JP2000196903A - Method and device for color conversion - Google Patents

Abstract

PROBLEM TO BE SOLVED: To perform color conversion at high speed by performing arithmetic interpolation in parallel in the course of a color converting process which is performed by also using arithmetic interpolation for a lookup table. SOLUTION: The lattice point data corresponding to the maximum value, intermediate value, and minimum value of the low-order bit data of inputted color data indicating each term of an arithmetic interpolation formula and the relations among the values are respectively stored in prescribed registers 402, 404, and 406 are registers 403, 405, and 407. Consequently, the appearance of a condition branching process resulting from the magnitudious relation among the values in the succeeding sum of products operation (registers 408, 409, and 410) can be prevented. Therefore, parallel processing becomes possible, because interpolation can be made through one unitary operation.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a color conversion method for converting a plurality of (multi-dimensional) color signals into different color signals by using a lookup table (hereinafter also referred to as "LUT") together with an interpolation operation. And a color conversion device.

[0002]

2. Description of the Related Art Non-linear conversion such as gamma conversion and log conversion of a digitized image signal is often performed using an LUT. This is because if the above-described non-linear conversion is to be obtained by an arithmetic circuit, the calculation becomes relatively complicated and the processing time becomes long. Further, such an LUT can be realized if a memory having a capacity of 256 bytes is used, for example, when performing some kind of non-linear conversion on an 8-bit video signal. Incidentally, since the conversion for the 8-bit video signal is for converting one image signal into another image signal having another property, the LUT used there is called a one-dimensional LUT.

[0003] With the recent remarkable progress in the desktop publishing (hereinafter abbreviated as "DTP") environment, anyone can easily handle color images. Input devices for color images in DTP are mainly scanners, video cameras, and the like, and output devices are various color printers and the like using respective systems such as inkjet, dye thermal sublimation, and electrophotography. Each of these color input / output devices has a unique color space. For example, when color image data obtained by a certain scanner is directly transferred to another color printer and an image is output, The colors rarely match the colors of the original image read by the scanner. In order to match the two colors, a process of converting a color space of a so-called input device (input device) into a color space of an output device (output device) is required (hereinafter, this process is referred to as a “color conversion process”). ").

The color conversion processing specifically refers to a whole series of image processing such as input γ correction, luminance / density conversion, masking, black generation, UCR, output γ correction, or a fixed processing therein. However, the three colors of the input device (generally,
By simultaneously referring to image signals of three colors of Red, Blue, and Green (hereinafter, abbreviated as "R", "G", and "B"), three colors (e.g., cyan, magenta, and yellow) on the output device side are referred to. Alternatively, the image signal is converted into an image signal of four colors (for example, cyan, magenta, yellow, and black).

The process of converting the three-color image signal of the input device into one of a plurality of colors of the output device is performed by an LUT.
If one image signal is represented by 8 bits, the L of input 24 bits and output 8 bits is used.
It becomes a UT, and requires a memory of 16 M (mega) bytes. In addition, the above memory is required for the number of colors of the output device.
The capacity is as large as M bytes, which is not practical in terms of cost.

Therefore, in the color conversion processing using the LUT,
It is common to use an interpolation process together, thereby achieving a significant reduction in the memory capacity for the LUT.

A conventional example of a color conversion process using an interpolation operation for input / output correspondence using an LUT will be described below.

First, as input color data, for example, R,
G, 3 one color signal B (each color n + m bits), respectively ^{Xi = Xh · 2 m + Xf} , Yi = Yh · 2 m + Yf, Z
i = Zhｉ2 ^m + Zf. Here, Xh, Yh,
Zh represents the upper n-bit signal of each of the Xi, Yi, Zi data, and Xf, Yf, Zf represents the lower m-bit signal of each of the Xi, Yi, Zi data.

In the LUT, Xh = 0, 1, 2,.
^2n- 1, Yh = 0, 1, 2,..., ^2n- 1, Zh
= 0, 1, 2,..., 2 ⁿ −1, the converted color data is stored as grid point data for each of the combinations (2 ³ⁿ ways). The data is read by a 3n-bit address signal connecting Xh, Yh, and Zh.

When the lower m bits of each of the input color signal data (Xi, Yi, Zi), that is, Xf, Yf, Zf are all "0", the grid point data read from the LUT by the above address signal. Is the color data after conversion as it is. Otherwise, the interpolation processing is performed according to the values of Xf, Yf, and Zf.

There are several methods for this interpolation calculation processing, depending on how many pieces of grid point data are used and on what kind of relation grid point data is used. Generally, when a large amount of grid point data is used, the interpolation accuracy is improved, but the interpolation calculation processing is complicated and the processing time is long. As an interpolation method capable of performing relatively simple arithmetic processing without significantly lowering the interpolation accuracy, a method called four-point interpolation has been conventionally known. This performs interpolation processing using four grid point data, and is disclosed, for example, in Japanese Patent Publication No. 58-16180. The outline will be described below.

When all three lower bit signals Xf, Yf, Zf are not 0, ie, Xf ≠ 0, Yf ≠ 0, Zf
In the case of 色 0, the color data before conversion is, as shown in FIG.
It is located in a cube formed by eight grid points. This cube forms a partial interpolation space designated by the upper bit signals (Xh, Yh, Zh), and has a grid point interval of 2 ^m and lower bit signals Xf, Yf, Zf.
Defines one of the points obtained by dividing each grid point of the cube into 2 ^m pieces.

The cube shown in FIG. 1 is divided into three planes (Xf = Y
f, a plane defined by Yf = Zf, and a plane defined by Zf = Xf), six tetrahedrons are defined, and each tetrahedron is, for example, as shown in FIG. Has four grid points. The point (Xi, Y) on the above-described partial interpolation space defined by the input color data to be converted
i, Zi) is included in any one of these six tetrahedrons (the boundary surface is assumed to be included in any one of the tetrahedrons sharing the boundary). Input color data (Xi, Y
i, Zi) belongs to which tetrahedron, Xf, Y
It is determined by the magnitude relationship between f and Zf. For example, X
When there is a relationship of f>Yf> Zf, the input color data
And (Xh, Yh, Zh), (Xh + 1, Yh, Z)
h), (Xh + 1, Yh + 1, Zh), (Xh + 1, Y
h + 1, Zh + 1). Data associated with each grid point coordinate by the LUT, that is, grid point data is represented as dd (X coordinate, Y coordinate, Z coordinate), and LUT is applied to the input color data (Xi, Yi, Zi).
When the data after interpolation using the expressed and P (Xi, Yi, Zi) , the data P after the interpolation, P (Xi, Yi, Zi ) = 2 -m {(2 m -Xf) · dd ( Xh, Yh, Zh) + (Xf-Yf) ・ dd (Xh + 1, Yh, Zh) + (Yf-Zf) ・ dd (Xh + 1, Yh + 1, Zh) + Zf ・ dd (Xh + 1 , Yh + 1, Zh + 1)｝ (1) It goes without saying that the above-mentioned interpolation formula is different for each of the above-mentioned six tetrahedrons.

