JPH0969961A - Color conversion device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce black dotts generated by a method which uses a three- dimensional look-up table and interpolation for fast, high-precision conversion of color signals of color printing, color hard copying, etc., and reduce a color conversion table memory. SOLUTION: A plurality of grating points data which are read out at the same time are stored in different memories in a color conversion table memory 108 and simultaneous access is gained by a memory access part 106 to shorten a processing time. The output values of grating points outside an input space which become necessary for oblique prism interpolation by an outer shell block detection part 108 and an extrapolation part 110 need not be kept directly in the memory, whose memory capacity is therefore reducible.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an application for inputting a color image signal or a color video signal and performing arbitrary color coordinate conversion and color conversion in real time, for example, a color scanner requiring high-speed color correction and color correction. , Color cameras, color hard copy devices, color display devices that require accurate color calibration, color correctors that change the colors of video images in real time, video editing devices, and color recognition devices that perform color identification. is there.

[0002]

2. Description of the Related Art Conventionally, in the fields of color printing and color hard copy, a table lookup method using a three-dimensional interpolation method has been proposed as a method for performing simple and high-speed conversion of a variety of complex color signals. These are three-dimensional interpolation methods,
Divide the color space into a plurality of unit interpolation solid groups, select a unit interpolation solid containing the input color, and use the output values at multiple vertices of the unit interpolation solid to continuously perform arbitrary color conversion over the entire color space. Is ensured and interpolated. At present, 8-point interpolation that divides the color space into a plurality of cube groups, 6-point interpolation that further divides the cube into two triangular prism groups, 5-point interpolation that divides the cube into three pyramid groups, and 5 cubes. Four-point interpolation is known which divides into one or six tetrahedral groups.

The background to which these methods have been devised one after another lies in the fact that the most common three-dimensional interpolation method, the 8-point interpolation method using a cube, has many problems. This problem is obvious from the viewpoint of cost, because the interpolation using eight points is a heavy burden on the calculation time and the hardware. But there are significant performance deficiencies. When this interpolation method is used to generate cyan C, magenta M, yellow Y, and black K in the color hard copy field, if the input is density (Dr, Dg, Db), the traditional black K generation means is used. In an operation such as a skeleton black, input 3 variables (Dr, Dg, D
The black table is created by using the MIN operation that outputs the minimum value of b) as black. However, the MIN operation belongs to the category that is the most difficult to interpolate among the non-linear transformations that are difficult to interpolate, and when interpolating using 8 points, the interpolated result sags while maintaining continuity in the interpolated section. It will be interpolated into a wavy shape. Since this shape resembles a "ripple wave", it will be referred to as "ripple" hereinafter. That is, in the 8-point interpolation method,
Ripple is generated in the gradation of the black plate, and a false contour that is visually unbearable is formed on a color image of four CMYK colors.

One of the major advantages of 4-point interpolation using tetrahedral division and 5-point interpolation using pyramid division is that this MIN operation can be interpolated without ripples. For example, the description of 4-point interpolation can be found in JP-A-2-286867, and that of 5-point interpolation is JP-A-56.
Detailed description can be found in Japanese Patent No. 14237.

In these contents, it is argued that both the tetrahedral division and the pyramid division have division boundaries in the diagonal axis direction of the unit cube, and therefore black interpolation in the achromatic (gray) direction can be favorably performed. There is. This only mentions the interpolation error in the achromatic direction of the MIN calculation interpolation, and is rather incomplete. To be precise, in these figures, the division boundary surface includes all the differential discontinuous surfaces of the surface in which the result of the MIN operation is constant in the color space, and therefore the MIN operation is performed in all directions, not only in the achromatic direction. The color conversion can be linearly interpolated without ripples. However, in any case, in the tetrahedron division and the pyramid division, the ripple occurrence in the MIN calculation is completely avoided not only in the achromatic direction but also in the omnidirectional interpolation, which is a technically solved problem.

Next, another proposed interpolation method, which is a 6-point interpolation method by triangular prism division, will be described in detail.
The first conventional example is a method of dividing an XYZ space into a plurality of triangular prisms and interpolating them (JP-A-5-75848). A second conventional example is a method of interpolating the YCrCb lightness color difference space by dividing it into a triangular prism having a principal axis in the Y direction and the other two axes set in the color difference plane (Japanese Patent Laid-Open No. 5-46750 and Japanese Patent Laid-Open No. 5-75050). No. 120416).

Although it is obvious that the hardware is simplified in these proposals as compared with the 8-point interpolation,
No mention is made of any statement regarding the avoidance of ripples in the MIN operation. However, in the second conventional example, due to the feature that the triangular prism main axis is set in the Y direction, various color conversions do not become unnatural bending line shapes in the interpolation in the lightness direction, that is, the direction parallel to the achromatic color axis. The advantages of being able to linearly interpolate are described.

On the other hand, in the first conventional example, a triangular prism is used, but since the characteristics in the main axis direction of the triangular prism as in the second embodiment are not utilized, there is no advantage in the interpolation performance and the interpolation is performed. There is no concrete description of characteristics.

[0009]

A first conventional example of a triangular prism interpolation method proposed in the past (Japanese Patent Laid-Open No. 7584/1993).
No. 8) has the following problems.

First, in the conventional example, a method of interpolating an XYZ space and an RGB space by dividing a three-primary color space in which the lightness color difference is not separated into trigonal prism groups whose main axes match one axial direction of the three-primary color space. Is taking. This causes a large problem that "ripple" is generated on a large scale in the case of the MIN calculation as described above. According to an experiment by the present inventor, as will be described later in detail in the graph, the ripple generated in the case of the triangular prism interpolation of the first conventional example is so large that it cannot be ignored in the achromatic direction. In the same publication of the first conventional example, "when applied to color correction here, XYZ corresponds to input R (red), G (green), B (blue) signals, and output P
Is equivalent to Y (yellow), M (magenta), C (cyan), and Bk (black) signals that control ink in the case of a four-color printer. " Despite being seen, there is no disclosure of how to avoid this ripple occurrence.

Secondly, in the effect of the same publication of the first conventional example, "Since the memory capacity is smaller than that of the conventional interpolation method, the whole hardware is reduced and the LSI can be easily implemented." There is a description, but there is no description about how much memory capacity reduction is expected for any reason. Rather, there is a drawback that the memory usage efficiency is about 1/8 because the redundant grid points to be accessed in parallel are generated during the transfer from the ROM to the RAM and the memory is not effectively used.

On the other hand, a second conventional example (Japanese Patent Laid-Open No. 5-4675)
No. 0 and Japanese Patent Laid-Open No. 5-120416), the triangular prism main axis is set in the lightness Y direction as described above, and in the achromatic direction, not only MIN calculation but ripple generation is avoided. Has already been accepted. However, there is a possibility that ripples may occur in directions other than the achromatic color direction in the color space, and there is a problem that has not been completely solved.

The present invention is often used in color conversion.
It is possible to completely avoid the occurrence of "ripple" in the calculation not only in the achromatic color direction but also in all directions of the input color space. At the same time, by using an oblique triangular prism in the interpolation space, it is necessary to access the output values of the grid points outside the input color space. It is an object of the present invention to provide a color conversion device capable of avoiding an increase in the color conversion table memory by extrapolating from the grid point output value in the input color space.

[0014]

To achieve the above object, the present invention provides a pixel input section for dividing a color image signal represented by various color signals into an upper bit section and a lower bit section, and the lower bit section. Comparison section for comparing the sections and outputting the magnitude relation, an interpolation origin setting section for determining the position of the interpolation origin from the upper bit section according to the output of the comparison section, and output on a grid point in the input color image signal space A color conversion table memory that stores values, and a memory access that generates an address of the color conversion table memory to be accessed from the interpolation origin for the color conversion table memory and supplies the address to the color conversion table memory Block and outer block detection for detecting whether or not the grid point selected by the memory access unit exists in the outer block outside the input color image signal space A difference value generation unit that generates a difference value of the grid point output values, an extrapolation unit that extrapolates output values of grid points outside the input color image signal space from the grid point output values, and the outer block detection The difference value selector that switches the output of the difference value generation unit according to the output of the unit, the origin selector that switches the output value of the extrapolation unit and the output value of the lattice point corresponding to the interpolation origin, and the lower bit unit forms an oblique triangular prism. A weight generation unit that generates an interpolation weighting coefficient along the line, an oblique triangular column determination unit that selects an oblique triangular column based on the magnitude relationship of the weighting factor, and an output value using the output of the difference value selector and the weighting factor. It is characterized by comprising an oblique triangular prism interpolation calculation unit.

