JP2004215222A - Color transform processing device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To reduce processing costs and to realize a high speed processing by approximating the color transformation process using a DLUT through a combination of a one-dimensional LUT and a linear matrix. <P>SOLUTION: In a color transformation processing device for generating a plurality of output color signals based on a plurality of input color signals, the device comprises a first LUT processing unit 1 for performing a color transformation by using the same number of one-dimensional look up tables as the number of the output color signals with respect to input color signals, a matrix processing unit 2 for performing a calculation on data transformed by the first LUT processing unit 1 by a linear matrix, a second LUT processing unit 3 for performing the color transform on data transformed by the matrix processing unit 2 by using the one-dimensional look up table, and a parameter calculating unit 4 for calculating parameters of the first LUT processing unit 1, the matrix processing unit 2, and the second LUT processing unit 3. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

  The present invention relates to a color conversion processing device that generates an output color signal from an input color signal, and in particular, uses a multidimensional lookup table (hereinafter, referred to as “LUT”) in which input color values are associated with output color values. The present invention relates to a color conversion processing device that replaces an existing operation with a simple operation such as a one-dimensional LUT or a matrix.

  In a digital color image processing system, various image input / output devices having different color space characteristics are connected, and it is necessary to convert image data in an input color space into image data in an output color space with conversion characteristics according to these devices. . In addition, digital color images have a large amount of data and require low-cost and high-speed color conversion.

  In a conventional color conversion processing device using a multidimensional direct lookup table (hereinafter, referred to as “DLUT”), an output after color conversion processing corresponding to a plurality of discrete input color values (hereinafter, referred to as grid points). By storing the color values in a table and interpolating using the output color values of a plurality of grid points that are vertices of the area including the input color values to be converted, the relationship between the input color space and the output color space is non-linear. Even if there is, the color conversion processing with high accuracy is performed.

  In general, when the input color space is three-dimensional, such as RGB (red, green, blue), the area includes a cube (8 grid points), a tetrahedron (4 grid points), or a triangular prism (6 grid points). Used. As a method of interpolating a plurality of grid points, a method of performing linear interpolation a plurality of times is general. Among these, the area selection methods are characterized by accuracy, calculation speed, and the like. Therefore, there is a demand to switch the area selection method depending on the application.

  In general, various input / output device color spaces can be considered as objects to be color-converted. However, there is a demand for a color space device capable of flexibly supporting various color spaces.

  In response to the above various requirements, the apparatus described in Patent Document 1 uses a triangular prism with a small number of interpolations when emphasizing speed as a region and a cube with many grid points used for interpolation when emphasizing color conversion accuracy. It is configured to select and perform linear interpolation to provide a configuration that allows selection of two regions, a triangular prism or a cube.

  Further, in the image processing device described in Patent Literature 2, the color correction process more suitable for each region is performed by switching the interpolation calculation method in the color correction process using the LUT according to the type of the input image. .

JP-A-8-307684 JP 2001-352456 A

  However, the apparatus described in Patent Literature 1 has a problem in that DLUT processing is required regardless of whether a triangular prism or a cubic region is selected, resulting in a large amount of calculation. Also, since the interpolation method is fixed to linear interpolation, for example, it is possible to take a plurality of regions continuously in each axis direction of the input color space and use a non-linear interpolation method using a polynomial equation using the plurality of regions. In other words, there is a problem that a smooth gradation expression at the region boundary cannot be performed.

  Further, the apparatus described in Patent Document 2 has a problem that it takes a long time to determine the type of an input image, and the processing time becomes longer when an interpolation operation is performed.

  The present invention has been made in view of the above circumstances, and is a color conversion processing apparatus that achieves processing cost reduction and high speed processing by approximating processing that has been conventionally performed color conversion with a DLUT by a combination of a one-dimensional LUT and a linear matrix. The purpose is to provide.

  SUMMARY OF THE INVENTION The present invention is a color conversion processing device for achieving the above-described object. In a color conversion processing device for generating an output color signal from an input color signal, the output color signal A first LUT processing unit for performing color conversion using the same number of one-dimensional look-up tables as the number of the first LUT processing unit, a matrix processing unit for performing an arithmetic operation on the data converted by the first LUT processing unit using a linear matrix, and the matrix processing unit And a second LUT processing unit for performing color conversion on the data converted in step (1) using a one-dimensional lookup table.

  In the present invention, the color conversion processing can be simplified by approximating the processing of the DLUT with a one-dimensional LUT and a linear matrix, and the color conversion processing without complicated processing such as interpolation of the DLUT can be performed. Can be realized.

