JP2009049765A - Signal converter and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To automatically determine a conversion factor that obtains a maximum contrast within its range while controlling a remaining amount of other images when the other images after rounding processing are allowed to remain in order to enhance the contrast of all images. <P>SOLUTION: An extended-table generation part 11 generates an extended table X' generated by extending a table X" restricting a combination range taken by a plurality of input signals. A maximum/minimum luminance detection part 13 extracts a contact point where a uniform signal contacts with a range of the extended table X'. A maximum/minimum luminance detection part 12 detects singularity that corresponds to the contact point about each input signal. A conversion factor determination part 14 determines a conversion factor of a DR conversion part 15 by mapping the singularity to the contact point. Input signals g1, g2, and s are subjected to DR conversion by the DR conversion part 15, and further, subjected to rounding processing by a rounding processing part 16. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a signal conversion apparatus and program, and more particularly to a signal conversion apparatus and program for converting a plurality of signals in which the range of possible combinations is limited to a combination of signals in which their dynamic ranges are converted.

  In order to make the difference in the signal easier to perceive, its dynamic range is converted. For example, in the case of a natural image, it can be changed to an image with high contrast by adding a process for widening the dynamic range. Dynamic range conversion for a plurality of signals in which the range of possible combinations is restricted is performed, for example, when generating a tally image.

  The tally image means an image that can be decrypted by calculating or optically superimposing a plurality of images for each pixel. In other words, a tally image does not decode a secret image from one or less than a predetermined number of tally images, but a secret image is decoded when a predetermined number of tally images are superposed with their pixel positions aligned. Is.

  FIG. 5 is an explanatory diagram conceptually showing decryption of a secret image by superimposition of tally images. As shown in the figure, for example, when the tally images W1, W2,..., Wn are expressed with pseudo-grayscale with black pixels as 0 and white pixels as 1, these tally images W1, W2,. , Wn is calculated for each pixel (AND, OR, NOR, XOR, etc.) or optically superimposed to generate a superimposed image (decoded image) T. The secret image S is decoded as a superimposed image T.

  Conventional tally image generation methods can be broadly classified into a method using only a pixel arrangement determining unit and a method using a luminance changing unit and a pixel arrangement determining unit, and the method in the present invention belongs to the latter.

  Non-Patent Documents 1 and 2 describe a method of generating a tally image using a luminance changing unit and a pixel arrangement determining unit. This method is mainly intended for the case where both the original image for tally and the secret image are grayscale images, and is intended to express the grayscale image as a pseudo gray image as a tally image and a superimposed image. According to this method, the tally image can be made difficult to see on the superimposed image, and the tally image can be a natural image. In addition, although the secret image can be a natural image and can be decoded satisfactorily, in order to do so, it is necessary to perform luminance change processing and maintain the dynamic range of the original image for tally and the secret image to some extent over the entire image.

  FIG. 6 is a block diagram showing a conventional configuration for generating a tally image using a luminance changing unit and a pixel arrangement determining unit. In the figure, tally original images G1, G2,..., Gn and tally images W1, W2,..., Wn are illustrated, but in the following, the tally original images are G1, G2, secret images. An example will be described in which T is set to S and tally images W1 and W2 are generated based on these images. In addition, the images are indicated by capital letters, and the luminance (pixel value) of each image is indicated by small letters.

  First, the original tally images G1 and G2 and the secret image S are input to the luminance conversion unit 51. The luminance conversion unit 51 includes a dynamic range (DR) conversion unit 511 and a rounding processing unit 512, and applies DR conversion and rounding processing to the input tally original images G1 and G2 and the secret image S, respectively. The tally generating original images G1 ″ and G2 ″ and the tally generating secret image S ″ are output. The luminance conversion unit 51 may include only the DR conversion unit and may not include the rounding processing unit.

  The pixel arrangement determining unit 52 receives the tally generating original images G1 ″ and G2 ″ and the tally generating secret image S ″ as input, and generates tally images W1 and W2. Here, the tally images W1 and W2 are made to be pseudo gray images close to the grayscale (multi-valued) tally generation original images G1 ″ and G2 ″, and the tally images W1 and W2 are combined. When a low-pass filter (human visual characteristics) is applied to the superimposed image T, an image of a pseudo gray expression that is close to the grayscale (multi-valued) tally generating secret image S ′ ′ is decoded (the human is the superimposed image T The tally images W1 and W2 are generated by determining the pixel arrangement so that an image close to the tally generating secret image S ″ is visually recognized. The tally images W1 and W2 are generally monochrome binary images, and the superimposed image T in which the tally images W1 and W2 are combined is also generally a monochrome binary image.

  As described above, the tally images W1 and W2 and the superimposed image T obtained by superimposing them (an image in which the secret image is visually decoded) are binarized by the tally generation process. The generation process consists of tally generating original images G1 ″, G2 ″ having a luminance g1 ″, g2 ″, s ″ given as input, a tally generating secret image S ″, and black and white binary tally. Since the local pseudo gray level representation of the image is almost the same, the combination of pixels with the given luminance g1 ″, g2 ″, s ″ cannot be expressed in the binary image. There is a case. In this case, the tally images W1 and W2 cannot be generated, or even if the tally images W1 and W2 can be generated, the degree to which other images remain (mixed) (hereinafter referred to as the remaining degree) increases, which may cause poor image quality.

