JP2947015B2 - Color conversion device and color conversion method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an application in which a color image signal or a color video signal is input and arbitrary color coordinate conversion and color conversion are performed in real time, for example, a color requiring high-speed color correction and color correction. It relates to scanners, color cameras, color hard copy devices, color display devices that require accurate color proofing, color correctors that change video images in real time, video editing devices, and color conversion devices that perform color identification. is there.

[0002]

2. Description of the Related Art Conventionally, in image processing of a monochrome image, information of one pixel of an image is one-dimensional information called lightness (density), and lightness conversion is so-called gamma curve conversion.
LUT (Look Up Table) for various nonlinear curves
Could be converted in real time.
Even if the image to be handled becomes a color image, it is treated as three monochrome images of an R (red) plane, a G (green) plane, and a B (blue) plane for the purpose of performing color conversion in real time, and each image is independent. Often converted by LUT. However, in this type of processing, the color conversion that can be handled is essentially out of the one-dimensional processing range, and is in the form of R ′ = hR (R), G ′ = hG (G), B ′ = hB (B) Only color conversion is possible.

In the color image processing, the information of one pixel is three-dimensional information (R, G, B), and the color conversion in the original meaning is a three-dimensional conversion R '= fR (R, G, B) G '= fG (R, G, B)
B ′ = fB (R, G, B).

For example, in color image processing of a hard copy system, complicated color conversion such as “colors belonging to a specific hue increase saturation” is required. I want to convert only the background into an achromatic color ". These color conversions mathematically have one output as a function of three inputs and belong to the three-dimensional conversion described above. But,
If these are converted using a general-purpose table, assuming that one color is an 8-bit signal, 16 (Mb
yte) memory capacity is required. Therefore, there is a demand for a configuration which is hardware capable of performing three-dimensional color conversion for any color conversion in general and in real time and has a memory capacity as small as possible. On the other hand, the input color space is divided into a plurality of color spaces for the main purpose of color hard copy and color correction of a color scanner, and color correction information located at the vertices is held as a color conversion memory, In the calculation, there is an example of a color signal interpolation method in which a plurality of pieces of color correction information are selected, weighted, and subjected to interpolation output (Japanese Patent Publication No. 58-16180).

According to the contents of the publication, a unit cube, which is a basic solid in a three-dimensional color signal space, is set for the interpolation process, and the unit cube is divided into a plurality of tetrahedrons to form a tetrahedron.
A concept is disclosed that simplifies interpolation calculations from output signals at vertices. This is an application to a color correction device of a color scanner device. An RGB 8-bit signal is divided into an upper signal and a lower signal by 4 bits each, and the lower signals are determined by a comparator, and at the same time, four kinds of weights are generated. Input to the container. On the other hand, the upper signal is address-modified by the selector and the adder, the address corresponding to the four vertices of the tetrahedron is obtained, and the color conversion table memory is read at these four different addresses a total of four times. An interpolation output is obtained by adding the weighted coefficients in parallel with a multiplier and adding them. One unit cube is divided into six tetrahedrons.

On the other hand, as a second example of the prior art, there is a technique for storing and processing difference values at four vertices in a color conversion apparatus using the same tetrahedral interpolation method (US Pat. 477, 833). In this example, the 8-bit input color signals cyan, magenta, yellow (CY
When converting from AN, MAG, YEl) to cyan, magenta, and yellow with 8-bit color correction, the input color space is divided into tetrahedrons, and interpolation is performed using output difference values at the vertices.

Further, as a third example of the prior art, there is a color separation image correcting apparatus using the same tetrahedron-based interpolation method (Japanese Patent Laid-Open Nos. 2-87192, 2-286866, and 2-28667). Gazette, 2-2-2868
No.). It is important that these color conversion devices are in a table format, and non-linear free color conversion can be performed at high speed and flexibly.

[0008]

By the way, in the above-mentioned prior art, in order to reduce the operation required for the interpolation, in the interpolation of the three-dimensional space, the reference point, that is, the representative grid point whose output value is stored in the memory is used. It uses the division into tetrahedrons that requires the least access. This simplifies the interpolation hardware the most.

[0009] However, tetrahedral division is convenient when treating three axes equally in a color mixture system such as RGB in a color space.
When it is desired to separate the color space on the lightness axis and the chromaticity axis and perform both interpolations independently, there is a disadvantage that the analysis of the interpolation result becomes complicated. For example, if a color is to be selectively controlled by weighting only in the chromaticity plane without changing the brightness, such as for a color corrector, it is necessary to perform subtle curve color conversion only in the chromaticity plane. Become. However, in the conventional tetrahedral interpolation, there is a problem that interpolation is always performed in a space of three axes and interpolation of lightness and chromaticity cannot be separated.

Second, not only the interpolation using the tetrahedron and the interpolation using the cube but also the conventional interpolation type color conversion technique has a disadvantage that the interpolation curve is not accurate in some cases. . For example, in a color reproduction process of hard copy, the input value is maintained at 0 until a certain threshold value, and the output characteristic rises from the threshold point T or suddenly rises as seen in a process called “clipping characteristic”. Therefore, a curve sticking to the maximum value at the input point T 'at the maximum value could not be accurately interpolated. Also, 8
In a system with a bit input and an 8-bit output, it is possible to realize the so-called “through characteristic” in which the same output is given to all color signal inputs because 256 cannot be expressed and stored next to 255 because both cannot be stored. could not. The through characteristic is fundamental in a digital circuit, and if it is not realized, a signal that has been passed many times due to the intention of through will rapidly deteriorate.

Third, in the application to a color corrector or the like, the color conversion of a moving image or a still image on a display is frequently changed. During the operation of rewriting the color conversion table from the CPU side, the CPU accesses the table memory, so that the displayed image and video are disturbed, and the display temporarily disappears. This has been a very serious problem in applications such as a color corrector in which a user wants to perform color conversion instantaneously while viewing an image.

Fourth, an application in which a display or a hard copy paper is divided into several areas and different color conversion is performed for each area, or as a more flexible application in which color conversion is changed for each pixel of an image. There is. In this case, it is necessary to switch the color conversion table at high speed for each area and each pixel. However, the prior art does not support this table switching.

Fifth, as a problem that has conventionally occurred in interpolation using a cube, there is a problem of an interpolation error in MIN (minimum value) calculation. In the field of color hard copy, when generating black (black BK), it is necessary to generate the minimum value among the three signals of the input signals cyan C, magenta M and yellow Y. However, in interpolation using a cube or interpolation using a triangular prism, a three-color signal is input and this MIN signal is input.
When the calculation is performed, a ripple-like interpolation error occurs in the interpolated curve, which is visually unfavorable.