  According to the present invention, it is possible to simplify the color conversion process by approximating the DLUT process with the LUT and the linear matrix, and to realize the color conversion process without complicated processes such as the DLUT interpolation operation. Can be. Further, by replacing the color conversion processing by the 3 × 6 matrix with the calculation of the LUT and the 1 × 3 matrix, the color conversion processing can be performed with a simple configuration without deteriorating the image quality.

  Further, when replacing the DLUT with a polynomial, the processing cost can be reduced by a method of eliminating polynomial terms that are less affected by image quality degradation or a method of lowering the maximum order of the polynomial so that the image quality is within the allowable range of the user. it can.

  Further, since the division process into the unit areas performed by the DLUT is not required, a smooth gradation expression at the boundary can be realized. As described above, by replacing the complicated color conversion processing with the LUT and the matrix, the processing cost can be reduced and the processing speed can be increased.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing a color conversion processing device according to the present invention. The color conversion processing device includes a first LUT processing unit 1, a matrix processing unit 2, a second LUT processing unit 3, and a parameter calculation unit 4.

The first LUT processing unit 1 receives a plurality of input color signals, performs a color conversion process on one input signal by using a plurality of LUTs, and outputs the result to the matrix processing unit 2. The input signal includes, for example, RGB and L ^* a ^* b ^* . Also, one input signal is provided with the same number of LUTs as the output signal. For example, when the output signal is CMY, three LUTs are provided for one input signal, and when the output signal is CMYK, four LUTs are provided for one input signal.

  The matrix processing unit 2 performs an operation on the signal received from the first LUT processing unit 1 using a 1 × 3 matrix, and outputs the signal to the second LUT processing unit 3.

  The second LUT processing unit 3 receives a signal from the matrix processing unit 2, performs a color conversion process on one signal using one LUT, and outputs a signal. Output signals include, for example, CMY and CMYK.

  The parameter calculation unit 4 replaces or approximates a process of performing color conversion using a DLUT or a matrix with a calculation using a LUT and a matrix, and performs parameter conversion of a LUT or a matrix of a first LUT processing unit 1, a matrix processing unit 2, and a second LUT process. To determine.

  As shown in FIG. 6, the parameter calculation unit 4 includes a first LUT processing parameter calculation unit 41, a matrix processing parameter calculation unit 42, a second LUT processing parameter calculation unit 43, a parameter calculation control unit 44, and a DLUT polynomial conversion unit 45. Is provided.

  The DLUT polynomial conversion unit 45 converts the processing of the DLUT into a polynomial that can obtain the same color conversion processing result. The first LUT processing parameter calculation section 41 calculates parameters of the first LUT processing section 1 and sets values.

  The matrix processing parameter calculation section 42 calculates parameters of the matrix processing section 2 and sets values. The second LUT processing parameter calculation section 43 calculates the parameters of the second LUT processing section 3 and sets a value.

  The parameter calculation control unit 44 controls the parameter calculation of the first LUT processing parameter calculation unit 41, the matrix processing parameter calculation unit 42, and the second LUT processing parameter calculation unit 43.

Next, color conversion processing for converting RGB, R ² , G ² , and B ² input signals into CMY signals using a 3 × 6 matrix as shown in Expression (1) is performed by the color conversion processing apparatus of the present invention. A method for replacing with "" will be described.

  Equation (1) represents an operation for converting from the RGB color space to the CMY color space using a 3 × 6 matrix. Here, the value of C is represented by equation (2).

When the LUT of the first LUT processing unit 1 is represented by a second-order polynomial, the results of passing the RGB through the LUT are aR ² + bR + c, dG ² + eG + f, and gB ² + hB + i, respectively.

  Next, when the matrix processing unit 2 multiplies a 1 × 3 matrix, the value of C becomes as shown in Expression (3).

  Here, the coefficient of the 1 × 3 matrix is 1, and the constant term of the second-order polynomial is 0 (j = k = l = 1, c = f = i = 0).

  Next, the second LUT processing unit 3 outputs the value sent from the matrix processing unit 2 as it is, and the value of C is as shown in Expression (4).

  Similarly, the values of the output color signals M and Y can be expressed as Expressions (5) and (6), respectively.

From the equations (2) and (4), the matrix coefficients m _{1,1 to} m _{1,6 in the} equation (1) are uniquely obtained, and are as shown in the equation (7).

Similarly, in the case of the output color signals M and Y, the matrix coefficients m _{2,1 to} m _2,6 and m _{3,1 to} m _3,6 of the equation (1) are uniquely obtained, and the equations (8) and ( ₃ ) are respectively obtained. The result shown in (9) is obtained.