  For example, when the luminance of the image is set to 0 to 255, the luminances g1 ″ and g2 ″ of the tally generating original images G1 ″ and G2 ″ are both equal to 128 (1/2 of white (gray) ), As shown in FIG. 7, the luminance of 0 (black) or 128 (gray) can be expressed uniformly by the combination of both pixel arrangements (FIGS. 7A and 7B). However, the brightness of 192 (light gray with a white density of 3/4) cannot be expressed (FIG. 7 (c)).

  In this way, when a grayscale (multi-value) luminance g1 ″, g2 ″ is expressed in a pseudo gray scale with a plurality of black and white pixels, by changing the Hamming distance of the pixels of g1 ″, g2 ″, Although the pseudo gray expression of s ″ can be changed, S ″ that can express the pseudo gray with a combination of black and white pixels is restricted. This constraint is expressed by equation (1).

  Max (0, g1 '' + g2 ''-255) ≤S''≤Min (g1 '', g2 '') (1)

  It is desirable that the tally images W1 and W2 represent the tally original images G1 and G2 almost in pseudo gray, and the superimposed image T represents the secret image S in almost pseudo gray, but from the equation (1), the superimposed image T represents The range of the pseudo density tT that can be obtained is limited by the luminances g1 ″ and g2 ″ of the tally images W1 and W2.

  FIG. 8 shows a range satisfying the formula (1), and a region surrounded by the tetrahedron defg (inside the tetrahedron) satisfies the formula (1). This is described in detail in Non-Patent Document 3. In the figure, points k ′, m ′, and n ′ correspond to the cases (a), (b), and (c) of FIG.

  Therefore, in order for the superimposed image T to express the secret image S in a pseudo gray scale with a certain dynamic range, it is usually necessary to narrow the dynamic range of the tally original images G1 and G2. Similarly, the dynamic range of the secret image S needs to be narrowed.

  FIG. 9 is a functional block diagram showing a specific configuration of the luminance changing unit of Non-Patent Document 2. This luminance conversion unit converts the DR of the tally images G1 and G2 and the secret image S so that the luminance after DR conversion satisfies Equation (1). To that end, first, the polyhedron extraction unit 82 extracts polyhedrons that surround the range of combinations that can be taken by the luminances g1, g2, s of the original images G1 and G2 for the tally and the secret image S (region satisfying the expression (1)). , Generate table X ″ that constrains the range of possible combinations of luminances g1 ″, g2 ″, s ″ of the original images G1 ″, G2 ″ for tally generation and the secret image S ″ for tally generation To do.

  Next, DR conversion coefficients (a1, b1), (a2, b2), (a3, b3) for g1, g2, and s of the DR conversion unit 83 are appropriately set, and the original images G1, G2 for tally and The luminance g1, g2, and s of the secret image S are DR converted by g1 ″ = a1 * g1 + b1, g2 ″ = a2 * g2 + b2, s ′ = a3 * s + b3, and after DR conversion The check unit 84 checks whether the luminances g1 ″, g2 ″, and s ″ are combinations within the range restricted by the table X ″. Here, if the luminance g1 ″, g2 ″, s ″ is not a combination within the range restricted by the table X ″, the DR conversion coefficient for the g1, g2, and s of the DR conversion unit 83 The processing is repeated by changing (a1, b1), (a2, b2), (a3, b3).

  The above DR conversion and check are performed in an exploratory manner. The DR conversion coefficient determination unit 85 is a combination of luminance g1 ″, g2 ″, s ″ after DR conversion for the tally generating original images G1 ″ and G2 ″ and the tally generating secret image S ″. Is the combination of coefficients within the range constrained by table X ′ ′, the coefficients (a1, b1), (a2, b2), and (a3, b3) are the DR conversion coefficients for g1, g2, and s, respectively. decide.

  After that, the DR converter 81 uses the tally original images G1, G2 and the brightness g1, g2, s of the secret image S for g1, g2, and s DR conversion coefficients (a1, b1), (a2, b2), respectively. , (a3, b3) to generate the brightness g1 '', g2 '', s '' of the original tally generating images G1 '' and G2 '' and the tally generating secret picture S '' To do. The luminances g1 ″, g2 ″, s ″ of the tally generating original images G1 ″, G2 ″ and the tally generating secret image S ″ thus generated are directly input to the pixel arrangement determining unit.

  FIG. 10 is a functional block diagram illustrating a specific configuration of the luminance conversion unit of Non-Patent Document 1. This luminance conversion unit performs DR conversion and rounding processing. In this case, the luminance after DR conversion does not have to satisfy the equation (1), but further, a rounding process is performed to distribute the luminance error component of the equation (1) to a plurality of images. That is, it is allowed that a part of the other image is visible on the tally image or the superimposed image, and the DR converter 91 performs the tally original image G1 in order to improve the visibility without reducing the contrast of the image too much. , G2 and the luminances g1, g2, and s of the secret image S are DR-converted by g1 ′ = a1 * g1 + b1, g2 ′ = a2 * g2 + b2, and s ′ = a3 * s + b3, respectively. Thereafter, the rounding processing unit 92 performs rounding processing for dispersing the luminance error component, and the luminances g1 ″, g2 ″ of the tally generating original images G1 ″, G2 ″ and the tally generating secret image S ″. , s ′ ′.