[0014]

In order to achieve the above object, the present invention divides a three-dimensional space formed by input color signals into a plurality of unit cubes and unit rectangular parallelepipeds, and then divides them into two triangular prisms. This is an interpolation solid. For this reason, a triangular-prism-division-type interpolation calculation unit is employed, which determines which triangular prism contains the input color, and employs a triangular-prism-division-type interpolation method using an output value at each vertex of the triangular-prism area.

In addition, not only a conventional positive number area but also a negative number area can be stored as a numerical value to be stored in the color conversion table memory, so that conversion by clipping characteristics can be realized by interpolation. By providing a displacement amount correcting means in the calculation unit, complete through characteristics by interpolation are realized. At this time, in order to reduce the amount of memory, a color conversion table memory capable of expressing and storing a numerical value in a negative area by reducing the accuracy even with the same number of bits,
It has a signed multiplier and an adder.

In order to cope with interactive color adjustment on a display such as a color corrector, a color conversion table is transferred within a horizontal blanking period and a vertical blanking period of a video signal, and an image or a video being displayed is transferred. It is provided with a host interface unit for performing color conversion table data transfer control for changing color conversion without disturbing.

Also, a plurality of color conversion tables are provided for different color conversion processes, so that the color conversion table used for each of a plurality of rectangular areas set in the internal area processing unit can be switched on a pixel basis. It is possible to switch the internal color conversion table for each pixel in response to a color conversion table compulsory setting signal from.

[0018]

In the color conversion apparatus of the present invention, the simplest linear interpolation is performed independently on the lightness axis and the color difference plane for the lightness color difference type signal using a triangular prism. As a result, the interpolation curve is clarified in the color control in the chromaticity plane such as the color adjustment, and the color adjustment, which has conventionally tended to make the interpolation result unclear, is easily performed. Furthermore, when performing color conversion using a color space that uses lightness and color difference as input, the interpolation accuracy is better than the conventional interpolation using tetrahedral division, especially in the visually important gradient direction. The effect is that

[0019]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The concept of the above-described operation in the present invention will be described below.

For example, a color conversion apparatus according to the present invention performs color coordinate conversion to an NTSC-rgb space using a CIE-LAB space in which a lightness L axis and chromaticity A and B axes are separated as inputs. Consider an experiment in which interpolation is performed using. In this case, the L-axis corresponds to the Y-axis, the A-axis corresponds to the Cr-axis, and the B-axis corresponds to the Cb-axis. The color conversion tables are completely the same, and only the difference in the interpolation method between tetrahedral division interpolation and triangular prism division interpolation is a problem. So that A as an 8-bit pixel input to the color converter
FIGS. 10 and 11 collectively illustrate three rgb curves after interpolation when pure gray in which = B = 0 is changed from L = 0 to 255. In FIG. 10 of the tetrahedral division type interpolation, rgb, particularly r, is wavy in an uneven shape.
In the triangular prism division type interpolation of 1, interpolation is performed smoothly.
As described above, the interpolation using the triangular prism division has an effect that color conversion including nonlinearity can be visually and smoothly executed by matching the lightness direction with the main axis of the triangular prism due to the feature of the shape.

Next, the operation of the present invention in a so-called "clipping characteristic", that is, a conversion in which a constant value is output when an input is equal to or less than a threshold value is considered. FIG.
2 indicates a grid point position (0, 6
12, 128, 192, 256). FIG. 12 (A) shows the interpolation result according to the prior art, and FIG. 12 (B) shows the interpolation result according to the present invention in one dimension. Indicates the input on the horizontal axis and the output on the vertical axis. The conversion to be realized is a conversion that increases from the threshold value T and reaches the maximum value at T ′, and is represented by a dotted line 1201.
In order to realize this conversion by appropriately selecting the stored value at the lattice point position according to the conventional technique, for example, 120
As shown in FIG. 2, near the threshold value T or T ′, the result is greatly different from the original transform. This is because the threshold value T or T ′ is changed to the nearest grid point position, for example, the threshold value T is changed to 64, so that the threshold processing which is the initial intention is performed at the grid point position. Because it is only done in
On the other hand, in the present invention, the color conversion table memory unit can store numerical values in the range from (-256) to 255 as the stored values at the lattice points. Is generated by an interpolation operation, and then the displacement amount
The threshold values T and T ′ are obtained by displacing the interpolated value in the positive direction only, and finally replacing the value of 0 or less and 255 or more with 0 and 255 by the output data limit unit.
Thus, an interpolation result 1205 having the following formula is generated.

The operation for realizing the conversion of the so-called through characteristic in which the input is output as it is will be described. FIG. 13A shows an interpolation result according to the conventional technique.
(B) shows the results of the present invention in a one-dimensional input / output relationship. For the desired through characteristic 1301, the output at each grid point position at (0, 64, 128, 192, 256) on the input axis is respectively (0, 64, 128, 192, 2).
56), a through characteristic is realized when linear interpolation is performed. However, in the prior art, the value that can be stored in the color conversion table memory is usually 8 bits, and in this case, 0 to 25
Since only values up to 5 can be stored, 256 has to be stored as 255 as the next best measure. As a result 19
In the last interpolation section from 2 to 256, the interpolation result cannot realize the through characteristic, and an interpolation result such as 1302 that cannot be said to have the through characteristic occurs. On the other hand, in the present invention, since a negative number can be stored, the stored value at the input grid point is set to (−128, −64, 0, 64, 128) as shown in FIG. Is realized, the displacement amount correction means can add the displacement amount value 128 indicated by 1304 as an offset, and as a result, a complete through characteristic of 1305 can be realized.
There is an effect.

When a so-called MIN operation is performed to obtain the minimum value from the three input signals, a special color difference reference signal is selected from the three input signals, and the difference from the remaining two signals is a color difference. By performing color difference conversion, a state of lightness color difference input is created, and by combining this with the feature of triangular prism division, there is an effect that unnatural ripples are not generated after interpolation.

Furthermore, when color conversion is performed on a video signal in real time, it is necessary for the image before conversion to instantly change to a different color without disturbing the image when the color conversion table is rewritten from the host. Data writing to the color conversion table in the host interface unit is controlled using a video blanking period a plurality of times, so that the color can be changed without disturbing the video.
Also, there is no time to rewrite the color conversion table.
Since different color conversions can be performed by switching the color conversion tables that exist in multiple planes depending on the image plane area while drawing a frame image, different color conversion processing is performed in different areas on the screen. Also make it feasible.