[0015]

In the color conversion apparatus of the present invention, the three-dimensional space created by the input color signals of the three primary colors is divided into a plurality of unit cube areas,
Assuming an oblique triangular prism or a parallelepiped composed of apexes forming upper and lower bases of unit cubes that are adjacent to each other along the diagonal direction of the space, the entire input color space is defined as the oblique solid. The output values corresponding to arbitrary input colors are interpolated by using the output values at the respective vertices of the oblique solid body. As a result, the oblique triangular prism or parallelepiped group in the three primary color input space is parallel to the three planes of the first color axis = second color axis, second color axis = third color axis, and third color axis = first color axis. It has a surface as a boundary surface. For example, assuming that the input color space of the three primary colors is the RGB color space, R = G, G = B, B
= 3 including R plane. These three planes are three surfaces that are differentially discontinuous in the MIN calculation, while the MIN calculation result can be interpolated in a completely linear manner in the three regions surrounded by the boundaries. Therefore, since the non-interpolable surface can be assigned only to the boundary of the interpolation unit solid, the interpolation operation is performed linearly in the entire input color space with no ripple. That is, it is possible to completely avoid the occurrence of "ripple" in the MIN calculation that is frequently used in color conversion, not only in the achromatic color direction but also in any direction in the input color space, and access is required by using an oblique triangular column in the interpolation space. By extrapolating the output values of grid points outside the input color space from the output values of grid points inside the input color space, it is possible to avoid an increase in the color conversion table memory.

[0016]

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

FIG. 1 is a block diagram of a color conversion device according to an embodiment of the present invention. In FIG. 1, 1
01 designates the image input signal RGB as the upper bits (RH, GH, B
In the present embodiment, the pixel input unit separates H) and lower bits (RL, GL, BL) into upper 3 bits and lower 5 bits, respectively. 102 is a comparison unit for comparing the magnitudes of RL and GL and BL and GL, and 103 is the upper bit R
It is an interpolation origin setting unit that determines the position of the interpolation origin by subtracting the comparison results C1 and C2 of the comparison unit 102 from H and BH. 10
Reference numeral 4 is a weight generation unit that outputs the subtraction RL-GL, BL-GL and GL of the lower bits, 105 is an oblique triangular prism determination unit that determines which of two types of oblique triangular prisms is used in oblique triangular prism interpolation, and 106 is High-order bit GH, high-order bits RH 'and BH' adjusted by the interpolation origin setting unit 103, and an oblique triangular prism determination unit 1
The memory access unit generates a memory address to access the color conversion table memory 108 based on the determination result of 05. Reference numeral 107 denotes an outer block detection unit that determines whether the grid points read from the color conversion table memory 108 exist inside or outside the input color image space, and the detection result EDB is stored in the color conversion table memory 108 and the origin. Output to the difference value selector 111. 109 is a grid point output value a, which is an output of the color conversion table memory 108,
A difference value generation unit that calculates a difference value among b (or d), c, e, f (or h), and g, and an extrapolation unit 110 that extrapolates an output value of a grid point outside the input color image space. , 11
Reference numeral 1 denotes an origin / difference value selector that switches between the grid point output value extrapolated according to the detection result EDB of the outer block detection unit 107 and the grid point output value in the input color space, and further selects a difference value necessary for extrapolation. Reference numeral 112 is an oblique triangular prism interpolation calculation unit, and 113 is an interpolation calculation result.

The operation of the color conversion device configured as described above will be described. The color conversion device of FIG. 1 is a device for performing color conversion of input pixel signals (R, G, B) by the oblique triangular prism interpolation method. The oblique-triangular-column interpolation calculation unit 112 is a block that executes the interpolation calculation, and the other blocks (101 to 111) all have a role of generating data necessary for the oblique-triangular prism interpolation calculation. Therefore, in order to describe the operation of the present embodiment, the oblique triangular prism interpolation method will be described first.

The oblique triangular prism interpolation method divides a three-dimensional space formed by three primary color input color signals into a plurality of unit cube areas, and the upper and lower bases of the unit cubes adjacent to each other along the diagonal axis direction of the space. Assuming an oblique triangular prism composed of vertices forming the bottom, the entire input color space is set to be covered by the oblique triangular prism, and an output value corresponding to any input color is
This is a method of interpolation using the output value at each vertex of the oblique triangular prism. The six vertices that form the oblique triangular column are the pixel input unit 1
It is selected by the upper bits (RH, GH, BH) generated in 01. Further, the lower bits (RL, GL, BL) generated by the image input unit 101 are input pixel signals (R, G,
B) represents the position in the oblique triangular prism, and the color conversion output value for the input pixel signal (R, G, B) is calculated from the color conversion output values of the six vertices of the oblique prism by interpolation calculation according to the input position. To do. 2 (A) and 2 (B) show an oblique triangular column selected by the upper bits (RH, GH, BH), and FIGS. 2 (C) and 2 (D) show the upper bits (RH, G).
H, BH) represents a rectangular prism selected by. a,
b, c, d, e, f, g, and h are the vertices of an oblique triangular prism or an oblique quadrangular prism, and each has a color conversion output value for an input pixel signal.

FIG. 3 shows an oblique coordinate system XYZ and an orthogonal coordinate system RG.
The relationship of B is shown, and the XYZ space and the RGB space have the following vector relationships.

[0021]

[Equation 1]

Input point (r, b, g) in RGB space =
The vector O of (RL, GL, BL) is

[0023]

[Equation 2]

## EQU3 ## This is expressed as XYZ using the equation (1).
When expressed in space

[0025]

(Equation 3)

## EQU1 ## Therefore, the input point (x, y, z) expressed in XYZ space is

[0027]

(Equation 4)

## EQU1 ## Therefore, (R
It suffices to perform interpolation calculation from the interpolation origin a in FIG. 2 as the movement amounts of L-GL), (BL-GL), and GL.

Here, when the value of x or y becomes negative, it is inconvenient, but it can be solved by the following method. In this case, the amount of movement from the interpolation origin is in the -R or -B direction. Therefore, this can be solved by retracting the origin position a by one in the -R or -B direction and converting the interpolation weight into a distance from the new interpolation origin position. The amount of movement thus positively expressed is represented as a weighting coefficient for oblique-triangular prism interpolation as follows.

[0030]

(Equation 5)

FIG. 4 is a diagram in which, when an input point determined by the lower bits (RL, GL, BL) is input to the unit solid 401, what kind of oblique triangular prism it belongs to is classified into four. When an input point is entered in AREA (0) shown in FIG. 4A, the input point belongs to the oblique triangular prism 402. FIG.
When an input point is entered in AREA (1) shown in (B), the input point belongs to the oblique triangular prism 403 which is obtained by moving the interpolation origin a of the oblique triangular prism 402 in the −B direction. AR shown in FIG. 4 (C)
When an input point is entered in EA (2), the input point is an oblique triangular prism 4 obtained by moving the interpolation origin a of the oblique triangular prism 402 in the −R direction.
Belong to 04. When an input point is entered in AREA (3) shown in FIG. 4D, the input point belongs to the oblique triangular column 405 which is obtained by moving the interpolation origin a of the oblique triangular column 402 in the −R and −B directions. Table 1 shows the moving direction (R, B) of the position a in relation to the magnitude of RL, BL, GL.

[0032]

[Table 1]

Therefore, the upper bits of RGB (RH, GH,
The origin position a of the oblique triangular column selected in (BH) is changed by shifting the position of a in the -X and -Y directions, respectively, according to the magnitude judgment of expression (4) of the lower bits of RGB, and includes the input point. The origin position of the oblique triangular column can be obtained.

When x is negative, a value obtained by adding 32 to x is used as the weighting coefficient. The same applies to the y direction. It is very easy to add 32 when x or y becomes negative because the sign bit of the output of the digital subtractor can be ignored. As described above, the upper bits (R
H, GH, BH) and the method of changing the lower bits (RL, GL, BL) are summarized in (Table 2).

[0035]

[Table 2]

In this way, the end point positions b to h of the two oblique triangular prisms corresponding to the origin position a of the oblique triangular prisms, and (5)
With the weighting factor calculated from the formula, it is ready to execute the linear interpolation operation in the oblique triangular prism. Next, the interpolation calculation method in the oblique triangular prism will be described with reference to FIG.

The color conversion table output values for a to h are (a)
~ (H). If the intersections of the straight line drawn from the interpolation output point o in parallel to the line segment ae and the bottom surface abcd and the top surface efgh are m and n, respectively, the outputs (m) and (n) at the points m and n are shown in FIG. When Type = 0 shown in

[0038]

(Equation 6)

It becomes Therefore, the output (o) at the target point o after linear interpolation between the m point and the n point is

[0040]

(Equation 7)

Is calculated as Similarly, when Type = 1,

[0042]

(Equation 8)

Is calculated as The above is the description of the oblique triangular prism interpolation method.