The DR conversion coefficients (a1, b1), (a2, b2), (a3, b3) for g1, g2, and s are manually set in the DR conversion coefficient determination unit 93. The rounding processing performed by the rounding processing unit 92 is a range of combinations that the luminances g1 ″, g2 ″, and s ″ of the original tally generating images G1 ″ and G2 ″ and the tally generating secret image S ″ can take. This is done by referring to the table X ″ that constrains. The luminances g1 ″, g2 ″, s ″ of the tally generating original images G1 ″, G2 ″ and the tally generating secret image S ″ thus generated are input to the pixel arrangement determining unit.
M. Nakajima and Y. Yamaguchi, “Extended visual cryptography for natural images,” Journal of WSCG vol.2, pp.303-310,2002. CWWu and GRThompson, “Digital watermarking and steganography via overlays of halftone images”, Proc.SPIE, Vol.5561, pp.152-163,2004

  The tally image generation method described in Non-Patent Document 2 basically uses a method such as a convex hull so as to satisfy Equation (1), and a combination of luminances that can be taken by each image. Find the polyhedron that encloses the range, that is, the region that satisfies Equation (1), and set the DR conversion coefficient (DR (a1, a2, a3) and the minimum value so that the pixel combination conversion result is the pixel combination in the polyhedron (b1, b2, b3)) are obtained exploratively, and DR conversion is performed using this DR conversion coefficient.

  In this tally image generation method, since a combination of luminances outside the range of Expression (1) is not allowed after DR conversion, there is a problem that DR tends to be narrow and it is difficult to visually recognize an object on a tally image or a superimposed image. There is also a problem that it takes time to search for a convex hull and a DR conversion coefficient.

  In the tally image generation method described in Non-Patent Document 1, in order to increase the contrast of each image, a pixel set that does not satisfy Expression (1) is allowed, and first, the image is changed to a wide DR image.

  According to this, the restriction of DR can be relaxed, and the object in the image can be easily recognized. Thereafter, a rounding process (error dispersion process) is performed on a pixel group that does not satisfy Expression (1) only by the DR conversion. For example, when the combination of the luminances g1 ′, g2 ′, and s ′ of the images G1 ′, G2 ′, and S ′ after DR conversion is the point n ′ that is not included in the tetrahedron defg in FIG. The point n ′ ′ (g1 ′ ′, g2 ′ ′, s ′ ′) = n ′ (g1 ′, g2 ′, s ′) + e (eg1, eg2, es) on the nearest tetrahedron defg In this way, the error e for n ′ with respect to n ′ is distributed to the images G1 ′, G2 ′, S ′ by eg1, eg2, es, and the tally generating original images G1 ″, G2 ′. The luminances g1 ″, g2 ″ and s ″ of the secret image S ″ for generating the t and tally are generated.

  At this time, the wider the DR, the higher the contrast, but the greater the degree that other images remain (mixed) (the dispersed luminance error components eg1, eg2, es appear in the tally image and the superimposed image) There is.

  In the tally image generation method described in Non-Patent Document 1, the ratio of the pixel group having the luminance satisfying the expression (1) is set under the condition that both the degree of remaining image (remaining degree) and the contrast are compatible. Constraint fulfillment rate (CFR ), And when the CFR value is 0.6 or less, it has been found that the remaining degree and contrast of other images are not compatible.

  However, even if the CFR is a constant value, there is actually a large difference in the remaining level depending on the combination of images, so the CFR is too effective to determine whether the remaining level and contrast of other images are compatible. It is not a good evaluation scale. For example, if there is a small object in the image and the brightness of the pixel set at that object's position is outside the range of equation (1), the higher the object contrast, the easier it will remain in other images even if the CFR is the same value. . Further, Non-Patent Document 1 does not propose a method that replaces the exploratory method that can automatically determine the DR conversion coefficient in accordance with the luminance combination of the actually input image.

  An object of the present invention is to control the remaining amount and to obtain the maximum contrast (dynamic range) within the range when the remaining of other images after the rounding process is allowed to increase the contrast of all images. To provide a signal conversion apparatus and program capable of automatically determining conversion coefficients in accordance with a combination of luminances of actually input images. It is another object of the present invention to provide a signal conversion apparatus and program that can determine a conversion coefficient at high speed and uniquely without using exploratory processing.

  In order to solve the above-described problems, the present invention provides a signal conversion apparatus that converts a plurality of signals in which the range of possible combinations is limited to a combination signal in which the dynamic range is converted and outputs the signals. An extension table generating means for generating an extension table by extending a table that restricts the range of combinations that can be taken to a range of combinations that the plurality of signals cannot take according to a preset value, and a plurality of inputted signals, A first feature is that signal conversion means for converting the combinations so that the combinations are within the range of the combinations according to the restrictions in the extension table is provided.