Hereinafter, a first embodiment of the color conversion apparatus of the present invention will be described with reference to FIG. Note that the input color signal is originally a lightness / color difference separation signal such as lightness Y and color difference CrCb. Actually, for the versatility, the three-color separation color signal Red (hereinafter referred to as R), Green (hereinafter referred to as G), Blue
(Hereinafter referred to as B), any three types of color signals can be input. However, the following description is based on the name of the axis fixed to the triangular prism divided area in the color space regardless of the type of the actually input color signal. Are used as names Y, Cr, and Cb. In this embodiment, the image input / output signal is configured to correspond to up to 10 bits.
Since bits are often used, both will be described as appropriate.

In FIG. 1, input signals Y, Cr, and Cb are input to a pixel input unit 101 as 10-bit digital signals representing a color image. Before these signals are input to the address generation unit 102, the Cr and Cb signals are separated into upper 3 bits and lower 7 bits, and the Y signal is converted into the upper 3 bits and lower 7 bits in the normal mode and the address extension mode. Sometimes it is divided into upper 4 bits and lower 6 bits. In FIG. 1, upper signals (UY, UCr, UCb)
111 and lower signals (DY, DCr, DCb) 112. The normal mode refers to YCrCb as shown in FIG.
As a result of dividing each axis of Y, Cr, and Cb in the color space into divided areas designated by 3, 3, and 3 bits by an upper signal, that is, 8, 8, and 8 areas, the entire color space is 512 cubes. Divide into regions. On the other hand, in the address expansion mode, in order to increase the interpolation accuracy in the Y direction, as shown in FIG. 3B, the Y, Cr, and Cb axes are respectively divided areas designated by upper bits of 4, 3, and 3 bits, that is, As a result of division into 16, 8, and 8, the entire color space is divided into 1024 cubic regions. With respect to the number of grid points corresponding to the vertices of the divided cube, the total number of axes is 9,729,9 for the normal mode, and the total number of axes 9,9,9 for the address extension mode is 1,377.
When the image is in 8-bit input and the normal mode, the coordinates of the grid points are (0, 32, 64, 96, 128,
160, 192, 224, 256) The image is 8-bit input, and the (0, 16, 32, 4)
8, 64, 80, 96, 112, 128, 144, 16
0, 176, 192, 208, 224, 240, 25
6), the CrCb axis is the same as in the normal mode. On the other hand, when the image is in the normal mode with a 10-bit input, all axes are (0, 128, 256, 384, 512, 640, 76).
8, 896, 1024) 10-bit image input, and (0, 64, 128, 192,
256, 320, 384, 448, 512, 576, 6
40, 704, 768, 832, 896, 960, 10
24), the CrCb axis is the same as in the normal mode. All divided regions are defined as a section from a grid point to the next (grid point coordinate value -1).

Accordingly, since the maximum grid points such as the coordinate values 256 and 1024 are virtually present, the numerical value of 255 or 1023 becomes the maximum as an actual input signal entering the maximum divided area. In the present embodiment, the output values on the grid points are stored in the color conversion table, but the address input to the color conversion table is referred to not in the grid points but in each divided area. There are features.

More specifically, in the case of the normal mode in which the above-mentioned image is 8 bits, for example, the divided areas from 0 to 31 are the area numbers 0, 32 to 63 are the area numbers 1 and 2 at the end.
Area numbers 7 are from 40 to 256. That is, there are eight areas from number 0 to number 7. Since it is actually three-dimensional, the region numbers (0, 0, 0) to (7, 7, 7),
Up to 512 exist. The color conversion table memory is read by this address. There are two reasons for implementing this.

First, in this embodiment, since the interpolation operation is performed in parallel, when an input is included in a certain divided area, output values at six grid points are read out in parallel at an address indicating the divided area. This is because it is easier to input the same address to each of the six memories.

Second, the number of grid points is 9 or 17 on each axis.
Although the number is an odd number, the divided area number becomes a power of 2 of 8 or 16 which is one less than that, and the digital signal line of the address of the table memory can be used effectively.

As a result, a grid point at a virtual maximum value position necessary for making the volume of each divided region the same and performing accurate interpolation, for example, a grid point at a coordinate value 256 position in the case of 8-bit input, 10 In the case of bit input, the grid point at the position of the coordinate value 1024 is naturally specified as one of the six grid points when the divided area numbers 0 to 15 are specified in the normal mode and 0 to 15 in the address extension mode. Done in

On the other hand, conventional techniques such as those disclosed in
-226866 or Japanese Patent Laid-Open No. 22686/1990.
No. 8 discloses the same kind of interpolation technology,
Since the method uses a grid point to refer to the color conversion table memory, the number of grid points that cannot be a power of 2 must be treated as an address, and the number of area divisions must be changed in order to use address lines effectively. The number of grid points is changed to a power of two, and as a result, an interpolation error occurs while taking complicated measures.

At the same time that the upper signal 111 is input to the address generator 102, two signals (DCr, DCb) out of the three signals of the lower signal 112 are output from the triangular prism division selector 10.
3 is input. The lower signal 112 has three types of signals D
Y, DCr, DCb are all triangular prism-divided interpolation calculators 1
05 is input. A 1-bit Prsmsel signal, which is an output from the triangular prism division selecting unit 103, is output.
The msel signal is input to the address generation unit 102 and the triangular prism division type interpolation calculation unit 105.

With the above configuration, the upper signal is used as shown in FIG.
, A single cube or a rectangular parallelepiped (hereinafter collectively referred to as a unit cube) is selected.

As shown in FIG. 4, the lattice points constituting each unit cube are named a, b, c, d, e, f, g, h. In the triangular prism division selecting unit 103, the triangular prism abc-efg obtained by dividing the unit cube shown in FIG.
In consideration of DCr and DCb, when DCr> DCb, triangular prism abc-efg, Pr
smsel = 0 When DCr <DCb, triangular prism add-egg, Pr
Determination is made as in smsel = 1, and a Prsmsel signal is determined.

In the area processing unit 107, the internal area RA
A color conversion table switching signal 115 is internally generated based on the coordinate values on the image plane written in the M table and input to the color conversion table compulsory setting unit 108, and different colors are set for different regions on the screen. Corresponds to conversion. In the case where the color conversion table is externally switched for each pixel, the color conversion table switching signal 115 is directly input to the color conversion table compulsory setting unit 108 from the outside. The output signal of the color conversion table compulsory setting unit 108 is output to the address signal generation unit 102, and generates an address of the color conversion table memory 104 together with other information. In this embodiment, since the color conversion table memory 104 accesses six surfaces in parallel corresponding to the number of vertices of the triangular prism, the color conversion table memory 104 draws a diagram composed of six surfaces. Color conversion table memory corresponds to one type of color conversion. In the present embodiment, in the case of the normal mode, in order to perform two kinds of different color conversions, the set of six sheets is internally divided into two planes # 0 and # 1, and
Is selected and used by the output of.