As described above, the interpolation operation is performed, as shown in the equation (1), between the four grid point data corresponding to each tetrahedron and the coefficient data generated from the lower bit signal of the input color data. It consists of product and addition of four terms.

[0015]

However, as is apparent from the above equation (1), the interpolation calculation processing includes:
(Xh, Yh, Zh),..., (Xh + 1, Yh + 1,
For example, a comparison process such as Zh + 1) for calculating an address necessary for reading grid point data from the LUT and determining a magnitude relationship between Xf, Yf, and Zf is required. Depending on the magnitude relation, the grid point data to be used and the processing contents differ,
For this reason, for example, there is a problem that it is difficult to perform an interpolation operation on a plurality of pixels in parallel.

That is, the calculation accuracy of the above equation (1) is 16
A bit is enough. Specifically, the number of bits of the grid point data is 8 to 10 bits, and the input lower bit data X
The number of bits of f, Yf, and Zf is 3 to 5 bits. In most other general image processing, 16-bit operation accuracy is sufficient, and in some cases, 8-bit operation accuracy may be sufficient.

On the other hand, there has been a remarkable improvement in the performance of recent CPUs, and it is becoming common for the general-purpose registers to have a length of 32 bits or 64 bits. Such a register length is over-spec in the image processing related to the interpolation,
This means that the performance of U is not effectively utilized.

On the other hand, information exchanged regarding information processing is changing from the former text base to the image base. In such an environment, JPEG or MPEG
Image compression / decompression processing becomes more important, and high-speed processing is also required. In order to meet such needs, a CPU having a function corresponding to multimedia has been provided. That is, in these CPUs, a 64-bit register is stored in a 32-bit ×
By using a configuration of 2, 16 bits × 4 or 8 bits × 8, a plurality of low-precision operations can be processed in parallel at a time, or a product-sum operation can be performed at a time.

For example, it is well known that the above-described multimedia-compatible functions are effectively used in the above-described processing of JPEG, MPEG, and the like. In addition, filtering that performs the same arithmetic processing on each pixel in the image,
Parallel processing is easy, and the multimedia-compatible function can be effectively used.

However, the above-described interpolation operation involving conditional branch processing such as the magnitude relationship of lower-order bit data cannot effectively utilize the multimedia-compatible function, and it is relatively difficult to speed up the processing. There was a problem that it was difficult.

The present invention has been made to solve such a problem, and an object of the present invention is to make full use of, for example, a CPU having a multimedia-compatible function to perform parallel processing of an interpolation operation. An object of the present invention is to provide a color conversion method and a color conversion device that can be realized.

[0022]

According to the present invention, there is provided:
A color conversion method for performing color conversion by using an interpolation operation together with a look-up table, wherein, among input color data, data indicating the position of the input color data in an interpolation space by its magnitude relation, Among them, the grid point data read from the look-up table based on the data indicating the address of the look-up table and the data indicating the position according to the magnitude relation are stored in a plurality of registers predetermined according to the magnitude relation. Storing, and performing an interpolation operation using data stored in the register according to a predetermined interpolation operation expression among the plurality of registers.

Also, the present invention is a color conversion apparatus for performing color conversion by using an interpolation operation together with a look-up table, wherein data indicating the position of the input color data in the interpolation space by the magnitude relation among the input color data is provided. In the input color data, grid point data read from the look-up table based on data indicating the address of the look-up table and data indicating the position according to the magnitude relation is determined in advance according to the magnitude relation. Storage means for storing data in a plurality of registers, and arithmetic means for performing an interpolation operation using data stored in the registers in accordance with a predetermined interpolation operation expression among the plurality of registers. And

According to the above arrangement, for example, the lower-order bit data indicating the position in the interpolation space of the input color data in the interpolation space and the grid point data read out from the look-up table in accordance with the size relation. Are stored in a plurality of registers predetermined according to the magnitude relation, and an interpolation operation is performed using data stored in the registers according to an interpolation operation expression predetermined among these registers. Therefore, the interpolation formulas that differ depending on the magnitude relation or the calculations of each term of the formulas can be unified into one formula that defines the calculation relationship between the registers. The calculation and calculation of each term of the calculation expression can be performed in parallel.

[0025]

Embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 3 is a block diagram showing an outline of the processing and hardware configuration performed in the color conversion processing of the first to fourth embodiments described below as one embodiment of the present invention.

As shown in the figure, in the color conversion processing according to the embodiment of the present invention, a register group 301 used for interpolation calculation and yellow (Y) and magenta (M) storing grid point data are described later. , Cyan (C), and black (K) lookup tables (hereinafter, referred to as “LUTs”) 302Y, 302M, 302
C, 302K.

In the size discriminating unit 303, for example, R, G,
Among the 8-bit input color data Xi, Yi, and Zi for each of the B colors, the magnitude relationship between them is determined based on m-bit lower-order bit data Xf, Yf, and Zf. In addition, the address generation unit 304 generates an address for reading grid point data related to four-point interpolation from the LUT with reference to the magnitude relation information based on the n-bit upper bit data of the input color data. . Note that the magnitude discrimination unit 303
The address generation unit 304 may be configured as software executed by a CPU, or may be configured as hardware.

Based on each of the four addresses generated by the address generator 304, the LUT 30 of each color
4 from 2Y, 302M, 302C, 302K
Two pieces of grid point data are read out and stored in predetermined registers of the register group 301 as described later. The register group 301 includes lower-order bit data Xf, Yf, Zf
Are also input according to the magnitude relation and stored in a predetermined register as described later.

