle estimation unit 46 obtains the average value ^{_{θ = Σ i = 0 17 c}} 0/18 of the difference c _i of slope, the average value theta, the optical axis of the frame 24 when the feature point template 54 as the reference position Output as an estimated value of the surrounding rotation angle. At the stage where the rotation angle estimation unit 46 estimates the rotation angle of the frame 24, the photographed image is in a state before the perspective distortion is corrected, so the coordinate value of the feature point 22 in the frame 24 extracted from the photographed image is Including errors due to perspective distortion. Therefore, the rotation angle estimation unit 46 reduces the estimation error of the rotation angle θ by taking the average of the inclination differences c _i .

  Returning to FIG. 14, the rotation angle estimation unit 46 gives the estimated rotation angle θ to the template rotation unit 48 and the corrected image rotation unit 42. The template rotation unit 48 reads out the feature point template 54 that matches the shape and actual size of the original image area 20 determined by the image shape / size determination unit 36 from the feature point template storage unit 50, and rotates the feature point template 54 at the rotation angle. It is rotated by θ and given to the perspective distortion correction unit 38.

  The perspective distortion correction unit 38 includes a plurality of feature points 22 in the frame 24 extracted from the captured image by the feature point extraction unit 34 and a plurality of references in the rotated feature point template 54 provided from the template rotation unit 48. The feature points 52 for use are associated one-to-one and applied to the perspective distortion correction function g to calculate a perspective distortion correction coefficient. The perspective distortion correction unit 38 corrects the perspective distortion of the original image area 20 using the calculated perspective distortion correction function g, and gives the corrected original image area 20 to the corrected image rotation unit 42.

  The corrected image rotation unit 42 rotates the original image region 20 corrected by the perspective distortion correction unit 38 by −θ based on the rotation angle θ estimated by the rotation angle estimation unit 46. The original image area 20 rotates counterclockwise by the rotation angle θ and is converted into a state where there is no rotation around the optical axis. The corrected image rotating unit 42 gives the counter-corrected corrected image to the watermark extracting unit 40, and the watermark extracting unit 40 extracts the watermark information X from the counter-rotated corrected image and outputs it.

  FIG. 16 is a flowchart for explaining a digital watermark extraction procedure performed by the digital watermark extraction apparatus 200 according to this embodiment.

  The photographing unit 30 photographs the printed image Q with the logo with a frame (S50). The lens distortion correction unit 32 corrects lens distortion generated in the captured image (S52). The feature point extraction unit 34 extracts a plurality of feature points 22 from the frame 24 of the captured image whose lens distortion has been corrected, and the logo mark extraction unit 44 extracts a logo mark from the original image region 20 of the captured image (S54). ). The image shape / size determination unit 36 determines the shape and actual size of the original image area 20 from the arrangement of the extracted feature points 22 (S56).

  The rotation angle estimation unit 46 reads the feature point template 54 that matches the shape and actual size of the original image region 20 from the feature point template storage unit 50 (S58). The rotation angle estimation unit 46 determines the reference feature points 52 in the feature point template 54 and the captured image based on the relationship between the position of the reference logo mark 56 in the feature point template 54 and the position of the logo mark 26 in the frame 24 of the captured image. The feature points 22 in the frame 24 are associated one-to-one, and the rotation angle θ around the optical axis of the frame 24 is estimated using the feature point template 54 as a reference (S60).

  The template rotation unit 48 rotates the feature point template 54 read from the feature point template storage unit 50 by the rotation angle θ (S62).

  The perspective distortion correction unit 38 includes one reference feature point 52 in the feature point template 54 rotated by the rotation angle θ estimated by the rotation angle estimation unit 46 and one feature point 22 in the frame 24 of the captured image. The perspective distortion coefficient g is applied to the perspective distortion correction function g to calculate a perspective distortion coefficient, and the perspective distortion generated in the original image area 20 is corrected by the calculated perspective distortion correction function g (S64). The corrected image rotating unit 42 rotates the original image area 20 after the perspective distortion correction by −θ (S66). The watermark extraction unit 40 extracts the watermark information X from the original image area 20 that has been corrected in the opposite direction by the rotation angle θ around the optical axis after correcting the lens distortion and the perspective distortion (S68).