  In addition, the present invention has a second feature in that the extension table generation unit generates the extension table by performing a reverse rounding process on the preset value and adding it to the table.

  In the present invention, the signal conversion means performs a search for determining a conversion formula in order to convert a plurality of input signals so that their combinations are within the range of combinations that comply with the constraints in the extension table. There is a third feature in that the target process is executed.

  Further, the present invention provides the first detection means for detecting the contact at each reference signal from the contact to the extension table in the plurality of reference signals within the range of the combination according to the restriction in the extension table, the signal conversion means, Second detection means for detecting a singular point associated with the contact detected by the first detection means for a plurality of input signals, and a singular point detected by the second detection means are detected by the first detection means. Conversion coefficient determining means for determining a conversion coefficient to be mapped to the contact point, and using the conversion coefficient determined by the conversion coefficient determining means, a plurality of input signals are combined in the extension table. A fourth feature is that the conversion is performed so as to be within the range of the combination according to the restriction.

  Further, in the present invention, the first detection unit extracts a contact point where a combination of the plurality of reference signals is in contact with a range restricted by the extension table, and performs vector scalar conversion on the plurality of reference signals at the contact point. A fifth feature is that a maximum value and a minimum value of each of the plurality of reference signals subjected to vector-scalar conversion are detected as contact points.

  In the present invention, the second detection unit extracts singular points of the plurality of signals associated with the contact points detected by the first detection unit, and performs vector-scalar transformation on the plurality of signals at the singular points. A sixth feature is that the maximum value and the minimum value of each of the plurality of signals subjected to vector-scalar conversion are detected as singular points.

  In addition, the present invention has a seventh feature in that the plurality of signals are images.

  Further, the present invention is a secret image in which the plurality of signals are embedded in a plurality of tally images and a tally image, and the secret image is decoded by superimposing the plurality of tally images. There is an eighth feature.

  In addition, the present invention further includes rounding processing means for rounding a plurality of signals converted by the signal conversion means so that a combination of the signals is a combination of a plurality of signals according to the constraints of the table. There is a ninth feature.

  Furthermore, the present invention has a tenth feature in that the rounding processing means applies a low-pass filter to the error component during the rounding processing.

  The present invention can be realized not only as a signal conversion apparatus but also as a program that causes a computer to function as each unit of the signal conversion apparatus.

  According to the present invention, a table that restricts the range of combinations that can be taken by a plurality of tally signals is expanded to a range of combinations that cannot be taken by a plurality of signals according to a preset remaining allowable value, and an extended range is obtained. Multiple signals are converted so that their combination has the maximum contrast within the extended range, so the contrast (difference) in multiple signals can be strengthened and combined to make rounding processing. Even in such a case, it can be converted into a plurality of signals in which the remaining level of other signals is limited.

  Also, conversions that find contact points for the extended range of multiple assumed reference signals, extract singular points that are likely to become contact points after conversion from the actual input signals, and map them based on them. By adopting a method of determining a coefficient, a signal conversion coefficient can be uniquely determined automatically and at high speed. According to this method, since the signal conversion coefficient can be determined with a light processing load, the signal conversion can be made suitable for the server application.

  In addition, when multiple signals are images, multiple images are generated that have high contrast, make the object easily visible, and even when rounded, the remaining level of other images is limited. it can.

  According to the present invention, it is possible to create a high-contrast tally image with limited remainingness from various input images, and a high-contrast tally image with limited remainingness increases the unexpectedness of a secret image decoded therefrom. It is suitable for entertainment and advertisements such as lotteries and greeting cards.

  The present invention will be described below with reference to the drawings. In the following, as shown in FIG. 6, DR conversion is performed on the images G1, G2, and S by the DR conversion unit, and the images G1 ′, G2 ′, and S ′ generated thereby are further input to the rounding processing unit to generate tally. The case where the original images G1 ″, G2 ″ and the tally generating secret image S ″ are generated will be described as an example. The signal conversion apparatus of the present invention can be used as a DR conversion unit in this case. As preprocessing, the original images G1, G2, and S may be subjected to histogram flattening, S-shaped tone curve correction, and the like, and the contrast is enhanced by extending the intermediate gradation.

  In the present invention, the remaining degree of other images is considered to be proportional to errors (dispersion errors) eg1, eg2, es distributed in the tally generating original images G1 ″, G2 ″ and the tally generating secret image S ″. The maximum values eg1max, eg2max, esmax of the absolute values of the dispersion errors eg1, eg2, es are used as the image quality evaluation scale.

  By calculating the DR conversion coefficient so that the pixel set (g1 ', g2', s ') of the image G1', G2 ', S' after DR conversion disperses errors below a predetermined value as a result Regardless of the input tally original images G1 and G2 and the tally secret image S, the remaining levels of other images in the converted tally generating original images G1 ″ and G2 ″ and the tally generating secret image S ″ Can be expected to be below a certain level.