In the register group 301, as will be described in detail in the following embodiments, an interpolation operation is performed by performing a subtraction, a product-sum operation, and a shift between registers under the control of the CPU. Finally, the converted color data P can be obtained.

Here, the interpolation formula used in each of the embodiments described below will be described. First, in order to simply express the interpolation formula, the grid point data located at the vertices of the interpolation cube of interest are represented by A, B, C, D, E,
Expressed as F, G, H. The interpolation operation expression is based on the magnitude relationship between the lower-order m-bit data Xf, Yf, and Zf in each of the three input color data Xi, Yi, and Zi, that is, the input color data in the drawing as shown in FIG. Xi,
Depending on which tetrahedron Yi and Zi belong to, they are represented by any of the following six different interpolation formulas.

When Xf ≧ Yf ≧ Zf, P = 2− ^m ｛(2 ^m −Xf) · A + (Xf−Yf) · B + (Yf−Zf) · D + Zf · HH (a-1) When Xf ≧ Zf ≧ Yf, P = 2 ^−m ｛(2 ^m −Xf) · A + (Xf−Yf) · B + (Yf−Zf) · F + Zf · H…… (a-2) Zf ≧ Xf When ≧ Yf, P = 2 ^−m ｛(2 ^m −Xf) · A + (Xf−Yf) · E + (Yf−Zf) · F + Zf · H… (a-3) When Zf ≧ Yf ≧ Xf In this case, P = 2 ^-m ｛(2 ^m -Xf) ・ A + (Xf-Yf) ・ E + (Yf-Zf) ・ G + Zf ・ H｝ (a-4) If Yf ≧ Zf ≧ Xf, P = 2 ^-m ｛(2 ^m -Xf) ・ A + (Xf-Yf) ・ C + (Yf-Zf) ・ G + Zf ・ H｝ (a-5) When Yf ≧ Xf ≧ Zf, P = 2 ^{− m} ｛(2 ^m -Xf) ・ A + (Xf-Yf) ・ C + (Yf-Zf) ・ D + Zf ・ H｝ (a-6) Also, the above six interpolation formulas simply transform them. Can also be expressed as follows:

When Xf ≧ Yf ≧ Zf, P = 2− ^m ｛2 ^m · A + Xf · (BA) + Yf · (DB) + Zf · (HD)｝ (b-1) Xf ≧ Zf ≧ In the case of Yf, P = 2 ^-m ｛2 ^m・ A + Xf ・ (BA) + Zf ・ (FB) + Yf ・ (HF)… (b-2) When Zf ≧ Xf ≧ Yf, P = 2 ^-m ｛2 ^m · A + Zf · (EA) + Xf · (FE) + Yf · (HF)… ... (b-3) When Zf ≥ Yf ≥ Xf, P = 2 ^-m ｛2 ^m · A + Zf ・ (EA) + Yf ・ (GE) + Xf ・ (HG)… (b-4) When Yf ≧ Zf ≧ Xf, P = 2 ^-m ｛2 ^m・ A + Yf ・ (CA) + Zf · (GC) + Xf · (HD)… ... (b-5) When Yf ≥ Xf ≥ Zf, P = 2- ^m ｛2 ^m · A + Yf · (CA) + Xf · (DC) + Zf (HD)｝ (b-6) Similarly, it can be expressed as follows by further modification.

When Xf ≧ Yf ≧ Zf, P = 2− ^m ｛2 ^m · A + Xf · (BA) + Yf · (DB) + Zf · (HD)｝ (c-1) Xf ≧ Zf ≧ In the case of Yf, P = 2 ^-m ｛2 ^m · A + Xf · (BA) + Yf · (HF) + Zf · (FB)｝ (c-2) When Zf ≧ Xf ≧ Yf, P = 2 ^-m ｛2 ^m · A + Xf · (FE) + Yf · (HF) + Zf · (EA)… ... (c-3) When Zf ≥ Yf ≥ Xf, P = 2 ^-m ｛2 ^m · A + Xf · (HG) + Yf · (GE) + Zf · (EA)… (c-4) When Yf ≧ Zf ≧ Xf, P = 2- ^m ｛2 ^m · A + Xf · (HG) + Yf · (CA) + Zf · (GC)｝ ... (c-5) When Yf ≥ Xf ≥ Zf, P = 2- ^m ｛(2 ^m · A + Xf · (DC) + Yf · (CA) + Zf · (HD)… (c-6) (c-1) to (c-6) simply denote the second to fourth terms in (b-1) to (b-6) by Xf, Yf, and Zf. In the following order.

The method for realizing parallel processing is described in (a-
1) to (a-6), (b-1) to (b-6), (c-
The realization method differs depending on which of the expressions 1) to (c-6) is used, but in the following first and second embodiments,
In the third embodiment, based on equations (b-1) to (b-6), based on equations (a-1) to (a-6),
In the embodiment, a method of performing an interpolation operation using a register based on the equations (c-1) to (c-6) will be described.

(First Embodiment) As described above, in this embodiment, a method for realizing parallel processing will be described based on equations (b-1) to (b-6). The outline of the parallel processing of the present embodiment is as follows.

(D-1) Using an arithmetic register capable of parallel operation of the CPU described above with reference to FIG.
Bits, and these are used as 16 bits × 4 registers.

(D-2) The input color data to be color-converted is
Three colors (Xi, Yi, Zi) are used.

(D-3) The color conversion processing of four consecutive pixels of the image relating to the processing is performed in parallel.

In the following description, the subscripts 1, 2, 3 and 4 indicate which one of the four consecutive pixels relates to the data in order to distinguish the above four pixels. For example, the grid point data A shown in FIG. 1 is described as A1, A2, A3, A4.

The color conversion processing of this embodiment is performed in the following procedure. FIGS. 4A to 4J are diagrams showing registers used in this procedure and their contents.

(1) For each of the four pixels, a partial interpolation space (cube, see FIG. 1) is determined from the upper bit data of the input color data Xi, Yi, Zi, and grid point data located at A in each cube dd (Xh, Yh, Zh) is read from the LUT shown in FIG. 3 and, as shown in FIG.
As 6-bit data, they are concatenated in the order of A1, A2, A3, and A4, and stored in the register 401 capable of parallel operation of the CPU as described above.