  As described above, the parameter calculation unit 4 completely replaces the color conversion by the 3 × 6 matrix by the LUT and the matrix processing.

  FIG. 3 is a flowchart illustrating a process of replacing a 3 × 6 matrix with an LUT and a matrix. That is, in step S11, the LUT of the first LUT processing unit is represented by a second-order polynomial, and in step S12, the 1 × 3 matrix of the matrix processing unit is set to (1,1,1).

  In step S13, the value sent from the matrix processing unit is output as it is by the second LUT processing unit. In step S14, each parameter is obtained from the equation of the output signal from the second LUT processing unit and the coefficient of the 3 × 6 matrix. In step S15, processing is performed in the order of the first LUT processing unit, the matrix processing unit, and the second LUT processing unit.

  Further, here, the coefficient of the 1 × 3 matrix is set to 1, and the constant term of the second-order polynomial is set to 0 (j = k = l = 1, c = f = i = 0). More accurate color conversion can be performed without changing the basic configuration.

Next, the input signals of RGB, R ² , G ² , B ² , R * G, G * B, B * R are converted into CMY signals using a 3 × 9 matrix as shown in Expression (10). A method of approximating the color conversion processing by the color conversion processing device of the present invention will be described.

  Equation (10) represents an operation for converting from the RGB color space to the CMY color space using a 3 × 9 matrix.

  First, the first LUT processing unit 1 multiplies each of the input color signals RGB by a constant. Since there are three output color signals, CMY, three LUTs are used for one input signal. Therefore, when the input color signal is R, there are three output values aR, a'R, and a "R.

  Next, the matrix processing unit 2 multiplies the values sent from the first LUT processing unit 1 by a 1 × 3 matrix. The output value of matrix 1 is aR + bG + cB, the output value of matrix 2 is a'R + b'G + c'B, and the output value of matrix 3 is a "R + b" G + c "B.

  Next, the second LUT processing unit 3 performs one-dimensional LUT processing that can be represented by a second-order polynomial. The output value of LUT4 becomes the value of C, and can be expressed as in equation (11).

  Since all of the same terms as the nine input values shown in equation (10) are available, color conversion processing using a 3 × 9 matrix can be approximated by a combination of a one-dimensional LUT and a 1 × 3 matrix. Examples of the approximation method include a least squares method.

  Similarly, in the case of M and Y, as shown in equations (12) and (13), all the same terms as the nine input values shown in equation (10) are prepared, so that the one-dimensional LUT and It can be approximated by a combination of × 3 matrices.

  As described above, it is possible to approximate the color conversion processing using the 3 × 9 matrix with a simple configuration of the one-dimensional LUT and the linear matrix.

  FIG. 4 is a flowchart illustrating a process of approximating a 3 × 9 matrix by a combination of an LUT and a matrix. That is, in step S21, the LUT of the first LUT processing unit is represented by a constant multiple, and in step S22, the 1 × 3 matrix of the matrix processing unit is set to (1, 1, 1).

  In step S23, the LUT of the second LUT processing unit is represented by a second-order polynomial. In step S24, each parameter is approximated from the expression of the output signal from the second LUT processing unit and the coefficient of the 3 × 9 matrix. , A first LUT processing section, a matrix processing section, and a second LUT processing section.

  Next, an example in which a three-dimensional DLUT is approximated by a one-dimensional LUT and a linear matrix will be described. First, as a simple example, a method of replacing a two-dimensional DLUT of three grid points with a combination of an LUT and a matrix operation will be described.

  First, the values of the grid points of the DLUT are replaced with polynomials. In the example of the two-dimensional DLUT having three grid points shown in FIG. 2A, when the input signal is x and y and the output signal is z, it can be expressed by a polynomial as shown in Expression (14).

  Similarly, in the case of a 17-lattice DLUT, it can be represented by a 16th-order polynomial. Next, in the three-dimensional DLUT with three grid points shown in FIG. 2B, the value of C can be represented by a polynomial such as Expression (15).

  The highest order of each of RGB is a second-order polynomial. Also, in the case of M and Y, the highest degree of each of RGB becomes a second-order polynomial. In the case of a three-dimensional DLUT having N (N is a natural number) grid points, the value of C can be represented by a polynomial in which the highest order of each of RGB is (N-1) th order as shown in Expression (16).

  Similarly, the values of M and Y can be represented by a polynomial of the highest order being N-1.