In the case of the address extension mode, the color conversion table memory is used as one plane, and there is no switching of the color conversion table memory. Address generation unit 10
2, a specific address of the color conversion table memory 104 is generated based on the above signals,
9, the color conversion table memory 104 is accessed. On the other hand, in the triangular prism division type interpolation calculation unit 105, the output data value from the color conversion table memory 104 and the Prsmse
The l signal and the lower signal 112 are input and triangular prism interpolation is performed to generate an interpolation output value 113. In FIG. 1, since a color conversion device of a frame sequential format is premised, the color conversion output is composed of three inputs and one output in one color conversion interpolation operation.
Even in the case of three inputs and three outputs, such as when converting an rCb signal into a CIE-LAB signal, only one of the converted L, A, and B signals is generated. However, by mounting three or four types of color conversion table memory 104 and triangular prism division type interpolation calculation unit 105, it is possible to support three inputs and three outputs or three inputs and four outputs such as conversion from RGB to CMYK. It is. Also, rewriting of the color conversion table memory 104, various register units 1
Since the rewriting of 10 is performed from the host computer via the internal bus 114 via the host interface unit 106, the contents of the color conversion table memory are rewritten for every three or four outputs in one image configuration in a frame sequential manner. Can correspond to three inputs, three outputs, and four outputs.

Next, the operation of the triangular-prism-division-type interpolation calculation unit 105 will be described in detail. 8 of the specified unit cube shown in FIG.
Six color conversion table memories 10 among the output values at the vertices
The four output values read from 4 are Prs as shown in FIG.
When msel = 0, the output values are at six points a, b, c, e, f, and g. When Prsmsel = 1, a, d,
These are output values of six points, c, e, h, and g. Thus Pr
Although the grid point value output differs depending on the smsel signal, this is already specified by the address generation unit 102. The output values at the lattice points read from the six color conversion table memories 104 are expressed by adding () to the lattice points. For example, the output value at the lattice point a is (a). Using this notation, in FIG. 2, (a),
[(D), (c)], (c), (e), [(f),
(H)] and (g) are input to the difference generation unit 201. Here, [] indicates that any one is selected by Prsmsel. In the difference generation unit 201, five subtractors 2
13 to 217, five types of difference values are calculated. The difference value differs depending on the Prsmsel signal as shown in (Table 1).

[0039]

[Table 1]

The difference value selection section 202 selects four types of difference value signals generated based on the Prsmsel signal. Pr
When A input is selected when smsel = 0 and B input is selected when Prsmsel = 1, Pr is selected as shown in (Table 2).
A part of the difference value when the smsel signal = 1 is exchanged, and has a role of matching a combination when a product with the DCr and DCb signals serving as weighting factors is formed.

[0041]

[Table 2]

Here, the principle formula of the interpolation calculation will be described. In this interpolation method, the input color signal lower signals DY, DC
r and DCb are coordinates inside the cube representing the position of the input color in the Y, Cr, and Cb axis directions with the point a as a reference point in each unit cube. This is used as a weight coefficient used for interpolation. In FIG. 6, the output value (O) corresponding to the input color O existing in the triangular prism is DY, DCr, which is a component of the vector aO from the point a to the input point O to each of the Y, Cr, and Cb axes. , DCb, interpolation can be performed in three stages as follows.

In the first stage, a straight line is drawn through O and parallel to the Y axis, and the intersections of this straight line with the triangle abc and the triangle efg are defined as points n and m, respectively. Then, within the triangle abc, the output value (n) at the point n is

[0044]

(Equation 1)

The interpolation is performed as follows. This has the following graphical meaning: FIG. 7 is a view of the triangular prism of FIG. 6 observed from the Y-axis direction. From this direction, the two bottom surfaces of the triangular prism completely overlap, and O and n and m, a and e, b and f,
c and g each overlap. Therefore, this triangle may be considered as triangle abc. The input first difference vector ab (701) of the vector aO in FIG.
And the input second difference vector bc (7
02) is obtained, and an output first difference value ｛(b) − corresponding to each input difference vector in the output color space.
(A)} and the output second difference value {(c)-(b)} by DCr
And DCb to obtain the first and second output increments,
It will be added to the output value (a) at a.

In the second stage, assuming that FIG. 7 is a triangle efg, the output value (m) at m is given the same weight, and the output first difference value {(f)-(e)} and the output second difference value ｛( g)-
(F) Perform on｝,

[0047]

(Equation 2)

Is obtained. In the third stage, (n) and (m) are linearly interpolated on the line segment nm in FIG.

[0049]

(Equation 3)

Substituting (Equation 1) and (Equation 2) for (Equation 3) and rearranging,

[0051]

(Equation 4)

Is as follows. As described above, when the input point O is within the triangular prism abc-efg, the output value (O) at O is determined. When the input color point is present in the other triangular prism add-eg obtained by dividing the rectangular parallelepiped, the result is as shown in FIGS. Similarly, a straight line parallel to the Y axis is drawn from O to a triangle acd and a triangle egh to find intersections n and m, and each axis component of Y, Cr and Cb of the vector aO is D
Y, DCr, and DCb are obtained, and the output value at n in the first stage is

[0053]

(Equation 5)

In the second stage, the output value at m is

[0055]

(Equation 6)

In the third step, the output interpolation value at O is calculated according to (Equation 3).

[0057]

(Equation 7)

Is as follows. Based on the interpolation principle equations of (Equation 4) and (Equation 7), the multiplication unit 203 calculates each term of (Table 3) which is a part of the equation.

[0059]

[Table 3]

In the adder 204, the signal 223 of (Table 3)
224 and the addition of signals 225 and 226 to generate signals 227 and 228 (Table 4).

[0061]

[Table 4]

In the adder 205, the output value (a) and the signal 22
7 to generate a signal 229 (Table 5). This corresponds to the generation of the interpolation value (n) within the triangle corresponding to the lower base of the triangular prism in FIGS.

[0063]

[Table 5]

The difference between the signal 228 and the signal 227 is calculated by the subtractor 206.
To generate a signal 230.

[0065]

[Table 6]

The signal 230 is added to the signal 220 to generate the signal 231.

[0067]

[Table 7]

The signal 231 is obtained by multiplying the product with the DY signal by the multiplier 20.
8 and a signal 232 is generated.

[0069]

[Table 8]

The signals 229 and 232 are added by the adder 209, and the output 233 expressed by (Equation 4) and (Equation 7) is obtained.
Is calculated (Table 9).