Returning to the explanation of the operation in FIG. The upper bits (RH, GH, BH) and the lower bits (RL, GL, BL) of the image input signal RGB generated by the pixel input unit 101 are (RH ', G) according to the rules shown in (Table 2).
H ', BH') and (RL ', GL', BL '). To perform the operation of (Table 2), first, the lower bits (RL, GL, B
It is necessary to judge the size relationship of L). This is performed by the comparison unit 102. FIG. 6 shows the operation block of the comparison unit 102, and performs the operation corresponding to (Table 1). That is, C1 outputs 1 when GL is larger than RL and outputs 0 otherwise, C2 outputs 1 when GL is larger than BL, and 0 otherwise.
Is output. C1 and C2 are given to the interpolation origin setting unit 103, and the interpolation origin setting unit 103 creates (RH ', GH', BH ') by the method shown in (Table 2). Where GH 'is 3
Bits, RH 'and BH' are expanded to 5 bits with a sign by subtraction from the output of the comparison unit.
Further, C1 and C2 are also given to the weight generation unit 104, and the weight generation unit 104 uses the method shown in (Table 2) (RL ', GL',
BL ') is created. FIG. 7 shows an operation block of the weight generation unit 104, which calculates (RL
-GL), (BL-GL) are calculated, and GL is output as it is. The input and output have the same bit width because if the subtraction result is negative, the sign bit is ignored and it is equivalent to borrowing from the upper bits. The result of borrowing from the upper bits (RL-GL) ', (BL-G
L) 'is output. From the above, the position (RH ', GH', BH ') of the interpolation origin a and the weighting factors (RL-GL)', (BL-GL) ', G used in the oblique triangular prism interpolation shown in FIG.
L'is decided.

Next, the position of the interpolation origin a (RH ', GH', B
H ′), the color conversion output values of the six vertices of the oblique triangular prism are read from the color conversion table memory 108. The memory access unit 106 generates addresses for accessing the six grid points required for the interpolation operation from the representative grid point group defined by the upper bits (RH, GH, BH) of the image input signal RGB. The address is given to the color conversion table memory 108, and the color conversion output values stored in the six grid points forming the oblique triangular prism are stored in the color conversion table memory 1.
It is read from 08. a, b (or d), c, e, f
(Or h) and g are color conversion table memory 1 respectively.
It is the color conversion output value of each grid point read from 08.
However, when determining the six grid points, (A) of FIG.
Two oblique triangular prisms (Type = 0 and T as shown in FIG.
It must be decided which of the ype = 1) is to be used. This judgment is performed by the oblique triangular prism judgment unit 105.
It is. FIG. 8 is an operation block of the oblique triangular prism determination unit 105. Among the outputs of the weight generation unit 104, (RL−GL) ′,
Due to the size relationship of (BL-GL) ', the oblique triangular prism Type =
0 or Type = 1 is determined, and the determination output (PRISM) is obtained. The color conversion output values a and b of the six grid points forming the oblique triangular prism read from the color conversion table memory 108.
(Or d), c, e, f (or h) and g are given to the oblique triangular prism interpolation calculation unit 112 via the difference value generation unit 109, the extrapolation unit 110 and the origin / difference value selector 111, and the expression ( 6) to (11) are used for the oblique triangular prism interpolation calculation. Further, the weighting coefficient (R
L-GL) ', (BL-GL)', and GL 'are also interpolation calculation units 11
2 and is also used for the diagonal triangular prism interpolation calculation of the equations (6) to (11).

Then, the output of the interpolation origin setting unit 103 (R
H ', GH', BH ') to color conversion table memory 108
The operation of generating the outputs a, b (or d), c, e, f (or h), and g will be described. The output (RH ', GH', BH ') of the interpolation origin setting unit 103 is input to the memory access unit 106 and is input to the color conversion table memory 108.
Generate an address to access. The color conversion table memory 108 stores the color conversion outputs a, b (or d), c, e, f (or h), g according to the address.
Is output. First, the relationship between the address of the color conversion table memory and the grid point output value will be described.

In this embodiment, the output values of 729 (= (2 ³ +1) ³ ) grid points obtained by dividing each RGB axis into eight equal parts are stored in the color conversion table memory 108. Six grid points are used for one interpolation calculation. Therefore, in order to increase the calculation speed of the color conversion device, it is necessary to simultaneously read out these six grid point data from the memory. In order to realize this, 6 grid point output values read simultaneously are set to 6
The color conversion table memories (M0 to M5) may be stored separately. From the above, the color conversion table memory 1
It is necessary to provide six memories in 08 and allocate lattice points that are simultaneously accessed to different memories. Therefore, with reference to FIG. 9, it will be described in which of the color conversion table memories M0 to M5 each of the 729 grid points is stored.

(A) shows the RGB space of the color conversion table memory composed of 729 lattice points. Two (RH, BH, GH) and (RH ', BH', GH ') are written together on the space axis. Of the 6 lattice points forming the oblique triangular prism, 3 lattice points on the upper surface and 3 lattice points on the lower surface will be considered separately. That is, GH = 0, 2, 4, 6, 8 (even surface) is assigned to the color conversion table memories M0 to M2, and GH = 1, 3, 5, 7
The surface (odd surface) is assigned to the color conversion table memories M3 to M5.

In (B), GH = 0, 2, where GH becomes an even number.
The method of allocating to the memory on 4, 6 and 8 planes is shown. First, the origin position (on the G-axis) 901 is stored in M0, and M1, M2 ,. . . And change the memory to store. And when it is full in the R-axis direction, R
Go back to H = 0 and proceed to the B axis by one. Start B axis is M1
, And again in the R-axis direction, M2, M0 ,. . . And move on. When moving to B axis further, it starts from M2. According to this rule, the 9 × 3 area 902 is read out at the same time, and the three lattice points (the bottom surface on one side of the oblique triangular prism) can be drawn from different memories M0 to M2.
The remaining area of (B) can be filled up by repeating the area 902. In (C), GH is an odd number GH =
The method of allocating the memory to the first, third, fifth, and seventh planes is shown. Allocation method is exactly the same as (B), M0 to M3,
It simply replaces M1 with M4 and M2 with M5. As shown in (B) and (C) above, if the output values of each grid point are stored in the memories M0 to M5, the 6 grid point data that are simultaneously read in one interpolation calculation can be retrieved from different memories. You can

The memory access unit 106 creates a memory address in order to read the color conversion output value arranged on the color conversion table memory in this way. The memory access unit 106 outputs the output of the interpolation origin setting unit (RH ', BH',
An address for accessing the color conversion table memory is generated from GH ') and output to the color conversion table memory 108.
FIG. 10 shows an internal configuration diagram of the memory access unit 106, the outer block detection unit 107, and the color conversion table memory 108. The memory access unit 106 includes a slide unit 1001, an address generation group 1002 that supplies an address to each color conversion table memory, and a memory interface group 1003 that connects the color conversion table memory to an external bus. The color conversion table memory 108 includes a color conversion table memory group 504 including six memories (M0 to M5) and a memory data selector group 1005 for switching an output path from the color conversion table memory.

First, the description of the slide portion 1001 will be started. FIG. 11 is a view of the color conversion table memory arranged in the RGB space of FIG. 9A viewed from the (−B) direction,
FIG. 12 is a view seen from the G direction. In FIG. 11, A indicates an oblique triangular column used when GH ′ is an even number, and B indicates an odd number. In FIG. 12, the position A of the lower surface of the interpolated solid is indicated by (RH'BH'GH ') indicating the interpolation origin a, and the upper surface is the position A'advanced by 1 in the R and B directions, respectively. . In order to form the oblique triangular column, one of the M0 to M2 planes or the M3 to M5 planes must be moved by one in the R and B directions. The slide portion 501 plays this role. As can be seen from FIG.
If H'is even, slide M3 to M5, and if GH 'is odd, slide M0 to M2. Therefore, the slide unit 501 determines whether GH 'is even or odd by using the least significant bit (LSB) of GH' and the least significant bit of GH 'is 0.
Since GH ′ is an even number, the interpolation origin signal RH ″ 2 for accessing the color conversion table memories M3 to M5
And BH ″ 2 are given by RH ′ + 1 and BH ′ + 1 respectively. Further, when the least significant bit of GH ′ is 1, GH ′ is an odd number, so the interpolation origin for accessing the color conversion table memories M0 to M2. Signals RH "1 and BH" 1 are R respectively
It is given by H '+ 1 and BH' + 1. GH "1 and GH" 2 receive GH 'as it is. The logic of the slide unit 1001 is summarized in (Table 3) above.

[0052]

[Table 3]

By the above processing, the slide unit 10
After 01, the path to the memories M0 to M2 and the memory M3
~ The route for M5 is divided into two systems. Therefore, the address generation group 100 following the slide unit 1001
2, memory interface group 1003, color conversion table memory group 1004, memory data selector group 1005
Are both composed of two systems corresponding to M0 to M2 and M3 to M5.