  Conversion formula g1 ′ = a1 * g1 from the luminance g1, g2, s of the tally original image G1, G2 and the secret tally image S to the luminance g1 ′, g2 ′, s ′ of the image G1 ′, G2 ′, S ′ + b1, g2 ′ = a2 * g2 + b2, s ′ = a3 * s + b3 conversion coefficients (a1, b1), (a2, b2), (a3, b3) are represented by images G1 ′, G2 ′, When S ′ is rounded and the tally generating original images G1 ″, G2 ″ and the tally generating secret image S ″ are generated, the variance error is less than the maximum values eg1max, eg2max, esmax. Ask.

  In the following, for simplicity of explanation, the maximum dispersion error value eg1max = eg2max = esmax = emax, and the ratio of DR after conversion is DR (g1 ′): DR (g2 ′): DR (s ′) = 1: 1: r (cuboid).

  FIG. 1 is a functional block diagram showing an example of a signal conversion apparatus according to the present invention, and FIG. 2 is a flowchart showing processing in FIG. FIG. 3 is a diagram for explaining the process.

  Hereinafter, a method of obtaining a conversion coefficient of a linear function for obtaining the images G1 ′, G2 ′, S ′ from the input images G1, G2, S will be described with reference to FIG. 1 and FIG. In the present invention, the remaining of other images is allowed to be constant by the following means, the maximum dynamic range (contrast) is taken within the range, and the conversion coefficient is obtained at high speed.

  First, as shown in FIG. 2, a maximum value emax (emax = eg1max = eg2max = esmax) allowing the remaining of other images is set (S11). The extension table generation unit 11 restricts the range of combinations that the luminances g1 ″, g2 ″, and s ″ of the tally generating original images G1 ″ and G2 ″ and the tally generating secret image S ″ can take. From the table X ″ and the allowable maximum remaining value emax set earlier (emax = eg1max = eg2max = esmax), the luminance (g1) allowed for the pixel set of the images G1 ′, G2 ′, S ′ after DR conversion An extension range (extension table) X ′ that constrains the range of the combination of ′, g2 ′, s ′) is generated (S12).

The extended table X 'luminance g1 which constrains', g2 ', s' combination of range of, for example, be in the tetrahedral d _{e e e} f _e g _e shown in FIG. 3, the tetrahedral d _e e _e f _e g _e is the plane efg, def, dfg, deg of the tetrahedron defg in the normal vector direction respectively (0,0, emax), (emax, 0, emax), (0, emax, emax), (emax , emax, emax). That is, the extension table X ′ is extended to a range where the combination of the luminances g1 ′, g2 ′, and s ′ cannot be taken. This extended table X ′ can be generated by de-rounding emax and adding it to the table X ″. The inverse rounding process is a process opposite to the error distribution process (the process in the rounding processing unit 512 in FIG. 6), and uses a vector having a sign opposite to the vector from the pre-error distribution to the pixel group after the error distribution. realizable.

  Next, the pixel set (g1, g2, s) of the input image is converted so as to exist in this extended range X ′, and the remaining maximum value is limited to emax, and then the images G1 ′, G2 after DR conversion A method will be described in which ′ and S ′ have the maximum DR in the extension table X ′, that is, the maximum contrast for facilitating visual recognition of the object. For this purpose, it is necessary that the pixel groups (g1 ′, g2 ′, s ′) of the images G1 ′, G2 ′, S ′ after DR conversion are inscribed in the range restricted by the extension table X ′. The conversion formula for inscribing the pixel group (g1 ', g2', s') of the image G1 ', G2', S 'after DR conversion inscribed in the range constrained by the extension table X' Can be obtained exploratively by re-determining the conversion coefficient again if it is not in contact with the extended range. A method for determining the coefficient of the conversion formula uniquely and with a light processing load will be described. For this purpose, not only an actual input image but also an image preliminarily assumed that the distribution of combinations of input signals is uniform is used. Then, a contact point with the extended range is obtained from an image assumed in advance, only a singular point that is likely to be a contact point is obtained from an actual input image, and the conversion coefficient is uniquely obtained by mapping the singular point and the contact point.

  When the histogram of the luminance y1, y2, y3 of the assumed image Y1, Y2, Y3 is uniform or biased at both ends, the existence range Y of the pixel set (y1, y2, y3) is the same as that of each image Y1, Y2, Y3. When the luminances y1, y2, and y3 are three-dimensionally distributed in a three-dimensional shape and distributed in a halftone, it is confirmed that the cubes are distributed in an expanded shape. Therefore, the existence range Y of the pixel group (y1, y2, y3) of the image Y1, Y2, Y3 takes the shape of a rectangular parallelepiped, and the pixel group (y1 ′) of the image Y1 ′, Y2 ′, Y3 ′ after the DR conversion , y2 ′, y3 ′) is a rectangular parallelepiped shape, and Y ′ is inscribed in the range constrained by the extension table X ′.

  When the existence range Y ′ of the pixel set (y1 ′, y2 ′, y3 ′) of the image Y1 ′, Y2 ′, Y3 ′ after DR conversion is in the shape of a rectangular parallelepiped and touches the expansion table X ′, Y ′ is It becomes a rectangular parallelepiped a'-h 'shown in FIG. At this time, the four vertices a ′, b ′, c ′, h ′ and the two sides a′b ′, b′c ′ are in contact with the extension table X ′, and at the point of contact, eg1, eg2, es are the largest. Take dispersion error emax. In the images G1, G2, and S, there is a high possibility that the pixel sets that are in contact with the extension table X ′ after DR conversion are selected from the six points closest to the vertices a, b, c, h and the sides ab, bc. This is a singular point candidate of the actual input image.