[0071]

[Table 9]

The signal 232 is added to the positive or negative displacement amount written and stored in the displacement amount register 211 from the host by the adder 210, corrected for the displacement amount, and then inputted to the output data limit unit 212. After the dynamic range is limited, the interpolation output value 113 is obtained. Note that, for the sake of explanation, it is assumed that the weight coefficients of DY, DCr, and DCb all have a maximum value = 1, but in practice, this numerical value is 1
When a 0-bit image is input, the value is 128 in the normal mode and 64 in the DY in the address extension mode. The correction for this is all performed by the bit shift of the interpolation result. The multiplier used in the present invention can perform a multiplication operation of a signed number (data) and an unsigned number (weight coefficient).

Next, the configuration and data storage format of the color conversion table memory 104 in this embodiment will be described with reference to FIG.

The color conversion table memory 104 having a six-plane configuration shown in FIGS. 1 and 2 is a CRAM shown in FIG.
0, CRAM1, CRAM2, CRAM3, CRAM
4, CRAM5, and 6 color conversion RAMs, of which CRAM0, CRAM2, CRAM3, CRAM
5 stores the output values (a), (c), (e), and (g) at the vertices a, c, e, and g of the unit cube in FIG. 4, respectively, and has a capacity of 512 words. x2 = 1024 (word)
It is. CRAM1 and CRAM4 have a capacity of 2048 (word) twice as large as the other four planes, and output two banks (b) and (d), which are outputs at b and d, respectively, and outputs at f and h ( Two banks f) and (h) are stored. CRA
The output value banks of M1 and CRAM4 are switched by the Prsmsel signal.

In the case of the normal mode shown in FIG. 14A, two types of tables, different color conversion # 0 and color conversion # 1, can be stored in the CRAM, and the tables are switched by the LUTNO signal. Outputs in the case of color conversion # 1 are indicated by dotted lines as (a) ′ to (h) ′.

On the other hand, in the address expansion mode shown in FIG. 14B, two memories for the color conversion # 1 in the normal mode are stored.
Used for doubled grid points, Prsmse
LUTNO signal besides the bank switching by 1 signal
It plays the role of the Y6 signal, which is the sixth bit of the signal. Therefore, in the address extension mode, only one type of color conversion # 0 is stored.

FIG. 15 shows a main color conversion table memory 104.
Is an external address configuration which is an address when the address is viewed from the host side. In FIG. 14, the memory assigned (0) to (15) is arranged in order. In the normal mode, (0) to (7) correspond to the color conversion # 0 table 0,
(8) to (15) correspond to Table 1 of color conversion # 1.

On the other hand, in the address extension mode, all the tables from (0) to (15) are tables for performing the color conversion # 0.

FIG. 16 shows the relationship between the external address and the color conversion RAM address. In this embodiment, the host stores the 14-bit external address ADR
0 is accessed as ADR13. Color conversion RAM
Since the bit width is 10 words and 1 word, in a byte configuration address, bit 0 is not used because one output value data is stored in two addresses of lower byte when bit 0 is 0 and upper byte when bit 0 is 0 .

In the case of the normal mode shown in FIG. 16A, the LUTNO signal N and the bits 12, 11, 10
Is the color conversion RAM number nram, bits 9, 8, and 7 are Cb
Upper 3 bit signals (Cb9, Cb8, Cb7) of the signal,
Bits 6, 5 and 4 contain the upper 3 bit signals (Cr
9, Cr8, Cr7) and bits 3, 2, and 1 are assigned the upper 3 bit signals (Y9, Y8, Y7) of the Y signal. The Prsmsel signal is embedded in bit 10 as part of the color conversion RAM number. The memory capacity represented by bits 0 to 9 corresponds to 512 (Word). On the other hand, the actual color conversion RAM address has an address configuration based on a 10-bit configuration, and the CRA
In M1 and CRAM4, 512 (Word) addresses occupy bit 0 to bit 8, and bit 9 indicates LUTN.
O, bit 10 corresponds to Prsmsel, so total 20
48 (Word) address space. CRAM0,
Prsmse for CRAM2, CRAM3 and CRAM5
Since l does not exist, 102 from bit 0 to bit 9
8 (Word) address space.

On the other hand, in the address extension mode shown in FIG.
Other than that, the role of bit 9 of the AM address is changed from LUTNO to Y6, and the rest is the same as the normal mode.

Next, the bit width set in the color conversion table memory 104 will be described with reference to FIG. The color conversion table memory 104 has a configuration capable of storing up to 10 bits as the maximum capacity, but has a storage data bit width of 10 bits.
Bit and 8-bit can be selected. This selection can be set irrespective of whether the color signal input / output value of the color image is 10 bits or 8 bits. This is because all interpolation calculations of color signals are performed in a 10-bit format, and when the image color signal is 8 bits, only the upper 8 bits need to be considered. That is, the following four combinations can be selected as combinations of the number of input / output bits of the image color signal and the bit width of the color conversion table memory data.

[0083]

[Table 10]

As can be seen from the case of FIG. 17A, when data of 10-bit width is stored, the upper byte B9,
B8 is an external address and is stored at an odd address. For this reason, the host must perform 16-bit transfer or 8-bit transfer twice in succession, increasing the data transfer amount.

On the other hand, in FIG. 17B, since 8-bit data is stored only in the even address of the external address, it is sufficient to perform 8-bit transfer only for the even address, and the transfer amount is reduced. The 8-bit width storage is effective for applications such as interactive color correction where the rewriting time of the color conversion table memory is to be reduced as much as possible and the data transfer amount is to be reduced as much as possible.

However, the configuration of the output data value at the grid point used for the interpolation operation is 10 bits with a sign.
Therefore, a value stored with an 8-bit width always includes an error. As a result, numerical values given on grid points when color space is interpolated are restricted. For example, in case 1, for an 8-bit input / output of an image color signal, signed 8-bit data having an 8-bit width, that is, 7-bit data is substantially stored, and data on a grid point is an even number from 0 to 254. It is unavoidable that the interpolation accuracy will deteriorate.
Table 11 shows the trade-off between the table transfer amount and the values that can be taken at the lattice points in cases 1 to 4.

[0087]

[Table 11]

Next, a method for actually creating a signed color conversion table memory storage data will be described. In this embodiment, when the numerical values are stored in the color conversion table memory 104, 1201 in FIG. 12A and 1301 in FIG.
In order to realize the clipping characteristics and the through characteristics as shown in FIG. In this case, assuming that 8-bit color image processing is usually assumed, data (0 to 25) is stored in the color conversion table memory 104.
In addition to 5), in order to store a numerical value of (-1) or less, 9-bit data must be stored because it is necessary to represent a negative number in two's complement. This eliminates the aforementioned advantages, especially when stored in 8-bit width. In order to solve this problem, the present embodiment reduces the precision of each numerical value,
The storage width data extends the original storage range while keeping the same.
Hereinafter, description will be made based on embodiments.