Next, how to apply the color conversion table address to each grid point, the relationship between each grid point and the color conversion table address will be described with reference to FIGS. 13D and 13E. FIG. 13D shows M0 arranged in FIG. 9B.
It is a figure explaining how to give an address to ~ M2. The rule is to make three blocks in the R-axis direction starting from the B-axis, and to number these blocks in the R-axis direction from the origin. This M0, M
Hereinafter, three blocks each including 1 and M2 will be referred to as a unit block. When the allocation of the address along the R axis is completed, the address is moved in the B axis direction by one. In this embodiment, each of the RGB axes is divided into eight equal parts, so that there are 27 unit blocks (= 9 × 9 ÷ 3) on the entire RB surface. Therefore, No. 26 is added to the unit block farthest from the origin. Here, regarding the bottom surface (GH = 0) of M0 to M2, one number is moved in the G axis direction because the numbering is completed, but the M3 to M5 surfaces (GH = 1) immediately above M0 to M2 are Skip and move to the upper M0-M2 surface (GH = 2). The number will continue to be number 27. When numbers are added to the unit blocks in this way, as shown in FIG. 13F, the last unit block 903 located diagonally to the origin is No. 134 (= 26 + 2).
7 × 4). The unit block number added in the above manner is used as it is as the address of M0 to M2.

The method of creating the addresses M3 to M5 is exactly the same. The numbering starts at GH in Fig. 9 (A).
It becomes the RB plane of = 1 and this corresponds to (E) of FIG.
When the unit block numbers are numbered in the same manner as M0 to M2, the unit block 904 farthest from the origin on the M3 to M5 planes is numbered 107 (=) as shown in FIG.
26 + 27 × 3). The reason that the numbers are different from those of the M0 to M2 surfaces is that the M0 to M2 surfaces are five surfaces and the M3 to M5 surfaces are four surfaces, as can be seen from FIG.

From the above, since the type of memory in which the output value of each grid point is stored and the address in each memory are determined, the output (RH "1, BH" 1, G of the slide unit 1001 is determined.
A method of generating an address from H "1) and (RH" 2, BH "2, GH" 2) will be described.

FIG. 14 shows an address generation group 1002 (see FIG.
(A) is an internal configuration diagram of address generation units 0 to 2 corresponding to the memories M0 to M2, and (B) is an internal configuration of address generation units 3 to 5 corresponding to the memories M3 to M5. Indicates. In order to explain the operation of FIG. 14, it is necessary to first explain the method of executing the oblique triangular prism interpolation calculation. Although the principle of the oblique-triangular prism interpolation method has been described with reference to FIG. 5, FIG. 15 shows a method of allocating the color conversion table memory and a method of executing the oblique-trigonal prism interpolation calculation in consideration of minimizing the hardware scale of the interpolation calculation circuit. In FIG. 15, four adjacent triangular prisms are drawn. Of these, the lower two oblique triangular columns 1401 and 1402 are the same as those shown in FIG. 5, the lower surface is abc or adc, and the upper surface is efg or ehg. On the other hand, the two oblique triangular columns 140 above
In 3 and 1404, the upper surface is abc or adc, and the lower surface is efg or ehg. Therefore, if the vertical interpolation, that is, the interpolation between the point m and the point n is performed in the upward direction (abc surface → efg surface or adc surface → ehg surface) as in the case of the lower oblique triangular prism 1401 or 1402, the upper oblique triangular prism is generated. Conversely, in 1403 or 1404, interpolation such as efg surface → abc surface or ehg surface → adc surface must be performed. The interpolation calculation circuit does not distinguish this direction. For example, the ab (d) c surface is fixed to the bottom surface and the ef (h) g surface is fixed to the upper surface. I don't want to increase the burden. Therefore, as shown in FIG. 15, the ab (d) c surface is the lower surface of the interpolation calculation, and ef (h) g
The surface is fixed to the upper surface of the interpolation calculation and the interpolation calculation is performed. The ef (h) g plane is a like the upper oblique triangular columns 1403 and 1404.
Even if it is located below the b (d) c plane, the interpolation calculation is ab (d) c.
Surface → ef (h) Execute in the g-plane direction. As a result, the memories M0 to M2 always give the grid point output values of the lower surface ab (d) c surface of the interpolation calculation, and the memories M3 to M5 always give the output values of the grid point of the upper surface ef (h) g surface of the interpolation calculation. Will be given.

From the above, by using the interpolation method shown in FIG. 15, the ab (d) c plane → ef (h) g plane and ef
(H) It is possible to reduce the size of the hardware by avoiding the circuit configuration that copes with the two systems of g-plane → ab (d) c-plane. The weighting factor GL ′ in the G-axis direction needs to be switched depending on the direction of the interpolation calculation.
When performing interpolation between the points mn in the downward direction as in 1404,

[0059]

[Equation 9]

The following weighting coefficient is used in the interpolation calculation. In FIG. 7, GL 'is given as it is to GL'.
As shown in 6, when the least significant bit of GH is 0, GL is given to GL 'as it is, and when the least significant bit of GH is 1, GL' is given by equation (12). When the oblique triangular column B in FIG. 11 is selected, the ab (d) c plane is ef (h)
The interpolation in the vertical direction (between the points mn) is inverted to come on the g-plane, and the weighting coefficient GL is given by the equation (12).

Returning to the explanation of FIG. In the address generators 0 to 5 of FIG. 14, as described in FIGS. 13D to 13F, six grid points to be selected according to the input pixel signal are stored in any of the six memories M0 to M5. Output from the slide unit 1001 (R
H "1, BH" 1, GH "1) and (RH" 2, BH "2, GH" 2)
Must generate an address from. The creation of this address means the calculation of the unit block number defined in FIG. 13 as described above. Therefore, the address generators 0 to 5 are (RH "1, BH" 1, GH "1) and (RH" 2, BH "2, G).
How to calculate the unit block number to which each grid point belongs from H ″ 2) will be described.

First, FIG. 14A will be described. FIG.
(A) is an internal configuration diagram of the address generation units 0 to 2. Since the output (RH "1, BH" 1, GH "1) of the slide unit 1001 indicates the position of the interpolation origin a,

[0063]

(Equation 10)

The unit block number to which the grid point corresponding to the interpolation origin a belongs can be obtained by the above. However, R
H ″ 1/3 represents the quotient of division of 3. Now, assuming that the triangle 905 in FIG. 9B is selected, the output value of the grid point corresponding to the interpolation origin a is stored in the memory M0. , The address of the equation (13) must be supplied to the memory M0.However, if the triangle 906 of FIG. 9B is selected, the output value of the grid point corresponding to the interpolation origin a is stored in the memory M2. Therefore, the memory M2 must be supplied to the memory M2. Thus, the address of the interpolation origin a calculated by the equation (13) must switch the memory to be supplied according to the selected triangle type. RH "1 divided by 3 and BH" 1
It is summarized as shown in (Table 4) using the remainder obtained by dividing by.

[0065]

[Table 4]

RH "mod3 is the remainder of RH" 1 divided by 3,
BH ″ mod3 indicates the remainder when dividing BH ″ 1 by 3.
As shown in the area 902 in FIG. 9B, the memory M
In the arrangement pattern of 0 to M2, a 3 × 3 region is repeated in the R-axis direction. Therefore, the remainder obtained by dividing RH ″ 1 by 3 and B
It can be defined as shown in (Table 4) by looking at the remainder obtained by dividing H ″ 1 by 3.

Next, a method of calculating addresses for the remaining grid points b, c and d will be described. When the triangle 1907 is selected in (D) of FIG. 13, the output value of the grid point b is read from the memory M1, and the address supplied to the memory M1 at this time is the same as the address given to M0. This is because M1 exists in the same unit block as M0. Therefore, the address of Expression (13) may be supplied as it is. On the other hand, although the output value of the grid point c is stored in M2, the address needs to be added to the address given by equation (13) by three. This is because M2 is located one above in the B-axis direction of the unit block to which M0 belongs, and the unit block number is increased by 3. In this way, memories M0-M
It can be seen that the addresses given to 2 need only be incremented with respect to the grid point deviating from the unit block to which the interpolation origin a belongs, based on the address calculated by the equation (13). Therefore, which memory is to be incremented according to the input RGB signal, and the increment amount is determined severally to generate the address. Interpolation origin a
The rules for incrementing the address are summarized as shown in (Table 5) using RH ″ 1 divided by 3, the remainder RH ″ mod3 and BH ″ 1 divided by 3, the remainder BH ″ mod3, and PRISM.

[0068]

[Table 5]

Since the memory corresponding to the interpolation origin a is not incremented, the increment value of the memory to which the address of the interpolation origin a shown in (Table 4) is supplied is 0.

From the above, the address generators 0 to 2 (Table 5)
The address increment should be executed for each memory according to the logic of. First, it is necessary to generate a quotient obtained by dividing RH ″ 1 by 3 according to the equation (13). This is executed by DIV3 (1301) in FIG. 14A. RH for performing case classification in (Table 5) "1" and "BH" 1 need to be divided by 3. The MOD3RH (130
2) and MOD3BH (1303). (Table 5)
As shown in FIG. 14, addition of the increment values shown in FIG.
Disassemble into axis system and execute. That is, the color conversion table address MAi (i = 0 to 5) output from the address generator is

[0071]

[Equation 11]

Is calculated in the form

[0073]

(Equation 12)

[0074]

(Equation 13)

And execute. INCR is an increment value in the R-axis direction, INCB is an increment value in the B-axis direction, and these are given by (Table 6) based on (Table 5).