  The digital watermark embedding apparatus 100 according to the present embodiment pastes a frame 24 in which a plurality of feature points 22 are arranged around an original image in which the digital watermark is embedded, and then places a logo mark 26 on a predetermined image of the original image. Superimpose on position. The presence of the logo mark 26 allows the user to easily know that the image is a watermarked image. The digital watermark extraction apparatus 200 can correct the rotation direction by detecting the rotation angle around the optical axis by using the relative positional relationship between the logo mark 26 and the feature point 22 in the frame 24. In addition, by making the type of the logo mark 26 different depending on the shape and the actual size of the original image, the shape and the actual size of the original image taken according to the type of the logo mark 26 can be determined, and the original image area is accurately cut out. Thus, the watermark can be extracted.

  The present invention has been described based on the embodiments. The embodiments are exemplifications, and it will be understood by those skilled in the art that various modifications can be made to combinations of the respective constituent elements and processing processes, and such modifications are within the scope of the present invention. .

  In the second embodiment, the rotation angle estimation unit 46 estimates the rotation angle θ around the optical axis of the frame 24 from the relative positional relationship between the logo mark 26 and the feature point 22. In this case, when it can be assumed that the rotation angle θ around the optical axis is within ± 45 °, each feature point 22 of the frame 24 is assumed on the assumption that the rotation angle θ is within ± 45 °. Since the reference feature points 52 of the feature point template 54 can be associated with each other, the rotation angle θ can be calculated without using the logo mark 26.

1 is a configuration diagram of a digital watermark embedding device according to Embodiment 1. FIG. It is a figure explaining the printing image with a frame output by the digital watermark extraction apparatus of FIG. It is a figure explaining another example of the printing image with a frame. It is a figure explaining another example of the printing image with a frame. It is a figure explaining another example of the printing image with a frame. 3 is a flowchart for explaining a digital watermark embedding procedure by the digital watermark embedding apparatus in FIG. 1. 1 is a configuration diagram of a digital watermark extraction apparatus according to Embodiment 1. FIG. It is a figure explaining a mode that the geometric distortion which arose in the picked-up image is correct | amended by the lens distortion correction part and perspective distortion correction part of FIG. It is a figure explaining a mode that perspective distortion is amended based on a plurality of feature points extracted from a photography picture. It is a flowchart explaining the digital watermark extraction procedure by the digital watermark extraction apparatus of FIG. 6 is a configuration diagram of a digital watermark embedding device according to Embodiment 2. FIG. It is a figure explaining the printed image with the logo mark with a frame output by the digital watermark extraction apparatus of FIG. It is a figure explaining another example of the printed image containing a logo mark with a frame. 6 is a configuration diagram of a digital watermark extraction apparatus according to Embodiment 2. FIG. It is a figure explaining the estimation method of the rotation angle around the optical axis at the time of imaging | photography by the rotation angle estimation part of FIG. It is a flowchart explaining the digital watermark extraction procedure by the digital watermark extraction apparatus of FIG.

Explanation of symbols

  10 image forming unit, 12 watermark embedding unit, 14 printing unit, 16 frame pasting unit, 17 logo mark overlapping unit, 18 frame database, 20 original image area, 22 feature points, 24 frames, 26 logo mark, 30 shooting unit, 32 lens distortion correction unit, 34 feature point extraction unit, 36 image shape / size determination unit, 38 perspective distortion correction unit, 40 watermark extraction unit, 42 corrected image rotation unit, 44 logo mark extraction unit, 46 rotation angle estimation unit, 48 Template rotation unit, 50 feature point template storage unit, 52 reference feature point, 54 feature point template, 56 reference logo mark, 100 digital watermark embedding device, 200 digital watermark extraction device

Claims (12)