(2) Next, the maximum value of the lower-order bit data of the input color data is obtained for each pixel, and as shown in FIG.
FIG. 4B shows an example in which the maximum value of the first pixel is Yf, the maximum value of the second pixel is Xf, the maximum value of the third pixel is Zf, and the maximum value of the fourth pixel is Xf.

(3) Further, if the maximum value is Xf in accordance with the maximum value for each pixel, the grid point data B = dd (Xh + 1, Yh,
Zh), the lattice point data C = dd when the maximum value is Yf
(Xh, Yh + 1, Zh), and when the maximum value is Zf, grid point data E = dd (Xh, Yh, Zh + 1) is selected and read from the LUT, and as shown in FIG. Are connected to each other and stored in the register 403.
In the example shown in FIG. 4C, C1, B2, and C3 correspond to the lower-order bit data having the maximum value for each pixel shown in FIG.
E3 and B4 are stored in the register 403.

(4) Similarly, the median value of the lower-order bit data of the input color data is obtained for each pixel, and four of them are connected as shown in FIG.
The example shown in FIG. 4D shows a case where the median value of the first pixel is Zf, the median value of the second pixel is Yf, the median value of the third pixel is Xf, and the median value of the fourth pixel is Zf. ing.

(5) Next, the maximum value and the median value are represented by Xf, Y
Based on the combination of f and Zf, grid point data D = dd (Xh + 1, Yh + 1, Zh), F = d
d (Xh + 1, Yh, Zh + 1), G = dd (Xh, Y
h + 1, Zh + 1) for each pixel,
The data is read from the LUT, and the four are connected as shown in FIG. For example, FIG.
In the content of the first pixel shown in (e), G is selected corresponding to the combination of Yf and Zf shown in FIGS. 4B and 4D. Similarly, for the second to fourth pixels, D2, F3,
F4 is stored in the register 405.

(6) Similarly, the minimum value of the lower bit data of the input color data is obtained for each pixel, and as shown in FIG.
The example shown in FIG. 4F shows a case where the minimum value of the first pixel is Xf, the minimum value of the second pixel is Zf, the minimum value of the third pixel is Yf, and the minimum value of the fourth pixel is Yf. ing.

(7) Finally, for each pixel, grid point data H = dd (Xh + 1, Yh + 1, Zh +) in the partial interpolation space
As shown in FIG. 4 (g), 4)
H2, H3, and H4 are stored in registers.

As described above, the processes (1) to (7) are based on the magnitude relation of the lower bit data of the input color data in the partial interpolation space defined by the input color data. This is a process for determining a tetrahedron used for interpolation calculation, or a process for preparing grid point data of the tetrahedron. In the present embodiment, it can be said that such a process is replaced by a process of storing in a predetermined register.

(8) After the above storage processing, the register 4 shown in FIG.
The value of the register 401 shown in FIG. 4A is subtracted from the value of 03, then multiplied by the value of the register 402 shown in FIG. 4B, and held in the register 408 (FIG. 4H).

(9) Similarly, the value of the register 403 shown in FIG. 4C is subtracted from the value of the register 405 shown in FIG. 4E, and then the value of the register 404 shown in FIG. And hold it in the register 409 (FIG. 4)
(I)).

(10) Further, the value of the register 405 shown in FIG. 4E is subtracted from the value of the register 407 shown in FIG. 4G, and the result is multiplied by the value of the register 406 shown in FIG. Is held in the register 410 (FIG. 4 (j)).

By the above processes (8) to (10),
By calculating the difference between the registers and then calculating the product between the registers, the second to fourth terms in parentheses in the above formulas (b-1) to (b-6) are obtained. Become.

(11) Next, the value of the register shown in FIG. 4A is shifted to the left by m bits in each of the 16 bits, and the parentheses in the above equations (b-1) to (b-6) are obtained. First
The term, ie, 2 ^mA, is formed and determined above in FIG.
Registers 4 indicated by (h), (i) and (j) respectively
Add the values 08, 409 and 409. Further, the result is shifted right by m bits. As a result, color data P for, for example, cyan after conversion for four pixels can be obtained simultaneously by performing the interpolation operations for four pixels in parallel.

As described above, in the present embodiment, the number is related to the lattice point data and the magnitude relation thereof according to the magnitude relation (maximum value, median value, minimum value) of the lower bit data of the input color data. Since the registers for storing the lower-order bit data are determined, only the processing ((1) to (7)) for storing the respective data in these predetermined registers in accordance with the magnitude relation is performed. It is possible to prevent the magnitude relationship or the determination process from appearing in the interpolation calculation (the processes of (8) to (11)) itself. This is the same as (8) to (1) in the present embodiment.
In the interpolation processing performed in 1), the six equations (b-1) to (b-6) that differ according to the magnitude relation are treated as one unified interpolation equation, and the calculation is performed according to the equation. Means equivalent. As a result, even though the input color data and therefore the magnitude relationship determined by the lower-order bit data are different for each pixel, it is possible to perform the interpolation calculation for four pixels in parallel according to the one unified interpolation formula. I have.

As is clear from the above description,
The number of pixels to be processed in parallel is determined by being restricted by the register length of the register, and does not constitute the essence of the present invention. Therefore, if the register length allows, parallel processing of four or more pixels is possible, and parallel processing with a smaller number of pixels than four pixels according to the register length is also applicable to the present invention.

The color conversion data obtained as described above is
In the case of four colors per pixel (cyan (c), magenta (m), yellow (y), black (k)), the above processing is further performed for three colors. For example, when the above-described processing corresponds to cyan and the next color is magenta, the grid point data (Am) corresponding to magenta is stored in the register 401 of FIG.
1Am2Am3Am4) is loaded, and (Cm1Bm2Em3Bm4) is stored in the register 403 in FIG.
(Gm1Dm2Fm3Fm)
4) and the register 407 in FIG.
(Hm1Hm2Hm3Hm4) Data is loaded, and lower bit data corresponding to other registers is loaded. After that, parallel operations similar to the above (8) to (11) are performed, and the magenta color interpolation operation is executed.
Next, interpolation processing for yellow and black colors is performed in the same manner. Hereinafter, parallel processing is performed for each of the next four pixels (one pixel, four colors), and color conversion is performed.