  The maximum / minimum luminance detection unit 12 including the singular point extraction unit 121, the vector / scalar conversion unit 122, and the max / min determination unit 123 is a pixel set (g1, g2, s) of the images G1, G2, S before DR conversion. From the candidate points (singular points) that are most likely to have the maximum dispersion error after rounding, and determine the minimum and maximum luminance for each image G1, G2, S from the candidate points (S15, S16) .

  In the maximum / minimum luminance detection unit 12, first, the singular point extraction unit 121 extracts points close to the vertices a, b, c, h and the sides ab, bc as candidate points as described above. With reference to the extension table X ′, a singular point that is most likely to have the maximum dispersion error after the rounding process may be extracted as a candidate point.

  Since the candidate point extracted by the singular point extraction unit 121 is a vector, the vector / scalar conversion unit 122 extracts the g1, g2, and s scalars of the candidate points. The max / min determination unit 123 determines the maximum / minimum luminance of each image G1, G2, S from the scalar extracted by the vector / scalar conversion unit 122, and the maximum / minimum luminance (g1max, g1min), Output (g2max, g2min), (smax, smin).

  When the maximum and minimum luminance in each image G1, G2, S is expressed as n [p] g1, n [p] g2, n [p] s (p = a, b, c, h, ab, bc), The minimum and maximum brightness of each image G1, G2, S is (g1min, g1max) = (n [ab] g1, max (n [c] g1, n (h) g1)), (g2min, g2max) = ( n [bc] g2, max (n [a] g2, n [h] g2)), (smin, smax) = (n [h] s, max (n [a] s, n [b] s, n [c] s, n [ab] s, n [bc] s)). Thus, the mapping source information is obtained from the actual input images G1, G2, S.

  Mapping destination information is obtained from the assumed images Y1, Y2, and Y3 by the following method. The maximum / minimum luminance detection unit 13 including the contact extraction processing unit 131, the vector / scalar conversion unit 132, and the max / min determination unit 133 corresponds to the minimum / maximum luminance obtained by the maximum / minimum luminance detection unit 12. The minimum and maximum luminances of the points inscribed in the range restricted by X ′ are obtained (S13, S14).

  In the maximum / minimum luminance detection unit 13, first, the contact extraction unit 131 uses a pixel set (y1 ′, y2 ′, y3) for an image whose pixel histogram is uniform and the possible range is Y ′ (FIG. 3). ′) Is input, and a pixel set (y1 ′, y2 ′, y3 ′) in contact with the expansion table X ′ is extracted as a contact. In this case, a pixel set (y1 ′, y2 ′, y3 ′) that touches the four vertices a ′, b ′, c ′, h ′ and the two sides a′b ′, b′c ′ is extracted as a contact point. .

  Since the contact extracted by the contact extraction unit 131 is a vector, the vector / scalar conversion unit 132 extracts the scalar of each contact. The max / min determination unit 133 determines the maximum / minimum luminance from the scalar extracted by the vector / scalar conversion unit 132, and (y1′max, y1′min), (y2′max, y2′min), (y3 'Max, y3'min) is output as the luminance after DR conversion corresponding to the maximum and minimum luminances (g1max, g1min), (g2max, g2min), (smax, smin) of the images G1, G2, S.

  Here, when smin is converted to y3′min = 0 and smax is converted to y3′max, (g1min, g1max) = (y3′max-2emax, 128 + 3emax / 2), (g2min, g2max) = (y3′max-2emax, 128 + 3emax / 2), (smin, smax) = (0, y3′max). The fraction is rounded off.

  FIG. 4 is an explanatory diagram showing processing in the contact extraction unit 131, and FIGS. 4A and 4B show cross sections at s = 0 and s = y3′max in FIG. 3, respectively. Here, in the same figure (a), the square of the outer solid line corresponds to the plane efgh of the rectangular parallelepiped ah, and the shaded portion corresponds to the cross section of the tetrahedron defg at s = 0 (plane efg), the dotted line part Corresponds to a cross section of s = emax of the tetrahedron defg. In FIG. 6B, the shaded portion corresponds to the cross section of s = y3′max of the tetrahedron defg, and the dotted line corresponds to the cross section of s = y3′max-emax of the tetrahedron defg. To do.

  When the luminance of the image is 0 to 255, as is clear from FIG. 4, the point h ′ is represented by (128 + 3emax / 2,128 + 3emax / 2,0), and the point b ′ is represented by (y3′max− 2emax, y3′max-2emax, y3′max). The fraction is rounded off. The values 128 + 3emax / 2 and 128 + 3emax / 2 of the point h ′ in the g1 and g2 axis directions are y1′max and y2′max, respectively, and the values of the point b ′ in the g1 and g2 axis directions y3′max-2emax and y3 ′ Max-2emax is y1′min and y2′min, respectively.