[0076]

[Table 6]

As shown in FIG. 14 (A), the table of increment values shown in (Table 6) is provided for each memory, INCB is added to BH "1 to obtain BN, and INCR is added to RH" 1. To obtain RN. The table of the increment values of (Table 6) has the increment value generation unit 0 (1304), the increment value generation unit 1 (1305), and the increment value generation unit 2 (1306), and the MOD3R
RH ”mod3 is given from H (1302), and MOD3BH
BH ″ mod3 is given from (1303), and PRISM is input from the oblique-triangular-pillar determination unit 105 in FIG. 1. Increment value generation unit, addition of increment values of formulas (15) and (16), and color conversion of formula (14) The part that executes the calculation of the table address MAi is called a linearizer, and since (Table 6) of the increment value generation unit differs for each memory, this linearizer has the memories M0 to M2.
Provide 3 for each. The linearizer that outputs the address MA0 to the memory M0 is the linearizer 0 (1407), the linearizer that outputs the address MA1 to the memory M1 is the linearizer 1 (1408), and the linearizer that outputs the address MA2 to the memory M3 is the address MA2.
Linearizer 2 (1409) that outputs a linearizer that outputs
Call. RH "mod3, BH" mod3 and PRISM are shared for each linearizer, and the same value is used for linearizers 0-2.
Supply to Further, GN is M0 to M as shown in FIG.
2, M0 is a scale in the G-axis direction set for each of M3 to M5.
The GNs of ~ M2 are installed on the surface of even GH, and the GNs of M3 to M5 are installed on the surface of odd GH. The GN is obtained by shifting GH to the right by one bit.
If input is made to the oblique triangular column A in FIG. 11, GH = 1, and therefore GN = 1, and the lower surface of the oblique triangular column A is GN (M
From 0 to M2) = 1, the upper surface is GN (M3 to M5) =
Each is given from the 1st side. On the other hand, if there is an input to the slanted triangular prism B, GH = 1, and thus GN = 1.
Therefore, the lower surface of the oblique triangular column A is given from the surface of GN (M3 to M5) = 1, but the M0 to M2 surfaces are inappropriate. In this case, GN (M0 to M2) must be 2.
That is, if GH is odd, then GN (M0-M2) must be incremented by one. Therefore, in the address generators 0 to 2, after shifting GH ″ 1 by 1 bit to the right,
The least significant bit (LSB) of H ″ 1 is added to adjust GN.

FIG. 14B is a diagram showing the internal structure of the address generators 3 to 5 corresponding to the color conversion table memories M3 to M5 of the address generator group 1002 of FIG. This is almost the same configuration as (A) for the color conversion table memories M0 to M2. M0 is set to M3, M1 is set to M4, and M is set to M4.
2 by replacing each with M5.
As can be seen by comparing (B) and (C) of FIG. 9, the GH even surface and the GH odd surface have a symmetrical relationship, and M0 is set to M3, and M0 is set to M3.
By replacing 1 with M4 and M2 with M5, the operation of the GH even plane can be completely reproduced on the GH odd plane. However, there is only one difference. It is a process for GH "2. GH" 2 is the R to which the memories M3 to M5 belong.
Although the B plane is determined, unlike the RB plane to which the memories M0 to M2 belong, it is not necessary to move the RB plane in the G axis direction by the least significant bit of GH "1. Therefore, GH" 2 is GN shown in FIG. It only shifts one bit to the right to produce (M0-M2) and GN (M3-M5). From the above, the address MAi given to the color conversion table memories M0 to M5 can be generated from the outputs (RH "1, BH" 1, GH "1) and (RH" 2, BH "2, GH" 2) of the slide unit 1001.

Next, the memory data selector group 10 of FIG.
The internal structure of 05 is shown in FIG. The color conversion table memories M0 to M5 receive the addresses of the address generators 0 to 5 and give the output values of the six grid points necessary for the triangular prism interpolation calculation, but each memory outputs which of the six grid points. It is necessary to grasp whether or not, and appropriately pass the grid point output value to the interpolation calculation circuit accordingly. In other words, the memories M0 to M5 of the color conversion table memory 108 and the grid points a, b (or d), c, e, f (or h) and g must be associated with each other. The memory data selector group 1005 in FIG. 10 is responsible for associating this memory with the grid points.

Now, description will be made from FIG. 17 (A). FIG.
7A shows the internal configuration of the memory data selector 0 (FIG. 10) that selects the output values of the color conversion table memories M0 to M2. When the triangle 905 is selected in FIG. 9B, the output value of the memory M0 is the output value of the grid point a, the output value of the memory M1 is the output value of the grid point b, and the output value of the memory M2 is It is the output value of the grid point c. On the other hand, when the triangle 906 is selected, the output value of the memory M0 is the output value of the grid point b, and the output value of the memory M1 is the grid point c.
, And the output value of the memory M2 are the output values of the grid point a. Therefore, the correspondence relationship changes depending on the position of the triangle on the bottom surface of the oblique triangular prism selected according to the input RGB, and this correspondence relationship can be described as in (Table 7).

[0081]

[Table 7]

That is, the position of the triangle is the remainder RH "m obtained by dividing RH" 1 by 3 as in the address generators 0 to 5 in FIG.
It can be specified by using the remainder BH "mod3 obtained by dividing od3 and BH" 1 by 3, and the memories M0 to M can be compared with FIG. 9B.
It suffices to find the correspondence between 2 and the grid points. As can be seen from (Table 7), there are three connection relationships between the memories M0 to M2 and the lattice points a, b (or d) and c. Therefore, a signal indicating the type of connection relation, SELType0, is provided for the three types of connection relations, and this SELType0 is supplied to the selectors A to C (1604 to 1606). FIG.
(A) DEC0 (1903) is a decoder according to (Table 7) and gives a remainder obtained by dividing BH ″ 1 by 3 MOD3BH
MOD that gives the remainder when (1901) and RH ″ 1 are divided by 3
3RH (1902) is input and SELType 0 is output. The logic of the selectors A to C is as shown in (Table 8).

[0083]

[Table 8]

FIG. 17B shows the color conversion table memory M.
Memory data selector 1 for selecting output values 3 to M5
The internal structure of FIG. 10 is the same as that of FIG. SELType1 indicating the connection relationship between the memories M3 to M5 and the grid points e, f (or h), g is generated according to (Table 9), and the selector D according to (Table 10).
~ F distributes the output values of the memories M3 to M5 to the grid points e, f (or h), g.

[0085]

[Table 9]

[0086]

[Table 10]

From the above, the color conversion table memories M0 to M
5 and grid point data a, b (or d), c, e, f
(Or h) and g are associated with each other, and six grid point output values corresponding to the input RGB data can be supplied to the interpolation calculation.

By the way, the oblique triangular prism interpolation has a positional relationship in which the upper surface and the lower surface are slanted diagonally as shown in FIG. Therefore, the diagonal triangular prism interpolation executed in the outer portion of the input RGB space requires grid points located outside the input RGB space. FIG. 18 is a diagram for explaining the relationship between the input RGB space and grid points. FIG. 18 (A) shows B or R
FIG. 19 is a view of the input space viewed laterally from the axial direction.
(B) is a view looking down from the G direction. In this way, when the dividing method of the present invention is used, 1 from the input definition space
It is necessary to hold the data outside the division unit in the interpolation table. For example, when each axis of the RGB input space is divided into eight equal parts and grid points are set as in this embodiment, (8 + 1)
³ = 729 grid point output values are held in the color conversion table memory, but if data outside the input definition space by one division unit is added to cover the diagonal triangular prism, (8 + 1 + 1 + 1) ² × (8 + 1) = 1089 grid points are required, which requires about 1.5 times as much memory. Therefore, in the present invention, the output value of the grid point located outside the input space is not directly held in the memory, but the memory value is increased by extrapolating the output value of the grid point outside the input space from the output value of the grid point inside the input space. Avoided. Therefore, the extrapolation method will be described next.

Similar to FIG. 18, FIG. 19 shows a case where the RGB input space is viewed from the G axis direction. The shadowed portion is the projection of the upper and lower surfaces of the unit interpolated solid, and the thin shadow oblique triangular prism is selected when Type = 0, and the dark shadow oblique triangular prism is selected when Type = 1.
The black circles represent the interpolation origin a or the top origin e, and the solid arrows represent the difference values between the grid points. When the oblique triangular prism selected according to the input point exists in the RGB input space as shown in FIG. 19, the data in the color conversion table memory may be used as it is.