With the parallel processing of the present embodiment, it is possible to increase the speed by four times by simple calculation as compared with the conventional sequential processing.

(Second Embodiment) In a second embodiment of the present invention, these color conversion processes are not performed on four pixels in parallel as in the first embodiment described above, but one pixel at a time. The color conversion process is performed by the CPU described above.
The present invention relates to a method for effectively performing an interpolation operation process by effectively utilizing the parallel operation function of (1).

In this embodiment, the aforementioned (b-1) to (b)
The interpolation calculation processing is performed based on the equation -6), and the processing performed for each pixel in that case is as follows.

(1) Calculate an address for accessing the four grid point data.

(2) Coefficients to be multiplied by the difference value of grid point data are arranged in descending order and stored in a register.

(3) Four grid point data are stored in the read register based on the address.

(4) Interpolation processing is performed using the parallel calculation function to obtain converted color data.

The processing of the above (1) and (2) is based on Xf,
The processing is performed by conditional branching based on the magnitude relation of Yf and Zf, and the processing of (3) and (4) is processing performed without the need for conditional branching based on the magnitude relation. That is, the process in which the conditional branch appears due to the magnitude relation of the lower bit data is replaced by the process of storing the grid point data and the like in each predetermined register, as in the first embodiment.

FIG. 5 is a flowchart showing the above-mentioned processes (1) and (2).

As shown in the figure, first, in step S501,
Now, address information A for reading grid point data A
dr Generate A. That is, the input color data Xi, Y
The value of 3n bits obtained by concatenating the upper n-bit data Xh, Yh, Zh of i, Zi is Adr A. next,
In step S503, address information Adr for reading grid point data H Generate H. This processing is based on the Adr Add (2 ²ⁿ +2 ⁿ +1) to A and add Adr
H.

Next, in steps S505 to 513, processing for determining the magnitude relation of the lower-order bit data Xf, Yf, Zf is performed. That is, step S505 is performed by setting Xf ≧ Y
comparison processing for determining whether or not f holds, step S5
07 is a comparison process for determining whether or not Yf ≧ Zf is satisfied. Step S509 is a comparison process for determining whether or not Xf ≧ Zf is satisfied. Step S511 is a comparison process for determining whether Zf ≧ Yf.
Comparison processing for determining whether or not the condition is satisfied, step S51
No. 3 is a comparison process for determining whether or not Xf ≧ Zf is satisfied.

Based on each of the above determinations, in step S521, a process to be performed when steps S505 and S507 are established (positive determination), step S52
In step 2, step S505 is established, and step S50 is performed.
7 is not satisfied, and the processing is performed when step S509 is satisfied. In step S523, the processing performed when step S505 is satisfied is not satisfied and the processing is performed when determination 509 is not satisfied. In step S524, step S505 is performed.
Does not hold, and the processing is performed when step S511 is satisfied. In step S525, the processing is performed when step S505 and step S501 are not satisfied, and when step S513 is satisfied. In step S526, step S505, step S507, and step S509 are all satisfied. Perform the processing to be performed when there is not.

That is, when the magnitude relation of Xf, Yf, Zf is Xf ≧ Yf ≧ Zf, in step S521, the remaining two lattice point addresses of the tetrahedron corresponding to the magnitude relation are generated and the lower bit data is generated. Is stored in a register in accordance with the magnitude relation of. Similarly, Xf ≧ Zf>
If Yf, the process of step S522, Zf> Xf ≧ Yf
In step S523, the processing in step S524 is performed when Zf ≧ Yf> Xf, the processing in step S525 is performed when Yf> Xf ≧ Zf, and the processing in step S526 is performed when Yf>Zf> Xf. .

FIG. 6 is a diagram showing a configuration of a register according to the present embodiment. The lower-order bit data and the like obtained in the processing of steps S521 to S526 are stored in the register R0 shown in FIG. That is, the register R0 (6
4 bits) is 2m, X in step S521.
f, Yf, and Zf (16 bits each) are connected. In step S522, values of 2m, Xf, Zf, and Yf are connected. In step S523, 2m, Zf, Xf, and Xf are connected.
A value obtained by concatenating the values of Yf. In step S524, 2m,
A value obtained by concatenating the values of Zf, Yf, and Xf, step S525
Is a value obtained by connecting the values of 2m, Yf, Xf, and Zf. In step S526, the value becomes a value obtained by connecting the values of 2m, Yf, Zf, and Xf.

The address calculation described in the above (1) is performed.
The result is an algorithm for reading grid point data adjacent to grid point A.
Adr Dress next A, grid point data adjacent to grid point H
Address to read data from Adr before If H
In step S521, Adr next A = Adr A +
2²ⁿ, Adr before H = Adr next A + 2ⁿ, Stay
In step S522, Adr next A = Adr A + 2²ⁿ,
Adr before H = Adr next A + 1, step S52
In 3, Adr next A = Adr A + 1, Adr before
H = Adr next A + 2²ⁿ, In step S524, A
dr next A = Adr A + 1, Adr before H = Adr
next A + 2ⁿIn step S525, Adr next
A = Adr A + 2ⁿ, Adr before H = Adr next
A + 2²ⁿIn step S526, Adr next A =
Adr A + 2ⁿ, Adr before H = Adr next A + 1
Becomes

After any one of the processes of steps S521 to S526 is performed according to the magnitude relation of the lower-order bit data, the above-described process (3) is performed. That is, in step S527, a process of reading out the four grid point data of the tetrahedron selected in the above process and storing them in the register is performed.

In this processing, the intermediate result and the final read result
The two values of the result are stored in registers R1 and R2 shown in FIG.
Stored. That is, the address value Adr Read by A
[Adr A], the register R
1 has 0, [Adr A] [Adr next A] [Adr
before H] (16 bits each), register
[Adr A] [Adr next A], [Adr
before H], [Adr H] is stored.
I do. Here, the contents of register R1 are shifted left by 16 bits.
Aft Grid point data H read from H
Is the contents of the register R2.

Finally, in step S528, the interpolation processing described in the above (4) is performed. As shown in FIG. 6, first, the processing of R3 = R2-R1 is performed in units of 16 bits. As a result, each item in parentheses in the above-described equations (b-1) to (b-6) is stored in the register R3. Thus, regardless of six different interpolation expressions based on the magnitude relationship of the lower bit data, the interpolation operation itself can be performed by one unified process.