  Here, the ratio of DR after conversion is set to DR (y1 ′): DR (y2 ′): DR (y3 ′) = 1: 1: r (cuboid). r and emax are given as y3′max = {(128 + 5emax / 3) r / (1 + r)}, and the maximum and minimum luminances (y1′max, y1) after DR conversion of each image G1, G2, S 'Min), (y2'max, y2'min), (y3'max, y3'min) are obtained and used as mapping destination information.

  The conversion coefficient determining unit 14 obtains a conversion coefficient for converting the mapping source information of the actual input image and the mapping destination information to the contact point of the assumed image. Maximum / minimum brightness (g1max, g1min), (g2max, g2min), (smax, smin) detected by the maximum / minimum brightness detection unit 12 is the maximum / minimum brightness (y1 ′ Max, y1′min), (y2′max, y2′min), and (y3′max, y3′min) are mapped to DR conversion coefficients (a1, b1) for g1, g2 and s, (a2, b2) and (a3, b3) are determined (S17).

  The luminance after DR conversion is generally given by the linear expression ax + b using the luminance x before conversion, so g1 ′, g2 ′ and s ′ are g1, g2, s and y1 ′, y2 respectively. It can be uniquely obtained from a linear equation passing through the minimum and maximum luminance points of ′, y3 ′. For example, the conversion equation for generating g1 ′ is (g1min, y1′min) = (n [ab] g1, y3′max-2emax), (g1max, y1′max) = (max (n [c] g1, n [h] g1), 128 + 3emax / 2).

  The DR conversion unit 15 performs DR conversion on each image G1, G2, S using the conversion formulas of the coefficients (a1, b1), (a2, b2), (a3, b3) determined by the conversion coefficient determination unit 14. (S18). The rounding processing unit 16 distributes the error to the luminance g1 ′, g2 ′, and s ′ so that the pixel group (g1 ″, g2 ″, s ″) after the rounding processing is in the table X ″. The process is applied (S19).

  As mentioned above, although embodiment was described, this invention is not limited to the said embodiment. For example, in the above embodiment, the number of tally images is two, but the present invention can also be applied to a case where there are n tally images (n ≧ 3). The pseudo density S ″ that can be expressed when the tally image is n sheets (n ≧ 3) is expressed by Expression (2).

  Max (0, g1 '' + g2 '' + ... + gn ''-255) ≤S''≤Min (g1 '', g2 '', ..., gn '') (2)

  In addition, when rounding processing is applied to the images G1 ′, G2 ′, S ′ that have been DR-converted by the DR converter 15 in FIG. 1, only the error component has a standard deviation of 5 × 5 to 20 × 20 elements. Apply the blurring filter (low pass filter) such as the above Gaussian filter, filter with the same element value, lens filter, etc., and add the error component with the low pass filter to the images G1 ', G2', S ' By doing so, the edge component of only the error component can be relaxed, so that the error component can be made less noticeable.

  In the above embodiment, the input signal is an image and the tally image is generated. However, even if the input signal is a signal other than an image, for example, an audio signal, the level of the signal is set. By treating it as corresponding to the luminance of the image in the above embodiment, the DR can be widened to make it easier to perceive the difference, and mixing of other signals can be suppressed. Even in this case, the number of input signals is not limited to three.

It is a functional block diagram which shows an example of the signal converter which concerns on this invention. It is a flowchart which shows the process in FIG. It is a figure where it uses for description of the process in FIG. It is explanatory drawing which shows the process in a singular point extraction part. It is explanatory drawing which shows notionally the decoding of a secret image by the superimposition of a tally image. It is a block diagram which shows the conventional structure which produces | generates a tally image using a brightness | luminance change part and a pixel arrangement | positioning determination part. It is explanatory drawing which shows the pseudo density which can be expressed with the combination of a monochrome pixel. It is explanatory drawing which shows the range of the pseudo density which can be expressed with the combination of a monochrome pixel. It is a functional block diagram which shows the specific structure of the conventional brightness | luminance change part. It is a functional block diagram which shows the other specific structure of the conventional brightness | luminance change part.

Explanation of symbols

11 ... Extended table generation unit, 12, 13 ... Maximum / minimum luminance detection unit, 14, 85, 93 ... Conversion coefficient determination unit, 15, 81, 83, 91, 511 ... DR conversion unit, 16 , 92, 512 ... rounding processing unit, 51 ... luminance change unit, 52 ... pixel arrangement determining unit, 82 ... polyhedron extraction unit, 84 ... check unit, 121 ... singular point extraction unit, 122,132 ... vector / scalar conversion unit, 123,133 ... max / min determination unit, 131 ... contact extraction unit, G1 to Gn ... original image for tally, S ... secret image, G1 '', G2 ″: Tally generating original image, S ″: Tally generating secret image, W1, W2: Tally image, T: Superimposed image

Claims (20)