FIG. 20 shows a case where the lower surface of the oblique triangular prism is out of the RGB input space as a result of the movement of the interpolation origin a described in FIG. In the unit interpolation solid 1901, the input point enters the outer block including the G axis, and the AR shown in FIG.
Formed when belonging to EA (3), lower surface a of the triangular prism
bcd is out of the input space. Unit interpolation solid 1902
Is formed when the input point enters the outer block including the R-axis and belongs to AREA (1) or AREA (3) shown in FIG. 4B, and the lower surface abcd of the oblique triangular prism is out of the input space. ing. The input point of the unit interpolation solid 1903 is B
It is formed when it enters the outer block including the axis and belongs to AREA (2) or AREA (3) shown in FIG. 4 (C), and the lower surface abcd of the oblique triangular prism is out of the input space. The unit interpolation solid 1904 has an input point in the outermost block farthest from the B axis including the R axis, and AREA (1)
Formed when belonging to the lower surface abcd and upper surface efgh
Both go out of the input space. However, in this case, Type
= 1 only the oblique triangular prism is required, and Type =
No input point appears in the oblique triangular column selected by 0.
The unit interpolation cube 1905 is formed when the input point enters the outermost block from the R axis including the B axis and belongs to AREA (2), and both the upper and lower surfaces go out of the input space, but with Type = 0 Only the selected oblique triangular prism is required, and the oblique triangular prism selected at Type = 1 need not be considered.

The output values of the grid points which are outside these input spaces are extrapolated by the following method. First, the unit interpolation solid 190
1 will be described. The interpolation origin a is extrapolated by the difference value between the point c on the lower surface and the point e and the point g on the upper surface as in the following equation.

[0092]

[Equation 14]

Now, assuming that an oblique triangular prism of Type = 0 is selected, since the grid point b also exists outside the input space, it is necessary to extrapolate. However, in order to execute the oblique triangular prism interpolation, as shown in Expressions (6) and (7), the difference value (b)-(a) between the output value (a) of the interpolation origin a and the grid point output value,
Since (c)-(a) is required, (b)-(a) uses the difference value (f)-(e) of the upper surface instead of directly obtaining the output value (b) of the grid point b, For c)-(b), the difference value (g)-(f) on the upper surface is used. That is, (m) in equation (6) is

[0094]

(Equation 15)

It can be calculated by

The purpose of substituting the difference value on the lower surface with the difference value on the upper surface is to make the extrapolation operation circuit as small as possible. With this method, the adder / subtractor for the extrapolation operation uses the interpolation origin a. It is necessary only for, and the others can be replaced simply.

When the oblique triangular prism of Type = 1 is selected, the difference value (d)-(a) of the lower surface is changed to the difference value (h) of the upper surface.
-(E), and the difference value (c)-(d) on the lower surface may be replaced with the difference value (g)-(h) on the upper surface. (M in the case of Type = 1 obtained by the equation (9)) ) Is

[0098]

(Equation 16)

It can be calculated by

Next, an extrapolation method in the unit interpolation solid 1902 will be described. The solid is different from the unit interpolation solid 1901 in that the grid point d exists in the input space.
For this reason, when the oblique triangular prism of Type = 1 is selected, the difference values (c)-(d) actually exist, and it is not necessary to replace them from the upper surface. Therefore, the unit interpolation solid 1902 obtains the interpolation origin a by the equation (17), and when the oblique triangular prism of Type = 1 is selected, the difference value (d)-(a) is the difference value (h)-(e) of the upper surface. When the oblique triangular prism of Type = 0 is selected, it is interpolated by the equation (18).
The interpolation formula in the case of Type = 1 is shown below.

[0101]

[Equation 17]

Next, an extrapolation method in the unit interpolation solid 1903 will be described. The solid is different from the unit-interpolated solid 1901 in that the lattice point b exists in the input space.
For this reason, when the oblique triangular prism of Type = 0 is selected, the difference values (c)-(b) actually exist, and it is not necessary to replace them from the upper surface. Therefore, the unit interpolation solid 1903 obtains the interpolation origin a by the equation (17), and when the oblique triangular prism of Type = 0 is selected, the difference value (b)-(a) is the difference value (f)-(e) of the upper surface. Is replaced with, and when the oblique triangular prism of Type = 1 is selected, the interpolation is performed by the equation (19).
The interpolation formula in the case of Type = 0 is shown below.

[0103]

(Equation 18)

Next, an extrapolation method in the unit interpolation solid 1904 will be described. When obtaining the output value (a) of the interpolation origin a in the solid, the method used in the unit interpolation solids 1901-1903 cannot be applied. This is because the grid point g exists outside the input space and the difference value (g)-(e) does not actually exist.
Therefore, the interpolation origin a is the difference value (h) − from the grid point d to the upper surface.
Extrapolate using (e) as follows.

[0105]

[Equation 19]

As for the difference value, since (d)-(a) on the lower surface and (g)-(h) on the upper surface do not actually exist, (h)-(e) on the upper surface and (c)-(d) on the lower surface, respectively. Replace with.
From the above, (m) of Type = 1 in equation (9) is

[0107]

(Equation 20)

(N) of the equation (10) is obtained by

[0109]

(Equation 21)

It is calculated by As already mentioned,
In the unit interpolation solid 1904, since the input point does not appear in the oblique triangular prism of Type = 0, the expressions (23) and (24) are sufficient.

Next, an extrapolation method in the unit interpolation solid 1905 will be described. When obtaining the output value (a) of the interpolation origin a in the solid, the unit interpolation solids 1901-190 are present because the grid point g exists outside the input space as in the unit interpolation solid 1904.
Equation (17), which is the extrapolation method in 3, cannot be used. Therefore, extrapolation is performed as follows using the difference value (f)-(e) from the lattice point b to the upper surface.

[0112]

(Equation 22)

As for the difference value, since (b)-(a) on the lower surface and (g)-(f) on the upper surface do not actually exist, (f)-(e) on the upper surface and (c)-(b) on the lower surface, respectively. Replace with.
From the above, (m) of Type = 0 in equation (6) is

[0114]

(Equation 23)

The value (n) of Type = 0 in the equation (7) is given by

[0116]

(Equation 24)

It is calculated by As already mentioned,
In the unit interpolation solid 2405, since the input point does not appear in the oblique triangular prism of Type = 1, the expressions (26) and (27) are sufficient.

FIG. 21 shows a case where the upper surface of the oblique triangular prism comes out of the RGB input space as a result of the movement of the interpolation origin a described in FIG. However, the unit interpolation solids 2004 and 2005
Are the same as 1904 and 1905 in FIG. 20, respectively.
The unit interpolation solid 2001 is formed when the input point enters the outer block located at the farthest position on the diagonal from the origin and belongs to AREA (0) shown in FIG. 4A, and the upper surface efgh of the oblique triangular prism is formed. It is outside the input space. The unit interpolation solid 2002 is formed when the input point is the furthest from the B axis and enters the outer block that does not include the R axis and belongs to AREA (1) or AREA (0) in FIG. Is out of the input space. Similarly, the unit interpolation solid 2003 is formed when the input point is farthest from the R axis and enters the outer block that does not include the B axis and belongs to AREA (2) or AREA (0) in FIG. efgh is out of the input space.

The output values of the grid points outside these input spaces are extrapolated by the following method. When the upper surface of the oblique triangular prism is outside the input space, unlike the case where the lower surface is outside the input space, extrapolation can be executed only by replacing the difference value. In the unit interpolation solid 2001, four difference values (g) on the upper surface-
(F), (f)-(e), (g)-(h), (h)-
(E) is the difference value of the lower surface (c)-(b), (b)
-(A), (c)-(d), (d)-(a) are substituted. Therefore, (n) in equation (7) in the oblique triangular column of Type = 0 is

[0120]

(Equation 25)

It is calculated as follows. (N) in the formula (10) in the oblique triangular prism of Type = 1 is

[0122]

(Equation 26)

Is calculated.

Next, in the unit interpolation solid 2002, the difference values (g)-(f), (f)-(e), and (g)-(h) of the upper surface are respectively calculated as the difference values (c)-(b) of the lower surface. , (B)-
Substitute with (a), (c)-(d). Therefore, Type =
(N) of the formula (7) can be calculated by the formula (28) in the case of the oblique triangular column of 0, and the formula (10) can be calculated in the case of the type of the oblique prism of Type = 1.
(N) is

[0125]

[Equation 27]

Is calculated as

Similarly, in the unit interpolation solid 2003, the difference values (g)-(f), (g)-(h), (h)-(e) of the upper surface are obtained.
The difference values (c)-(b), (c)-
Substitute with (d) and (d)-(a). Therefore, Type =
(N) of the formula (10) in the oblique triangular prism of 1 is the formula (29)
Can be calculated by the formula (7) in the oblique triangular prism of Type = 0
(N) is

[0128]

[Equation 28]

Is calculated. Table 11 shows the extrapolation calculation method of the interpolation origin a and the replacement method of the difference value in the extrapolation of the lower surface described with reference to FIG. 20, and the case is determined by the position of the triangle by RH ′ and BH ′.