Next, the above-described register R3 and register R0
, And the sum of the respective contents is calculated, and the result is stored in the register R4. Thereby, the register R4
In the upper 32 bits, the product-sum operation result of the corresponding two 16-bit data is stored, and in the lower 32 bits, the product-sum operation result of the corresponding two 16-bit data is stored. Then, by adding these two sums of products and shifting them left by m bits, converted color data P can be obtained.

One input data (Xi, Y
The color conversion processing for i, Zi) is generated by conversion. For example, four colors of yellow, cyan, magenta, and black are performed dot-sequentially. In this case, the processes of (1) and (2) need only be performed once for all four colors. This is because the address for reading grid point data from the LUT does not change while input color data (for example, R, G, and B values) obtain conversion data for these four colors. On the other hand, in the processes (3) and (4) described above, since the contents (grid point data) of the LUT are different for each color obtained by the conversion, these processes are repeated for four colors, so that the processes for four colors are performed. Conversion color data can be obtained.

In this case, assuming that the above-mentioned repetitive processing is performed by loop control as shown in FIG. 7, the processing time increases because of the condition determination (step S707).
On the other hand, in the present embodiment, as shown in FIG. 8, the process of reading grid point data and storing the data in the register shown in FIG. 6 for each color is performed collectively for four colors (steps S801 to S801).
807) Then, the interpolation calculation processing may be performed for four colors at a time.

As a result, it is possible to significantly speed up the interpolation operation for four colors per pixel as compared with the conventional method.

(Third Embodiment) In a third embodiment of the present invention, an interpolation operation is performed based on the above-mentioned equations (a-1) to (a-6).

FIG. 9 shows Y in the interpolation calculation processing of the present embodiment after the magnitude relation of the lower-order bit data is determined.
FIG. 9 is a diagram illustrating processing in the case of f ≧ Zf ≧ Xf using a register configuration.

The lower bit data 2 ^m , Yf, Zf, and Xf are loaded into the register R10 in the order of magnitude. The register R11 shifts the contents of the register R10 to the left by 16 bits and stores Yf, Zf, Xf, and 0. Next, register R10-register R
11 is executed on the contents, and the result is stored in the register R.
12 is stored. Next, the grid point data A, C, G, and H shown in the equation (a-5) are stored in the register R3. Then, a product-sum operation (2 ^m −Y
f) · A + (Yf−Zf) · C and (Zf−Xf) ·
C + Xf · H is performed, and the result is stored in the register R14. Further, after an addition operation of the upper 32 bits value and the lower 32 bits value of the register R14 is performed, a right shift operation of m bits is performed to obtain converted color data.

In the above operation, the result of the subtraction operation and the result of the product-sum operation are stored in registers R12 and R14, respectively, in order to make the description easy to understand. However, the registers for storing the operation results are limited to R12 and R14. Not that
Any free register is acceptable.

When the above calculation is performed for each color of cyan, magenta, yellow, and black, the A,
The point at which the grid point data corresponding to C, G, and H is loaded and the process of storing the lower-order bit data according to the magnitude relation are the same as in the above-described second embodiment. This gives 1
With respect to the interpolation calculation for four colors of pixels, it is possible to greatly increase the speed as compared with the conventional method.

(Fourth Embodiment) In a fourth embodiment of the present invention, an interpolation operation is performed based on the aforementioned equations (c-1) to (c-6).

FIG. 10 shows that the lower bit data is Yf ≧ Zf
FIG. 9 is a diagram illustrating an interpolation operation of the present embodiment by a register configuration when ≧ Xf.

As shown in the above equations (c-1) to (c-6), the register R100 stores the grid point data A, H, C, and G, and the register R101 stores the grid point data. Data 0, G, A, and C are stored. Then, the contents of the register R0-register R1 are calculated, and the result is written to the register R102.

The resist R3 has 2 ^m , Xf, Y
f and Zf are stored. Then, the product-sum operation A + Xf (H
-G) and Yf (CA) + Zf (GC) are executed in parallel. Furthermore, the addition operation of those results and m
Bit right shift operation is performed to obtain converted color data.

In the above operation, the result of the subtraction operation and the result of the product-sum operation are stored in the resists R102 and R104, respectively, for easy understanding of the description. However, the resist for storing the operation result is limited to R102 and R104. It is not a matter of fact that any free register can be used.

For the above operations, cyan, magenta,
When the process is performed for each color of yellow and black, the storing process of the grid point data and the lower bit data for each color is performed in the same manner as in the above-described second embodiment. This gives 1
Interpolation calculation for four colors of pixels can be significantly speeded up as compared with the conventional method.

<Other Embodiments> The present invention, as described above,
The present invention may be applied to a system including a plurality of devices (for example, a host computer, an interface device, a reader, a printer, and the like) or may be applied to an apparatus including one device (for example, a copying machine and a facsimile machine).

Also, software for realizing the functions of the above-described embodiment is installed in a computer connected to the various devices or a system so as to operate the various devices so as to realize the functions of the above-described embodiment. The present invention also includes a program code supplied and executed by operating the various devices according to a stored program in a computer (CPU or MPU) of the system or apparatus.

In this case, the program code of the software itself realizes the functions of the above-described embodiment, and the program code itself and means for supplying the program code to the computer, for example, the program code The stored storage medium constitutes the present invention.

As a storage medium for storing such a program code, for example, a floppy disk, hard disk, optical disk, magneto-optical disk, CD-ROM, magnetic tape, nonvolatile memory card, ROM or the like can be used.

When the computer executes the supplied program code, not only the functions of the above-described embodiment are realized, but also the OS (Operating System) running on the computer, or another program. Needless to say, the program code is included in the embodiment of the present invention even when the functions of the above-described embodiment are realized in cooperation with application software or the like.

Further, the supplied program code is stored in a memory provided in a function expansion board of the computer or a function expansion unit connected to the computer, and then stored in the function expansion board or the function storage unit based on the instruction of the program code. It is needless to say that the present invention includes a case where a provided CPU or the like performs part or all of the actual processing, and the processing realizes the functions of the above-described embodiments.