First, the input / output signal of an image is 10 bits and the storage bit width of the color signal / color conversion table memory 104 is 1
"Case 4" in (Table 10) of using a 0-bit configuration is taken as an example.

Since the storage bit width is 10 bits, considering the sign, only a value of 9-bit numerical value (−512 to 511) + sign can be set. The grid point data to be actually output to the color conversion device is (−3072 to 307).
In the range up to 2), the value of the range included in the window having an arbitrary width of 1024 is displaced to an 11-bit expression range of (−1024 to 1023), and then halved to reduce the accuracy by half (−). The values are set in the range of 512 to 511) and set in the color conversion table memory. If the original grid point data is KO and the displacement is OFFSET, the set value is

[0091]

(Equation 8)

Is obtained. This is shown in FIG. KO
The axes represent expressible grid point data (from -3072 to 307)
2) is displayed. This range is the displacement amount OFFSET
Is in the range of the above (Equation 8). The range that can actually be stored is a width 1024 indicated by a rectangle of 1801.
Range. Just the middle position 0 of this range is 180
It is expressed by the black circle of 2. Black circles in other rectangles indicate even numbers, white circles indicate odd numbers, and indicate that only even numbers or only odd numbers are actually stored in the corresponding range. The upper and lower limits of the rectangle are shown in parentheses. Further, the cross hatched portion in the rectangle shaded with oblique lines indicates a numerical range 0 to 1023 that can be finally output in the output data limit unit in the present invention. The rectangle indicated by 1803 is the maximum OFFSET value of 20
47 shows a case where the largest positive and largest numerical value area is set. Odd-numbered grid point data from (1024 to 3072) can be stored. This is realized by subtracting the displacement amount OFFSET = 2047 expressed by 1804, returning to the position of 1801, and then multiplying by 1/2 to obtain a set value 1810 having a range from (−512 to 511). . Needless to say, even if such a large numerical value is stored, the output data is finally limited and has little meaning. Similarly 18
The rectangle indicated by 08 is the minimum value of OFFSET = −2048
Is a case where the smallest negative numerical value area is set using. Even numbers from (-3072 to -1026) can be stored. This is also all 0 because output data limit is applied
It is meaningless. The actual OFFSET is used not in such an extreme value but in the range between them.
Minimum value K of grid points that can be stored when SET is determined
Omin and the maximum value KOmax are calculated by the following equations.

[0093]

(Equation 9)

Further, reference numeral 1805 (from -512 to 153)
When storing up to 4), the displacement amount OFFSET expressed by 1806 is subtracted, returned to the position of 1801, and then similarly halved to set value 1810. Of course, OFFSET may be set to 0 and used in a state like 1807. The displacement amount OFFSET used when setting data in the color conversion table memory is set by changing the sign in the displacement amount register 211. That is, when the displacement amount is subtracted, a positive value is set, and when the displacement amount is added, a negative value is set. As a result, at the time of color conversion, the color conversion table memory set value is doubled by hardware and used for interpolation calculation, and finally returns to the original grid point data value KO even if the displacement amount is adjusted.

Next, "Case 1" in (Table 10) in which the input / output signal of an image is 8 bits and the storage bit width of the color signal / color conversion table memory 104 is 8 bits is used as an example. In this case, the color conversion table memory 10
4 can only be set to a value from (-128 to 127). Therefore, the grid point data KO to be stored is set to a value in a range having a width 512 determined by the upper limit and the lower limit (from -256 to 25
The 9-bit range 1903 including the sign of 5) is corrected by the displacement correction means, and then halved to reduce the accuracy to half and stored in the range (-128 to 127) and stored in the color conversion table memory. Set. The setting value is

[0096]

(Equation 10)

When the OFFSET is determined, the upper and lower limits of the storable numerical values are as follows.

[0098]

[Equation 11]

Is determined by This case is shown in FIG. Next, how to use the displacement will be described. "Case 1" in which both the image color signal and the storage bit width are 8 bits is taken as an example. In the state of displacement = 0 and 1901 (-256 to 254)
Although it has been described that this can be expressed, the accuracy is actually reduced when the numerical value is compressed to 1/2, so that only an even number in the range (-256 to 254) can be expressed as shown in 1901 in FIG. Therefore, a through characteristic requiring a value of 256 cannot be realized. In order to realize the through characteristic, the displacement amount = 2 is required at least to represent an even number of (−254 to 256). The fact that only even values can be expressed does not hinder the realization of the through characteristic. This is because the lattice point position in the input space is an even number in the first place, and the lattice point storage value can take only an even number in the case of through. In order to secure a dynamic range that can cope with various color conversions including the through characteristic, a positive margin range exceeding the maximum value of the output data limit portion = 255 and a negative margin range less than the minimum value of the output data limit portion = 0. There is a way of thinking that is set equal. This corresponds to setting the displacement amount to 128 as indicated by 1902, and in the case of FIG. 18 corresponds to the state of the displacement amount = 512 indicated by 1805.
The results of similar considerations for Case 2 and Case 3 in (Table 10) are summarized in (Table 12). This is (Table 11)
It can be said that OFFSET is added to the result.

[0100]

[Table 12]

Next, the area processing unit 107 will be described in detail. As shown in FIG. 20, the area processing in the present embodiment is such that when performing image color conversion for one screen, a plurality of rectangular areas are set on the screen and different color conversions are performed inside and outside the area. is there. In the present embodiment, two types of color conversions # 0 and # 1 can be stored in the color conversion table memory 104 in the normal mode in which address expansion is not performed. Therefore, the color conversion table memory 104 is switched according to an area RAM table set in advance so that color conversion # 0 is used outside the area and # 1 is used internally.
Each area has a start pixel address SN and an end pixel address EN on the main scanning line. Here, N indicates 1 to 8. FIG. 21 shows a detailed circuit configuration of the area processing circuit 107. As can be seen from the figure, the area processing circuit 107 includes an area RAM table 2100 and an area counter 210.
1 and comparator 2102, 13-bit pixel counter 210
3 The pixel counter 2103 counts the pixels on the main scanning line, and the comparator 2102 compares the start pixel address and the end pixel address stored in the area RAM table 2100 as shown in FIG. On the other hand, the area counter 2101 counts the current area number from 0 to F (meaning 15 in hexadecimal notation) and inputs it to the area RAM table 2100 to output the next pixel address. Of the 14 bits of the output value of the area RAM table 2100, one bit is the color conversion table N of FIG.
O indicates 0 and 1 indicates a LUTNO signal 115 for switching the color conversion RAM between # 0 and # 1. The area RAM table 2100 is rewritten by the host via the internal bus 114. By this software rewriting, area processing in the sub-scanning direction is performed.