Next, a method of eliminating a polynomial term that is less affected by image quality degradation will be described. First, there is a method in which the coefficients of the N grid point three-dimensional DLUT are replaced with an (N-1) _-th order polynomial by equation (16), and terms whose coefficients a _{i, j, k of} the polynomial are close to 0 are omitted.

That is, all the terms whose coefficients a _{i, j, k} are equal to or less than a predetermined threshold are omitted, and only the terms in which the coefficients a _{i, j, k} that easily affect the image quality have a large value (exceed the threshold) are left. In this case, instead of determining the threshold value, a method of omitting a predetermined number of polynomial terms from terms having small coefficients a _{i, j, k} may be used.

  Further, there is a method in which the number of inflection points is checked from the DLUT coefficient, and the maximum order of the polynomial is thereby limited. For example, when replacing the 17-lattice DLUT with a polynomial, if the number of inflection points is k, the polynomial of the highest order is replaced by a polynomial of order k + 1, so that useless terms can be omitted.

  Even when the number of inflection points is k, not only the highest order of the polynomial to be replaced is limited to the (k + 1) th order, but also the number of terms to be reduced to the kth or lower order may be increased. In this case, the order can be reduced so that the image quality degradation falls within the allowable range.

  Another method is to limit the polynomial to be replaced to the order or less specified by the user. When the user emphasizes the image quality, the order of the polynomial may be increased, and when the processing time is emphasized, the order of the polynomial may be decreased.

  Furthermore, there is a method of preparing a plurality of processes for replacing with polynomials of different orders so that the user can set the image quality for a plurality of images in the same document. In this case, the user is allowed to select a plurality of modes from a high image quality mode to a low image quality mode, and when the high image quality mode is selected, the image quality is improved using a polynomial having a large order, and the low image quality mode is changed. When selected, the processing time is reduced by using a polynomial of a small order.

  As described above, by a method of omitting a term whose coefficient is close to 0 or a method of limiting the highest order of a polynomial, it is possible to omit a term that is less affected by image quality degradation and to simplify the processing.

  Next, a method of replacing the polynomial expressing the processing of the three-grid point two-dimensional DLUT shown in Expression (14) with an LUT and a matrix will be described. First, the first LUT processing unit 1 performs one-dimensional LUT processing that can be represented by a second-order polynomial. Since there are three output color signals, CMY, three LUTs are used for one input signal.

When the input color signal is RGB, the output of the LUT 11 is aR ² + bR + c, the output of the LUT 21 is dG ² + eG + f, and the output of the LUT 31 is gB ² + hB + i. Similarly, the output of the LUT ¹² is a'R ² + b'R + c ', the output of the LUT 13 is a "R ² + b" R + c ", the output of the LUT 22 is d'G ² + e'G + f', The output of the LUT 23 is d "G ² + e" G + f ", the output of the LUT 32 is g'B ² + h'B + i ', and the output of the LUT 33 is g" B ² + h "B + i".

Next, the matrix processing unit 2 performs a 1 × 3 matrix operation. Assuming that all parameters of the 1 × 3 matrix are 1, the output signal of matrix 1 is aR ² + bR + dG ² + eG + gB ² + hB + j (j = c + f + i).

  Next, the second LUT processing unit 3 performs one-dimensional LUT processing that can be represented by a third-order polynomial. The output of LUT4 is as shown in equation (17).

  Since all terms of the polynomial shown in equation (15) appear in equation (17), the coefficients of each term can be approximated. Similarly, in the case of M and Y, the coefficient of each term can be approximated.

  The parameter calculation unit 4 determines the LUT and matrix parameters of the first LUT processing unit 1, the matrix processing unit 2, and the second LUT processing unit 3 in this manner.

  FIG. 5 is a flowchart illustrating a process of approximating a three-grid point two-dimensional DLUT by a combination of an LUT and a matrix. First, in step S31, the three-grid point two-dimensional DLUT is replaced with a polynomial. Next, in step S32, the LUT of the first LUT processing unit is represented by a second-order polynomial. Subsequently, in step S33, the 1 × 3 matrix of the matrix processing unit is set to (1, 1, 1), and in step S34, the LUT of the second LUT processing unit is represented by a three-dimensional polynomial.

  After that, in step S35, parameters are approximated from the expression of the output signal from the second LUT processing unit and a polynomial expression obtained by replacing the three lattice point two-dimensional DLUT, and in step S36, the first LUT processing unit, the matrix processing unit, and the second LUT processing unit The processing is performed in the order of.