[0130]

[Table 11]

Further, (Table 12) shows the extrapolation calculation method of the interpolation origin a and the replacement method of the difference value in the extrapolation of the upper surface described in FIG. 21, which is also classified according to the position of the triangle by RH 'and BH'. went.

[0132]

[Table 12]

20 and 21, the lower surface of the triangular prism is the ab (d) c plane, and the upper surface of the triangular prism is ef (h) g.
Since the case of the surface is described, the inversion of the vertical interpolation direction described in FIG. 15 is not added. That is, FIG.
Contrary to cases 0 and 21, the upper surface of the oblique triangular prism is ab (d).
This is the case of the c-plane, and the lower surface of the triangular prism is the ef (h) g-plane. This case corresponds to the case of GH = odd number, the interpolation origin that goes out of the input RGB space is e, and the relationship of exchanging the difference values is also reversed. (Table 13) and (Table 14) summarize the extrapolation methods when GH = odd.
(Table 13) corresponds to (Table 11), where a is for e and b is for f
, C is replaced by g, and d is replaced by h. (Table 1
4) corresponds to (Table 12).

[0134]

[Table 13]

[0135]

[Table 14]

Next, the operation of the portion related to extrapolation in FIG. 1 will be described. In the extrapolation, as described with reference to FIGS. 20 and 21, the interpolation origin a existing outside the input RGB space is extrapolated from the grid point c or b (d) using the difference value. Once the output value of the interpolation origin a is determined, the output values of other grid points are not directly obtained, but extrapolation is indirectly performed by using the difference value between the grid points that actually exist in the RGB space, and the oblique triangular prism is used. This is to execute interpolation. In order to execute such extrapolation, first, a case-specific signal of the extrapolation method must be generated according to the position of the interpolation origin a, and the calculation of the interpolation origin a and replacement of the difference value must be executed according to this signal. It is the outer block detection unit 107 of FIG. 1 that detects the case classification of this extrapolation method, and this output signal EDB corresponds to the case classification signal of the above-mentioned extrapolation method. Further, the extrapolation unit 110 in FIG. 1 has a role of extrapolating the interpolation origin a, and the origin / difference value selector 111 stores the data of the actual interpolation origin a according to the EDB and the data of the interpolation origin a as the extrapolated value. The selected difference value and the difference value necessary for extrapolation are supplied to the oblique triangular prism interpolation calculation unit 112.

FIG. 22 shows the internal structure of the outer insertion portion 110.
In the extrapolation unit 110, a, c, e,
Only four points g are input, and after addition and subtraction according to FIG. 22, an extrapolation output a ′ at the interpolation origin a and an extrapolation output e ′ at the interpolation origin e are output. According to (Table 11) to (Table 14), the extrapolation calculation of the interpolation origin a and the interpolation origin e

[0138]

(Equation 29)

[0139]

[Equation 30]

[0140]

[Equation 31]

[0141]

(Equation 32)

A total of four types are required. Therefore, four adders and four subtractors are required, for a total of eight arithmetic units. Therefore, in order to reduce the hardware scale, a method of utilizing the selectors A to F of FIG. 17 to execute extrapolation with four arithmetic units as shown in FIG. 22 will be described.
Figure 22

[0143]

[Expression 33]

[0144]

(Equation 34)

## EQU14 ## The equation (32) is executed, but it can be calculated by replacing a in the equation (36) with c. To execute the equation (33), a in equation (36) may be replaced with b (or d), and g may be replaced with f (or h). Similarly, when executing the equation (34), the equation (3
It can be calculated by replacing e in 7) with g, and when executing formula (35), e in formula (37) can be replaced with f (or h) and c can be replaced with b (or d). (The replacement content is also shown in FIG. 22).

By the way, since a which is replaced by the equation (36) cannot directly supply data from the color conversion table memory and is extrapolated using data of other grid points, data of grid point a is supplied from the memory. The line to do is free. Therefore, using the selectors A to F of FIG.
By passing replacement data necessary for extrapolation to the extrapolation unit 110 through this empty line, extrapolation interpolation of the interpolation origins a and e can be executed by the small-scale hardware in FIG. From the above, the operation rules of the selectors A to F ((Table 8),
(Table 10) needs to be changed by adding the output EDB of the outer block detection unit 107 of FIG. 1 to the criterion.

Then, how should the EDB be generated in order to replace the data as shown in FIG. According to FIG. 22, there are four types of data replacement.
Therefore, five types of switching signals may be output in combination with the case where extrapolation is not used. Also, the replacement of this data is switched depending on the position of the bottom surface as shown in FIGS. Therefore, it is possible to describe the position of the bottom surface of the oblique triangular prism R
HDB, HH ', GH0' which can judge whether the G axis is even or odd, and the output PRISM of the oblique-trigonal-column determining unit 105 are used for case classification so that five kinds of switching signals can be generated.
Should be generated. Table 15 shows the case classification method and the arithmetic expressions of the outputs a ′ and e ′ of the extrapolation unit in each case. The outer block detection unit 107 outputs five types of switching signals to the selectors A to F in FIG. From the above, the switching method of the selectors A to C (Table 8) is changed to (Table 16) by adding EDB0 to 4 and the switching method of the memory (Table 15). Also selector D
~ F shows how to switch (Table 10) EDB0 ~ 4 and (Table 15) showing how to switch the memory is added (Table 17)
Is changed to. The internal configuration of the memory data selectors A to F shown in FIG. 17 is changed as shown in FIG.
DB0 to 4 are input to the selectors A to F.

[0148]

[Table 15]

[0149]

[Table 16]

[0150]

[Table 17]

From the above, the extrapolation of the interpolation origin a and the extrapolation of the interpolation origin e are executed by the extrapolation section 110 by the output EDB of the outer block detection section 107 and the memory data selectors 0 and 1 configured as shown in FIG. it can.

Next, FIG. 24 shows the difference value generator 109 of FIG.
The internal configurations of the extrapolation unit 110, the origin / difference value selector 111, and the oblique triangular prism interpolation calculation unit 112 are shown. However, the extrapolation unit 11
The internal structure of 0 is the same as that of FIG. In the oblique triangular prism interpolation calculation, when PRISM = 0, equations (6) to (8) are executed, and when PRISM = 1, equations (9) to (11) are executed. According to these calculation methods, it is first necessary to create a difference value between grid points. The process of generating this difference value is the difference value generation unit 109.

When the extrapolation calculation is performed on the difference value and the interpolation origins a and e, the switching operation must be executed according to (Table 11) to (Table 14). This switching is performed by the origin / difference value selector 111. The origin / difference value selector 111 includes an origin switching unit 2301 that switches the interpolation origins a and e and a difference value switching unit 2302 that switches the difference value. The origin switching unit 2301 switches the interpolation origin a with SEL5 and switches the interpolation origin e with SEL6. The switching signals CNG5 and CNG6 given to the SEL5 and SEL6 are generated according to the EDB0 to 4 of the outer block detection unit 107 as shown in (Table 22).
The case of EDB = 0 is a case where all 6 grid points exist in the RGB input space and extrapolation does not occur. Decoder 1
(2304) follows EDB0 according to the rules of (Table 22).
CNG5 and CNG6 are generated from ~ 4. (Table 18) ~
Table 21 shows how to switch the difference value, and EDB5 to 18 are assigned to each pattern. The switching of the four difference values D1 to D4 is executed by the selectors SEL1 to SEL4 in the difference value switching unit 2302. According to (Table 11) to (Table 14), RH ', BH', G
Depending on H0 'and PRISM, it is possible to set the case classification of the difference value switching method, and if EDB is assigned to each case classification, the desired difference value switching can be performed. (Table 18) to (Table 21) show the method of replacing four difference values, and it can be seen that there are 14 patterns of difference value switching. EDB set for the extrapolated value of the origin is 0
Since it is 4, EDBs for switching the difference value use 5 to 18. Four selectors SEL1 to SE shown in FIG.
When comparing the configuration of L4 and the method of replacing the difference values in (Table 18) to (Table 21), the selectors SEL1 to SEL4
The switching signals CNG1 to CNG4 supplied to
It can be designed like 2). Decoder 0 (2303) uses EDB5-18 to CN according to the rules in (Table 22).
G1, CNG2, CNG3, and CNG4 are generated.

[0154]

[Table 18]

[0155]

[Table 19]

[0156]

[Table 20]

[0157]

[Table 21]

[0158]

[Table 22]

The oblique triangular prism interpolation calculation unit 112 receives the outputs D1 'to D4' of the origin / difference value selector 111 and uses the equations (6) to (8) (when PRISM = 0) or the equations (9) to (). 11) Execute (if PRISM = 1). The selectors SEL7 to SEL10 perform switching of D1 'to D4' according to PRISM, and this logic follows (Table 23).

[0160]

[Table 23]

The weighting factors (RL-GL) ', (BL-GL)' and GL 'given to the multipliers MUL1 to MUL5 are shown in FIG.
Is given from the weight generation unit 104 of.