In a signal conversion device that converts a plurality of signals in which the range of possible combinations is limited to a combination of signals whose dynamic range is converted and outputs the signal,
An extension table generating means for generating an extension table by expanding a table that restricts a range of combinations that can be taken by a plurality of signals to a range of combinations that the plurality of signals cannot take according to a preset value;
A signal conversion apparatus comprising: a signal conversion unit configured to convert a plurality of input signals so that a combination of the plurality of signals falls within a combination range in accordance with the restriction in the extension table.   The signal conversion apparatus according to claim 1, wherein the extension table generation unit generates the extension table by performing a reverse rounding process on the preset value and adding the value to the table.   The signal conversion means performs an exploratory process for determining a coefficient of a conversion formula in order to convert a plurality of input signals so that a combination thereof falls within a combination range that conforms to the restriction in the extension table. The signal conversion apparatus according to claim 1, wherein the signal conversion apparatus is a signal conversion apparatus.   The signal conversion means includes a first detection means for detecting a contact at each reference signal from a contact with respect to the extension table in a plurality of reference signals within a combination range in accordance with a restriction in the extension table; and a plurality of input signals A second detecting means for detecting a singular point associated with the contact detected by the first detecting means, and a conversion for mapping the singular point detected by the second detecting means to the contact detected by the first detecting means. Conversion coefficient determining means for determining a coefficient, and using the conversion coefficient determined by the conversion coefficient determining means, a plurality of input signals are combined within a range of combinations in which the combination conforms to the restrictions in the extension table. The signal conversion apparatus according to claim 1, wherein conversion is performed so that   The first detection unit extracts a contact point where a combination of the plurality of reference signals is in contact with a range constrained by the extension table, and performs vector-scalar conversion on the plurality of reference signals at the contact point. 5. The signal conversion apparatus according to claim 4, wherein a maximum value and a minimum value of each of the plurality of reference signals are detected as contact points.   The second detection means extracts singular points for the plurality of signals associated with the contact points detected by the first detection means, performs vector-scalar conversion on the plurality of signals at the singular points, and performs vector-scalar conversion. 6. The signal conversion apparatus according to claim 4, wherein a maximum value and a minimum value of each of the plurality of signals that have been detected are detected as singular points.   The signal conversion apparatus according to claim 1, wherein the plurality of signals are images.   The plurality of signals are a plurality of tally images and a secret image embedded in the tally image, and the secret image is decoded by superimposing the plurality of tally images. 8. The signal conversion device according to 7.   2. The apparatus according to claim 1, further comprising: rounding processing means for rounding the plurality of signals converted by the signal conversion means so that a combination thereof is a combination of a plurality of signals in accordance with the constraints of the table. Thru | or 8 the signal converter.   The signal conversion apparatus according to claim 9, wherein the rounding processing unit applies a low-pass filter to the error component in the rounding process. In order to convert a plurality of signals in which the range of possible combinations is constrained into a combination signal in which the dynamic range is converted, and output the computer,
Extended table generating means for generating an extended table by expanding a table that restricts a range of combinations that can be taken by a plurality of signals to a range of combinations that cannot be taken by the plurality of signals according to a preset value;
A program that functions as a signal conversion unit that converts a plurality of input signals so that a combination thereof falls within a combination range that conforms to the restrictions in the extension table.   The program according to claim 11, wherein the extension table generation unit generates the extension table by performing a reverse rounding process on the preset value and adding the result to the table.   The signal conversion means performs an exploratory process for determining a coefficient of a conversion formula in order to convert a plurality of input signals so that a combination thereof falls within a combination range that conforms to the restriction in the extension table. The program according to claim 11 or 12.   The signal conversion means includes a first detection means for detecting a contact at each reference signal from a contact with respect to the extension table in a plurality of reference signals within a combination range in accordance with a restriction in the extension table; and a plurality of input signals A second detecting means for detecting a singular point associated with the contact detected by the first detecting means, and a conversion for mapping the singular point detected by the second detecting means to the contact detected by the first detecting means. Conversion coefficient determining means for determining a coefficient, and using the conversion coefficient determined by the conversion coefficient determining means, a plurality of input signals are combined within a range of combinations in which the combination conforms to the restrictions in the extension table. The program according to claim 11 or 12, wherein the program is converted so that   The first detection unit extracts a contact point where a combination of the plurality of reference signals is in contact with a range constrained by the extension table, and performs vector-scalar conversion on the plurality of reference signals at the contact point. The program according to claim 14, wherein a maximum value and a minimum value of each of the plurality of reference signals are detected as contact points.   The second detection means extracts singular points for the plurality of signals associated with the contact points detected by the first detection means, performs vector-scalar conversion on the plurality of signals at the singular points, and performs vector-scalar conversion. The program according to claim 14 or 15, wherein the maximum value and the minimum value of each of the plurality of signals that have been detected are detected as singular points.   The program according to claim 11, wherein the plurality of signals are image signals.   The program according to claim 17, wherein the plurality of signals are a plurality of tally images and a secret image embedded in the tally image, and the secret image is decoded by superimposing the plurality of tally images. .   19. A rounding processing unit that rounds the plurality of signals converted by the signal conversion unit so that a combination of the signals is a combination of a plurality of signals according to the constraints of the table. The program described in.   The program according to claim 19, wherein the rounding processing means applies a low-pass filter to the error component during the rounding processing.

Images