  Here, a method of approximating a three-lattice two-dimensional DLUT with an LUT and a matrix has been described. However, a more complicated DLUT such as a 17-lattice three-dimensional DLUT can be approximated by a similar method.

  As described above, the present invention has been described using the embodiments, but the technical scope of the present invention is not limited to the scope described in the above embodiments. Various changes or improvements can be added to the above-described embodiment, and embodiments with such changes or improvements are also included in the technical scope of the present invention. In addition, the above embodiments do not limit the claimed invention, and not all combinations of the features described in the embodiments are necessarily essential for solving the invention.

For example, the input color signal is not limited to three signals such as RGB and L ^* a ^* b ^* , but may be any n signals. When the number of input signals is n, the number of LUTs in the first LUT processing unit 1 is n * the number of output signals.

  When replacing the color conversion processing by the 3 × 6 matrix with the LUT and the matrix processing, the input signal of the 3 × 6 matrix is not limited to the combination of the three input values and the respective square values. A combination of n-th power values may be used.

1 is a block diagram illustrating a color conversion processing system including a color conversion processing device according to the present invention. It is a figure which shows the outline of a three-grid point two-dimensional DLUT and a three-grid point three-dimensional DLUT. 9 is a flowchart of a process of replacing a 3 × 6 matrix with a combination of an LUT and a matrix. It is a flowchart of the process which approximates a 3x9 matrix by the combination of a LUT and a matrix. It is a flowchart of a process of approximating a three-grid point two-dimensional DLUT by a combination of an LUT and a matrix. FIG. 3 is a block diagram illustrating a configuration of a parameter calculation unit.

Explanation of reference numerals

  DESCRIPTION OF SYMBOLS 1 ... 1st LUT processing part, 2 ... matrix processing part, 3 ... 2nd LUT processing part, 4 ... parameter calculation part, 41 ... 1st LUT processing parameter calculation part, 42 ... matrix processing parameter calculation part, 43 ... 2nd LUT processing parameter calculation Unit, 44: parameter operation control unit, 45: DLUT polynomial conversion unit

Claims (7)

In a color conversion processing device that generates a plurality of output color signals based on a plurality of input color signals,
A first LUT processing unit that performs color conversion on one input color signal using the same number of one-dimensional lookup tables as the number of output color signals;
A matrix processing unit that performs an arithmetic operation on the data converted by the first LUT processing unit using a linear matrix;
A second LUT processing unit that performs color conversion on the data converted by the matrix processing unit using a one-dimensional lookup table;
A color conversion processing device comprising: a first LUT processing unit, a matrix processing unit, and a parameter calculation unit that calculates parameters of the second LUT processing unit. When the input color signal is composed of a plurality of color signals and a value obtained by raising each of the color signals to the nth power, the first LUT processing unit performs a color conversion process to m output color signals by an n-th order polynomial in the first LUT processing unit. The color conversion is performed using a one-dimensional lookup table that can be expressed as follows, the matrix processing unit performs an operation based on a 1 × m matrix, and the second LUT processing unit multiplies it by a constant to replace the processing. The color conversion processing device according to claim 1, wherein The first LUT processing unit performs color conversion on one input color signal using a one-dimensional look-up table with m output signals,
After performing an operation using a 1 × m matrix in the matrix processing unit, the second LUT processing unit performs conversion using a one-dimensional lookup table, thereby approximating color conversion processing using a three-dimensional lookup table. The color conversion processing device according to claim 1. Performing a color conversion process into m output color signals by a matrix operation when the input color signal includes a plurality of color signals, a value obtained by raising each signal to the nth power, and a value obtained by multiplying each signal; The 1LUT processing unit performs color conversion using a one-dimensional look-up table that can be expressed by an n-th order polynomial, performs an operation using a 1 × m matrix in the matrix processing unit, and then uses the one-dimensional lookup table in the second LUT processing unit. The color conversion processing device according to claim 1, wherein the conversion results approximate the calculation result. 2. The color conversion according to claim 1, wherein a term whose coefficient is close to 0 is omitted when the three-dimensional lookup table of N grid points (N is a natural number) is replaced by an N-1 degree polynomial by the first LUT processing unit. Processing equipment. When replacing the three-dimensional look-up table with a polynomial by the first LUT processing unit, the highest order of the polynomial to be replaced is determined according to the number of inflection points of the three-dimensional look-up table. The color conversion processing device as described in the above. 2. The color conversion processing device according to claim 1, wherein an arbitrary selected polynomial term is omitted when the first LUT processing unit replaces the three-dimensional lookup table with a polynomial.

Images