The color conversion apparatus based on the oblique triangular prism interpolation method configured as described above has a feature that it can calculate the minimum value frequently used in the printing system (MIN operation) without interpolation error. ing. FIG. 25 illustrates a surface where the MIN calculation result is constant with respect to the input signal in the rectangular coordinates of input RGB. As can be seen in this figure, three boundary surfaces 2 of R = B, B = G, and R = G shown by hatching 2
Three areas (0) divided by 401 to 2403
Although there is differential continuity in (2), a strong differential discontinuity occurs at the boundary of the three surfaces, so a large error occurs when the linear interpolation calculation is performed in the boundary area. FIG. 26 shows a case where the triangular prism interpolation is performed based on the calculated value of the MIN calculation on the gray axis from the O point to the W point in FIG. 25 and the first example of the prior art (Japanese Patent Laid-Open No. 5-75848). The error with the interpolation value is shown. On the gray axis, the intersection of the above three boundary planes is touched, so naturally a strong differential discontinuity occurs. Original M
An unbearable ripple error occurs in the IN calculation calculation.

FIG. 27 is an exploded view of the differential discontinuous surfaces. The cross-hatched portion corresponds to the differential discontinuous surfaces 2401 to 2403 shown in FIG. Therefore, if the input data space is divided and linear interpolation is performed so that the differential discontinuous surface is not included in the interpolation solid, the ripple generated in FIG. 26 will not occur.

Based on the above consideration, when the input space is divided, the division examples shown in FIGS. 28A and 28B are derived. FIG.
The differential discontinuity surface in 8 (A) and (B) is indicated by a thick line 24
01 to 2403, and this surface corresponds to the division surface position. Therefore, if the division of FIGS. 28A and 28B is performed, no error occurs in the MIN calculation.

FIGS. 2 (A) and 2 (B) show a large scale of the unit division solid of the division of FIG. 28 (A), and the upper surface efgh of the RGB unit cube is exactly in the R direction and B with respect to the lower surface abcd.
1 unit in the direction, and the R = B plane (240
In 1), it is divided into two types of oblique triangular prisms of Type = 0 and Type = 1.

On the other hand, FIGS. 2C and 2D show FIG. 28B.
The unit of division is a large representation of the divided solid, which is just RGB.
An upper surface efgh of the unit cube is displaced by one unit from the lower surface abcd in the R direction and the B direction, and is divided by the R = B surface (2401) shown in FIG. 27. Two types of parallelepipeds of Type = 0 and Type = 1 It shows that it is classified into.

As described above, the oblique triangular prism interpolation method is an effective interpolation method capable of calculating the MIN calculation frequently used in the printing system without any interpolation error. In the present invention, the input RGB that is always generated when the oblique triangular prism is set in the interpolation space.
By increasing the capacity of the color conversion table memory by extrapolating the output values of the grid points outside the space from the output values of the grid points inside the input RGB space.

[0168]

As described above, according to the present invention, the occurrence of "ripple" in the MIN operation frequently used in color conversion can be completely avoided not only in the achromatic color direction but also in any direction in the input color space. By executing the address calculation to generate a one-dimensional address from the address, the color conversion table can be used with continuous addresses, the memory is not wasted, and the oblique conversion prism interpolation method and the triangular prism interpolation method can be easily performed without rewriting the color conversion table. Can be switched, and multiple grid point data to be read simultaneously can be stored in different memories to realize simultaneous access to shorten the processing time. It is not necessary to have the output values of the grid points outside the required input space directly in the memory, and the memory capacity can be reduced. Rukoto can.

[Brief description of drawings]

FIG. 1 is a block connection diagram showing a configuration of a color conversion device according to an embodiment of the present invention.

FIG. 2 is a conceptual diagram of division of an input color space in the same embodiment.

FIG. 3 is a diagram showing a relationship between a skew coordinate form and a rectangular coordinate form in the embodiment.

FIG. 4 is a diagram showing a change in the origin of the oblique triangular prism according to the comparison result of the comparison unit in the example.

FIG. 5 is a conceptual diagram of an oblique triangular prism interpolation calculation in the same embodiment.

FIG. 6 is a block connection diagram of a comparison unit in the example.

FIG. 7 is a block connection diagram of a weight generation unit in the embodiment.

FIG. 8 is a block connection diagram of an oblique triangular prism determination unit in the same embodiment.

FIG. 9 is a diagram showing positions of color conversion table memory and grid points in the same embodiment.

FIG. 10 is a block connection diagram of a memory access unit and a color conversion table memory in the embodiment.

FIG. 11 is a diagram showing a correspondence of selection between an oblique triangular prism and a color conversion table memory in the embodiment.

FIG. 12 is a view showing a correspondence of selection between an oblique triangular prism and a color conversion table memory in the embodiment.

FIG. 13 is a diagram showing positions of color conversion table memory and grid points in the embodiment.

FIG. 14 is a block connection diagram of an address generation unit in the memory access unit according to the embodiment.

FIG. 15 is a diagram showing a calculation execution method of an oblique triangular prism interpolation calculation unit in the embodiment.

FIG. 16 is a diagram showing a change content of the weight generation unit according to the calculation execution method of the diagonal triangular prism interpolation calculation unit in the embodiment.

FIG. 17 is a block connection diagram of a memory data selector group in the color conversion table memory in the embodiment.

FIG. 18 is a diagram showing a state in which an input color space is included in an oblique triangular prism in the example.

FIG. 19 is a diagram showing extrapolation interpolation in the example.

FIG. 20 is a diagram showing extrapolation interpolation in the example.

FIG. 21 is a diagram showing extrapolation interpolation in the example.

FIG. 22 is a block connection diagram of an extrapolation unit in the example.

FIG. 23 is a diagram showing the contents of change of the memory data selector group according to the calculation execution method of the oblique triangular prism interpolation calculation unit in the embodiment.

FIG. 24 is a block connection diagram of a difference value generation unit, an origin / difference value selector, an extrapolation unit, and an oblique triangular prism interpolation calculation unit in the embodiment.

FIG. 25 is a diagram showing a constant output surface and a differential discontinuous surface when MIN calculation is performed by table reference and interpolation in the embodiment.

FIG. 26 is a diagram showing a calculation result of an error between an interpolated value and a calculated value when a MIN operation is performed by a conventional table reference and interpolation.

FIG. 27 is a diagram showing a differential discontinuity surface when MIN calculation is performed by table reference and interpolation in the same example.

FIG. 28 is a diagram showing division of an input color space in the example.

[Description of sign]

 101 Image Input Unit 102 Comparison Unit 103 Interpolation Origin Setting Unit 104 Weight Generation Unit 105 Oblique Triangular Judgment Unit 106 Memory Access Unit 107 Outer Block Detection Unit 108 Color Conversion Table Memory 109 Difference Value Generation Unit 110 Extrapolation Unit 111 Origin / Difference Value Selector 112 Oblique triangular prism interpolation calculation unit

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Katsuhiro Kanamori 3-10-1 Higashisanda, Tama-ku, Kawasaki, Kanagawa Matsushita Giken Co., Ltd. Chome 10-1 Matsushita Giken Co., Ltd.

Claims (3)

[Claims] 1. A color conversion device, wherein a grid point existing outside the input color image signal space is extrapolated by a grid point existing inside the input color image signal space and a difference value between the grid points. 2. A pixel input section for dividing a color image signal into an upper bit section and a lower bit section, a comparing section for comparing the lower bit section and outputting the magnitude relationship, and a comparing section from the upper bit section to the comparing section. Access from the output of the interpolation origin setting unit, the interpolation origin setting unit that determines the position of the interpolation origin in accordance with the output of, the color conversion table memory that stores the output value at each grid point in the input color image signal space, A memory access unit that generates an address of a color conversion table memory to be supplied and supplies the address to the color conversion table memory, and a grid point selected by the address generated by the memory access unit is an input color image signal. An outer block detection unit that detects whether or not the outer block exists outside the space, a difference value generation unit that generates a difference value between the grid point output values, and the case An extrapolation unit for extrapolating output values of grid points outside the input color image signal space from child point output values; a difference value selector for switching the output of the difference value generation unit according to the output of the outer block detection unit; An origin selector that switches the output value of the insertion unit and the output value of the grid point corresponding to the interpolation origin, a weight generation unit that generates an interpolation weight coefficient along the oblique triangular column in the lower bit portion, and a magnitude relationship between the weight coefficients. 2. A color conversion device, comprising: an oblique triangular prism determination unit that selects an oblique triangular prism, and an oblique triangular prism interpolation calculation unit that interpolates an output value using the output of the difference value selector and the weighting coefficient. 3. The output value at each grid point in the input color image signal space is stored in the color conversion table memory so that the output values corresponding to the six end points of the oblique triangular prism selected in the oblique triangular prism determination unit do not overlap. The memory access unit simultaneously supplies the addresses of the output values corresponding to the six end points of the selected oblique triangular prism to the six storage units. The color conversion device according to claim 2.