As described above, in the area processing circuit 107, processing in the main scanning direction requiring high speed is performed by hardware, and processing in the sub-scanning direction is performed by software.

The color conversion table can be forcibly switched for each pixel by an external color conversion table switching signal 115 even in an area other than a preset area. In this case, since the change unit is a pixel, there is an advantage that color conversion processing can be switched at an arbitrary position even in a shape other than a rectangle.

Next, the host interface unit 106 will be described. The role of the host interface unit 106 is to allow the host to write into the color conversion table memory 104 only during the blanking period of the image / video signal, thereby eliminating image / video disturbance at the time of writing.
Thus, for example, various color changes can be tried one after another while the still original image is displayed on the display, and a subtle difference in atmosphere from the original image can be confirmed.

More specifically, as shown in the timing chart of FIG. 23, the input to the color conversion RAM address is a color conversion RAM address determined from the pixel value during the scanning of the image and the blanking period (see FIG. 23 (b)) performs switching control so as to become an external address. The host monitors the NBUSY signal (FIG. 23 (f)) output from the color conversion apparatus of this embodiment, and starts writing to the color conversion RAM from the time when NBUSY becomes “H”, and NBU.
When SY = "L", the writing is interrupted (FIG. 23).
(G)).

Since the writing is repeated by repeating this cycle, no influence is exerted on the image / video display. This writing procedure is the same when using DMA.

Hereinafter, a second embodiment of the present invention will be described. FIG. 24 is a block diagram of a color conversion apparatus according to the second embodiment of the present invention.

FIG. 24 differs from the first embodiment of the present invention in FIG. 1 in that a color difference conversion unit 116 is newly provided.

Hereinafter, the color difference conversion section 116
This will be described in relation to the N operation. As shown in FIG. 25, the color difference conversion unit 116 generates lightness and a special simple color difference from the three input signals. Assuming now that the input is three signals of R, G and B, the color difference signals Cr and Cb compensate for the negative value generated after the subtraction by the subtractor 2401 and are 8 bits or 1 bits used in the interpolation unit.
Negative value compensating means 2402 for converting to 0 bit value

[0110]

(Equation 12)

Is generated. In this case, a difference from the other two signals R and B is generated using the G signal as the color difference reference value, but R or B may be used as the color difference reference value. The Y signal can be created from R, G, and B by various methods, and may be any signal that is independent of the Cr and Cb signals. Specifically, the brightness generation means 2403

[0112]

(Equation 13)

The lightness Y is generated by such an equation. Next, the specificity of the MIN operation will be described. An operation to take the minimum value from three input signals R, G, and B

[0114]

[Equation 14]

The following is an example. In order to interpret this operation graphically, an RGB cube is divided into three pyramid-shaped areas (0), (1), and (2) as shown in FIG.

[0116]

(Equation 15)

Since the area can be defined as

[0118]

(Equation 16)

Is obtained. Therefore, interpolation is easy because the plane having a constant output value in the MIN operation is a parallel plane inside each divided area as shown in FIG. On the other hand, at the boundary surface 2501 of the three divided areas, this constant surface changes rapidly, and when performing interpolation, a large discontinuity occurs in the lattice point value, so that interpolation becomes extremely difficult and as a result, a ripple-shaped interpolation error occurs. It was. This boundary surface is defined as W of RGB solid
When viewed from the (white) direction, it is as shown in FIG. 26 (B). The directions of the hues R, G, and B correspond to each other.

[0120]

[Equation 17]

Is obtained. In the conventional interpolation method using a tetrahedron, since the discontinuous surface naturally coincides with the tetrahedron division surface, the constant surface does not change within one interpolation section, so that linear interpolation can be performed without ripples. Had been avoided.

In this embodiment, the problem is avoided by the same division effect. That is, the state of division of the triangular prism on the color difference plane after color difference conversion is as shown in FIG.
= (RG) axis, Cb = (BG) axis, and Cr = C
It is divided on the b-axis. Here, when the boundary surface in FIG. 26B is mapped to the color difference plane, the boundary surface is as shown in FIG. 27B. However, the boundary surface of the three regions shown in FIG.
Are included.

For this reason, the MIN calculation can be interpolated without ripple by the color difference conversion of the present embodiment.

Hereinafter, a third embodiment will be described. FIG. 28 is a block diagram of a color conversion apparatus according to the third embodiment of the present invention.

FIG. 28 is different from the second embodiment shown in FIG.
2 is newly provided, and a selector 121, a selector 122, and a weight control unit 123 are provided.

Hereinafter, the address generation of the third embodiment will be specifically described. FIG. 29 shows the detailed configuration of the address generation unit 120. Note that the address generation unit 120 includes fixed amount subtraction means 120A and adders 120B and 120C. In the above configuration, the input signals (Y, Cr, Cb) are as follows.

[0127]

(Equation 18)

Each of the input color signals R, G, and B has an 8-bit configuration.
Bit. Creating a color difference signal increases the dynamic range by one bit.

[0129]

[Equation 19]

Changes up to Here, the color difference signal is compensated for a negative value by the negative value compensating means 2402 described above.

[0131]

(Equation 20)

Therefore, YH, CrH, CbH are 3,
4 or 4 bit signal and dynamic range of color difference is 2
Increase by a factor of two. This causes an increase in the amount of unused memory as described later. Address generation unit 12
At 0, the fixed amount subtracting means 120A for subtracting the numerical value 7 from these signals and the adding means 120B and 120C are used.

[0133]

(Equation 21)

Is calculated. Here, since the original RGB values are all positive, CrH ′ and CbH ′ do not become negative but take the following positive range and reduce the dynamic range by half. With this effect, in the present embodiment, a large increase in memory use efficiency can be obtained.

[0135]

(Equation 22)

Next, the graphical meaning and effect of this calculation will be described. First, a case where the RGB input color space is considered as three orthogonal axes will be described. At this time, the RGB input color space becomes a cube, and the YCrCb space formed by (Equation 18) becomes an oblique coordinate system.

The CrCb space is a parallelepiped formed by shifting the upper and lower bases formed by the B and R axes of the RGB cube as shown in FIG. That is, the Y axis indicates the diagonal direction of the RGB cube, the Cr axis is the same direction as the R axis, and the Cb axis is the same direction as the B axis.

In FIG. 30, Cb is set as indicated by the dotted arrow.
FIG. 31 shows a two-dimensional representation when observed from the axial direction. YCrCb space is RG including color difference positive and negative.
It is represented as a parallelogram that completely encompasses the B square. Here, points represented by black squares represent grid points in the RGB cube.