A frame database storing a plurality of types of frames for arranging a plurality of feature points at regular intervals along the contour of the image at a constant interval;
A frame pasting unit that selects the frame that matches the shape of the original image to be embedded with the digital watermark from the frame database and pastes the frame around the original image;
A watermark embedding unit for embedding a digital watermark in the original image;
An electronic watermark embedding apparatus comprising: an output unit configured to output the original image in which the electronic watermark is embedded and the plurality of feature points are arranged along an outline of the image on a print medium.   The digital watermark embedding apparatus according to claim 1, further comprising a mark superimposing unit that superimposes a symbol mark indicating that the image has an electronic watermark at a predetermined position of the original image.   The digital watermark embedding apparatus according to claim 2, wherein the type of the symbol mark to be superimposed on the original image differs depending on the shape or actual size of the original image. An image input unit for acquiring a print image obtained by printing an original image embedded with a digital watermark as a digital image;
A feature point extraction unit that extracts a plurality of feature points arranged at a constant interval that does not depend on the image size along the contour of the original image region of the digital image;
A determination unit that determines the shape and actual size of the original image based on the arrangement of the plurality of feature points;
For each of the plurality of feature points, a perspective distortion correction unit that corrects the perspective distortion generated in the original image area by detecting a deviation from the position where it should be,
An electronic watermark extracting apparatus comprising: a watermark extracting unit that extracts the electronic watermark from the corrected original image area according to the determined shape and actual size of the original image. A lens distortion correction unit that corrects lens distortion generated in the original image area;
The feature point extraction unit extracts the plurality of feature points from the original image region in which lens distortion is corrected by the lens distortion correction unit,
5. The watermark extraction unit extracts the digital watermark from the original image region in which lens distortion is corrected by the lens distortion correction unit and perspective distortion is corrected by the perspective distortion correction unit. The electronic watermark extraction apparatus described. A rotation angle estimation unit that estimates the rotation angle around the optical axis at the time of photographing by detecting the inclination when compared with the original arrangement for the arrangement of the plurality of feature points,
The perspective distortion correction unit corrects the rotation direction based on the estimated rotation angle, and detects a deviation from the original position for each of the plurality of feature points. 6. The digital watermark extracting apparatus according to claim 4, wherein the perspective distortion generated in the region is corrected. A mark extraction unit for extracting a symbol mark indicating that the image is a digital watermarked image from the acquired digital image;
The digital watermark extraction apparatus according to claim 6, wherein the rotation angle estimation unit estimates the rotation angle using a relative positional relationship of the plurality of feature points with respect to the symbol mark. A mark extraction unit for extracting a symbol mark indicating that the image is a digital watermarked image from the acquired digital image;
6. The digital watermark extraction apparatus according to claim 4, wherein the determination unit determines a shape or an actual size of the original image according to a type of the symbol mark.   When an original image of a predetermined shape is given as an embedding target of a digital watermark, a frame having the same shape as the original image in which a plurality of feature points are arranged at regular intervals along the outline of the original image without depending on the image size. An electronic watermark embedding method comprising: pasting around an original image, embedding an electronic watermark in the original image, and outputting the image for printing on a print medium.   A print image on which an original image embedded with a digital watermark is printed is acquired as a digital image, and a plurality of feature points arranged at regular intervals along the contour of the original image area of the digital image are extracted. Then, the original image determined based on the arrangement of the plurality of feature points is corrected by detecting a deviation from the original position of the feature points to correct the perspective distortion generated in the original image region. A digital watermark extracting method, wherein the digital watermark is extracted from the corrected original image area in accordance with the shape and actual size of the original. From the frame database that stores multiple types of frames for arranging multiple feature points at regular intervals along the contour of the image to match the shape of the image, the shape of the original image to be embedded with the digital watermark Selecting the matched frame and pasting it around the original image;
Embedding a watermark to embed a digital watermark in the original image;
A program for causing a computer to execute the step of outputting the original image in which the digital watermark is embedded and the plurality of feature points are arranged along an outline of the image for printing on a print medium. Obtaining a print image on which an original image embedded with a digital watermark is printed as a digital image;
Extracting a plurality of feature points arranged at regular intervals independent of the image size along the contour of the original image area of the digital image;
Determining the shape and actual size of the original image based on the arrangement of the plurality of feature points;
Correcting the perspective distortion generated in the original image area by detecting a deviation from the position where it should be for each of the plurality of feature points;
A program for causing a computer to execute the step of extracting the digital watermark from the corrected original image area according to the determined shape and actual size of the original image.

Images