[0097]

As described above, according to the present invention,
For example, the lower bit data indicating the position in the interpolation space of the input color data in the interpolation space, and the grid point data read from the look-up table in accordance with the magnitude relation are previously determined in accordance with the magnitude relation. Interpolation is performed by using data stored in a plurality of predetermined registers and using data stored in these registers in accordance with an interpolation operation formula predetermined between these registers. Thus, the different interpolation arithmetic expressions or the operations of each term of this expression can be unified into the arithmetic relation between each register and one prescribed expression, whereby, for example, the interpolation operation between a plurality of pixels can calculate the calculation of each item of the arithmetic expression. Can be done in parallel.

As a result, it is possible to effectively utilize the parallel processing function of the CPU having the multimedia-compatible function, and it is possible to greatly increase the processing speed as compared with the case where sequential color conversion processing is performed.

[Brief description of the drawings]

FIG. 1 is a schematic diagram illustrating a relationship between input color data and grid point data in an LUT used in a color conversion process according to an embodiment of the present invention.

FIG. 2 shows a tetrahedron (4
FIG. 3 is a schematic diagram illustrating an example of one grid point).

FIG. 3 is a block diagram illustrating a schematic configuration of a color conversion process according to an embodiment of the present invention.

FIGS. 4 (a) to (j) are diagrams showing a process for obtaining parallel color conversion data for four pixels from input color data for four pixels by performing an interpolation operation in parallel according to the first embodiment of the present invention; FIG. 3 is a diagram illustrating a configuration of a register.

FIG. 5 relates to a second embodiment of the present invention, a process of calculating an address for accessing four grid point data, and a process of arranging coefficients to be multiplied by a difference value of grid point data in descending order and storing the same in a resist; It is a flowchart which shows.

FIG. 6 is a diagram illustrating an interpolation operation process in the color conversion process according to the second embodiment using a register configuration.

FIG. 7 is a flowchart showing a comparative example with the second embodiment, and showing a process of converting input color data of one pixel into color data of four colors using a loop process.

FIG. 8 is a flowchart showing processing for converting color data of one pixel into color data of four colors without using the loop processing shown in FIG. 7 according to the second embodiment.

FIG. 9 is a diagram illustrating an interpolation operation process according to a third embodiment of the present invention by a register configuration.

FIG. 10 is a diagram illustrating a register configuration of an interpolation operation according to a fourth embodiment of the present invention.

[Explanation of symbols]

301 Register group 302Y, 302M, 302C, 302K LUT (Lookup table) 303 Size discrimination unit 304 Address generation unit 401, 402, 403, 404, 405, 406, 4
07, 408, 409, 410, R0, R1, R2, R
3, R4, R10, R11, R12, R13, R14,
R100, R101, R102, R103, R104
register

 ──────────────────────────────────────────────────続 き Continued on the front page F-term (reference)

Claims (11)

[Claims] 1. A color conversion method for performing color conversion by using an interpolation operation in combination with a look-up table, comprising: data indicating a position of the input color data in an interpolation space by a magnitude relation among input color data. In the input color data, grid point data read from the look-up table based on data indicating the address of the look-up table and data indicating the position according to the magnitude relation is determined in advance according to the magnitude relation. Storing the data in a plurality of registers, and performing an interpolation operation using data stored in the registers in accordance with a predetermined interpolation operation expression among the plurality of registers. . 2. The input color data is (n + m)
3-color data Xi bit, Yi, consists of Zi, represents the data Xi, Yi, Zi and the ^{Xi = Xh · 2 m + Xf} Yi = Yh · 2 m + Yf Zi = Zh · 2 m + Zf, the look-up When the data indicating the address of the table is Xh, Yh, Zh, and the data indicating the position by the magnitude relation is Xf, Yf, Zf, the color conversion method is the interpolation space defined by (Xh, Yh, Zh). Is a color conversion method for executing an interpolation operation using data of four grid points defined by the magnitude relation of (Xf, Yf, Zf) among the eight grid points of (Xh, Yh, Zf). Zh) storing grid point data in a register a; and registering the maximum value of (Xf, Yf, Zf) for four pixels in a register b.
And storing the grid point data adjacent to the (Xh, Yh, Zh) grid points for four pixels in the register c, and storing the median value of (Xf, Yf, Zf) for four pixels in the register d. And storing grid point data adjacent to (Xh + 1, Yh + 1, Zh + 1) grid points for four pixels in a register e. Registering the minimum value of (Xf, Yf, Zf) for four pixels in a register f
And storing the (Xh + 1, Yh + 1, Zh + 1) grid point data for four pixels in the register g. The (register c-register a) register b is calculated in parallel for four pixels, and the result is calculated. Is stored in the register h, and the register d is operated in parallel for four pixels (register e-register c) and the result is stored in the register i. Calculating the register f and storing the result in the register j; shifting the value of the register a to the left by m bits and paralleling the result by 4 pixels to the register h, the register i,
2. The color conversion method according to claim 1, wherein the step of performing the addition operation of the value of the register j performs the interpolation operation processing for four pixels in parallel. 3. The color conversion method according to claim 2, wherein data of said register a, register c, register e, and register g is subjected to color conversion processing for cyan, magenta, yellow, and black. 4. The input color data is (n + m)
3-color data Xi bit, Yi, consists of Zi, represents the data Xi, Yi, Zi and the ^{Xi = Xh · 2 m + Xf} Yi = Yh · 2 m + Yf Zi = Zh · 2 m + Zf, the look-up When the data indicating the address of the table is Xh, Yh, Zh, and the data indicating the position by the magnitude relation is Xf, Yf, Zf, the color conversion method is the interpolation space defined by (Xh, Yh, Zh). A color conversion method for performing an interpolation operation using data of four grid points defined by the magnitude relationship of (Xf, Yf, Zf) among the eight grid points of And storing the highest-order block 2 ^m of the register 0 in the second highest-order block of the register 0 (X
f, Yf, Zf), and (X
f, Yf, Zf); and (X) in the fourth block from the top of the register 0
f, Yf, Zf); storing 0 in the uppermost block of the register 1 divided into four blocks; and (X) in the second highest-order block of the register 1.
h, Yh, Zh) storing grid point data, and (X
(h, Yh, Zh) storing grid point data adjacent to the grid point; and storing (Xh
+1, Yh + 1, Zh + 1) storing grid point data adjacent to the grid point; loading the contents of the register 1 into the register 2; shifting the register 2 one block to the left; The (Xh + 1, Yh + 1, Zh)
+1) a step of storing grid point data; a step of performing the operations of the register 2 and the register 1 in parallel with four blocks; and storing the result in the register 3; (1 block of the register 0 × (of the register 3) 1 block + 2 blocks of the register 0 × register 3
2) and (3 blocks of the register 0 × 3 blocks of the register 3 + 4 blocks of the register 0 × 4 blocks of the register 3) are executed in parallel. A step of storing in the 1-2 block and the 3-4 block of the register 4; and a step of adding the 1-2 block and the 3-4 block value of the register 4 and shifting right by m bits. The color conversion method according to claim 1, wherein the color conversion is performed. 5. The color conversion method according to claim 4, wherein the color conversion processing is performed by changing grid point data of the register 1 and the register 2 for each of cyan, magenta, yellow, and black. 6. The input color data is (n + m)
3-color data Xi bit, Yi, consists of Zi, represents the data Xi, Yi, Zi and the ^{Xi = Xh · 2 m + Xf} Yi = Yh · 2 m + Yf Zi = Zh · 2 m + Zf, the look-up When the data indicating the address of the table is Xh, Yh, Zh, and the data indicating the position by the magnitude relation is Xf, Yf, Zf, the color conversion method is the interpolation space defined by (Xh, Yh, Zh). A color conversion method for performing an interpolation operation using data of four grid points defined by the magnitude relationship of (Xf, Yf, Zf) among the eight grid points of And storing the highest-order block 2 ^m of the register 0 in the second highest-order block of the register 0 (X
f, Yf, Zf), and (X
f, Yf, Zf); and (X) in the fourth block from the top of the register 0
f, Yf, Zf), storing the contents of the register 0 in the register 1 and shifting it to the left by one block.
Performing a subtraction operation of register 1) and storing the result in register 2; storing (Xh, Yh, Zh) grid point data in the uppermost block of register 3 divided into four blocks; In the second block from the top of the register 1, (X
(h, Yh, Zh) storing grid point data adjacent to the grid point; and (Xh
(1) storing grid point data adjacent to (+1, Yh + 1, Zh + 1) grid points; and (Xh) in the fourth block from the top of the register 1.
+1, Yh + 1, Zh + 1) storing grid point data; (1 block of the register 2 × 1 block of the register 3 + 2 blocks of the register 2 × the register 3
2) and (3 blocks of the register 2 × 3 blocks of the register 3 + 4 blocks of the register 2 × 4 blocks of the register 3) are executed in parallel. Storing the first, second, and third blocks of the register 4 into the first, second, and third blocks; and adding the data of the first, second, and third, and fourth blocks of the register 4 2. The color conversion method according to claim 1, wherein the interpolation operation is performed in parallel by the step of shifting right by m bits. 7. The color conversion method according to claim 6, wherein color conversion processing is performed by changing grid point data of the register 3 for each of cyan, magenta, yellow, and black. 8. The input color data is (n + m)
3-color data Xi bit, Yi, consists of Zi, represents the data Xi, Yi, Zi and the ^{Xi = Xh · 2 m + Xf} Yi = Yh · 2 m + Yf Zi = Zh · 2 m + Zf, the look-up When the data indicating the address of the table is Xh, Yh, Zh, and the data indicating the position by the magnitude relation is Xf, Yf, Zf, the color conversion method is the interpolation space defined by (Xh, Yh, Zh). A color conversion method for performing an interpolation operation using data of four grid points defined by the magnitude relationship of (Xf, Yf, Zf) among the eight grid points of Storing (Xh, Yh, Zh) grid point data in the uppermost block of the register 0;
+1, Yh + 1, Zh + 1); and (X) in the third block from the top of the register 0.
(h, Yh, Zh) storing grid point data adjacent to the grid point; and (Xh
(+1, Yh + 1, Zh + 1) storing grid point data adjacent to the grid point; storing 0 in the highest block of the register 1 divided into four blocks; (Xh
+1, Yh + 1, Zh + 1) grid point data; and (X) in the third block from the top of the register 1
h, Yh, Zh) storing grid point data; and (X,
(h, Yh, Zh) storing grid point data adjacent to the grid point; performing a subtraction operation of (register 0-register 1) in parallel for each block and storing the result in register 2; Storing 2 ^m in the uppermost block of the register 3 divided into two blocks, and (X
f, Yf, Zf); and (X) in the third block from the top of the register 3.
f, Yf, Zf), and (X) in the fourth block from the top of the register 3.
f, Yf, Zf); (1 block of the register 2 × 1 block of the register 3 + 2 blocks of the register 2 × the register 3)
2) and (3 blocks of the register 2 × 3 blocks of the register 3 + 4 blocks of the register 2 × 4 blocks of the register 3) are executed in parallel. Storing the first, second, and third blocks of the register 4 into the first, second, and third blocks; and adding the data of the first, second, and third, and fourth blocks of the register 4 2. The color conversion method according to claim 1, wherein an interpolation operation is performed in parallel by the step of right shifting m bits. 9. The color conversion method according to claim 8, wherein the color conversion processing is performed by changing the grid point data of the register 0 and the register 1 for cyan, magenta, yellow, and black. 10. A color conversion apparatus for performing color conversion by using an interpolation operation in combination with a look-up table, comprising: data indicating a position of the input color data in an interpolation space by a magnitude relationship among input color data. In the input color data, grid point data read from the look-up table based on data indicating the address of the look-up table and data indicating the position according to the magnitude relation is determined in advance according to the magnitude relation. Storage means for storing the data in a plurality of registers, and operation means for performing an interpolation operation using data stored in the registers in accordance with an interpolation operation expression predetermined between the plurality of registers. Color conversion device. 11. A storage medium storing a color conversion processing program for performing color conversion by using an interpolation operation together with a look-up table so that the color conversion processing program can be read by an information processing apparatus. Among them, data indicating the position of the input color data in the interpolation space by the magnitude relation, and data indicating the address of the look-up table among the input color data and the data indicating the position by the magnitude relation are obtained from the lookup table. The grid point data to be read is stored in a plurality of registers predetermined according to the magnitude relation, and the data stored in the registers according to a predetermined interpolation operation expression among the plurality of registers is stored. Performing an interpolation operation using the storage medium.