This lattice point coincides with the lattice point in the YCrCb oblique coordinate system. However, if it is attempted to completely include the RGB cube, (RG) is changed from -8 to 7;
H must be used from 0 to 15 and YCrCb
Points outside the RGB space that are not accessed while in the space occupy a large portion, and waste of memory occurs. This memory use efficiency is only about 25%. Therefore, in this embodiment, the inside of the RGB cube and a few additional memories are prepared, and the grid point coordinates input as the YCrCb oblique coordinates are converted so as to access the coordinates of the RGB orthogonal memory. .

First, how to select points necessary for interpolation will be described. The vertices of the unit triangular prism in the lightness color difference space for interpolating the input color point 3101 in the RGB space are P1, P2, P3, and P4 in this two-dimensional description. Where P1 is R
This is a virtual grid point outside the GB cube.
A base point of a grid point in the orthogonal system, and a unit interpolation section number (YH ', CrH') of the RGB orthogonal system represented by this base point
Is calculated. The input color point is

Since YH = 2 and CrH = 5 (Equation 2)
According to 1)

[0142]

(Equation 23)

In the RGB orthogonal system, (YH ′, Cr
H ′) = P1, P which are unit interpolation sections indicating (2, 0)
2, P4 and P5 are designated. Since the unit interpolation section may be considered to indicate the base point, in this case P
1 is indicated. If such an address translation mechanism is used, extra memory must be prepared for one section at both ends in the RGB space such as the virtual points P1 and P5, but there is a case where memory is prepared over the entire YCrCb space. You can save a lot of memory.

The above description will be similarly described in a two-dimensional manner using a method of considering the YCrCb space as an orthogonal system. In FIG. 32, the square surrounding the whole is Y-Cr in the YCrCb space.
A parallelogram, which is a projection plane and is filled with grid points indicated by black square points, is an RG projection plane of an RGB cube. In this expression, four points necessary for the triangular prism for interpolating the input color are rectangles P1, P2, P parallel to the YH axis and the CrH axis.
3, P4, and P1 is a virtual grid point required outside the RGB space. On the other hand, the above-described address conversion is performed, and the parallelograms P1, P2, P in the RGB oblique coordinate system are converted.
4, P5 and its base point P1 are designated.

The output from the address generator 120 is YH'C
rH'CbH 'YH' (3 bits) in orthogonal system, CrH '
(4 bits) and CbH '(4 bits) in the address generator 1
02 is output.

Each of CrH ', CbH', and YH 'indicates the position on each axis of the unit interpolation section in the YH'CbH'CbH' space shown in FIG.

The address generator 102 is the address generator 1
An address of the color conversion table memory required for the interpolation calculation is generated from the unit interpolation section number indicated by 20. In FIG. 33, the position of the color conversion table memory required when the unit interpolation section number indicates the position of P is indicated by a thick black line. In this figure, M0 to M7 indicate eight types of color conversion table data, and (i, j, k) indicates the type of the unit cube. As can be understood from FIG. 33, when the unit interpolation section number is designated, the upper surface of the required unit cube is
It is at a position shifted by one each in the CrH and CbH directions with respect to the lower surface. This is because the address generator 120 adds YH when CrH 'and CbH' are created from CrH and CbH, so that the lattice point positions corresponding to CrH and CbH at the position of YH + 1 are CrH '+ 1 and CbH' + 1. Because it became. Since this positional relationship always holds, it is necessary to take out the position of the upper surface of the unit cube from the unit cube advanced by one each in the CrH'CbH 'direction with respect to the bottom surface.

Eight address generation units 102 are provided, each corresponding to eight color conversion table memories (M0 to M7) 104. The color conversion table has eight grid point data for only the unit cube having an even number of three axes. Since all of the input grid points can be filled with these grid point data without excess or shortage, the color conversion table memory has a necessary and sufficient memory amount.

How the color conversion table memory is accessed from the unit interpolation section number will be described with reference to FIGS. 34 and 35.

FIG. 34 shows the color conversion table memory arranged in the YH'CrH'CbH 'space as viewed from the CbH' direction, and FIG. 35 shows the color conversion table memory as viewed from the YH 'direction. In FIG. 34, A indicates the position of the table data used when the unit interpolation section number (YH ') is an even number, and B indicates the odd number. When YH 'is an even number (A), the memories M0 to M3 serve as the bottom surface of the interpolated solid and M
4-M7 is the upper surface. When YH 'is an odd number (B)
, The memories M4 to M7 are on the bottom, and the memories M0 to M3 are on the top. In FIG. 35, the unit interpolation section signal (YH'CrH'Cb
When the position A of the bottom surface of the interpolated solid is designated by H ′), the unit interpolation section number on the top surface is point A ′ advanced by one in the CrH′CbH ′ direction.

FIG. 36 shows a block configuration of the address generation unit 102. The address generation unit 102 has M0
M7 is provided independently of each other, and reference numeral 3601 denotes a block number generator for generating a selected block number for extracting memory data from the unit interpolation section number. FIG.
4. From the description of FIG. 35, when YH 'is an even number, M0 to M3 are at the bottom and M4 to M7 are at the top, so that the block number generation unit 3601 performs unit interpolation when YH' is an even number for M0 to M3. The section number is output as it is as the selected block number, and when YH 'is an odd number, a value obtained by adding 1 to each of the unit interpolation section numbers of CrH' and CbH 'is output as the selected block number. For M4 to M7, the operation is the reverse of the above. If YH 'is an even number, a value obtained by adding 1 to each of the unit interpolation section numbers of CrH' and CbH 'is output as a selected block number, and YH' is an odd number. In the case of, the unit interpolation section number is directly output as the selected block number.

When the selected block number is odd for each axis, the selected block number section 3603 uses the selected block number to obtain grid point data from the next even block number. Perform addition. This operation is performed for eight types of unit cubes (i, j, k).
And M0 to M7 change the presence or absence of addition.
(Table 13) shows the selected block number adjustment signals (UX UY YZ) for M0 to M7. A value of 0 indicates a case where the block address is incremented by 1 for an odd-numbered block. FIG. 37 is a block diagram of the block number even-numbering section 360 of FIG.
2 and 2 shows a specific configuration of the block number adding unit 3603, and is an example of logic for converting blocks into even numbers.

[0153]

[Table 13]

After the addition of the selected block numbers, all the numbers become even numbers, and the color conversion table data of each grid point of only the even number blocks held in M0 to M7 can be extracted.

Further, in FIG. 36, the multiplier 3604 multiplies the output of the CrH 'block number adder and the output of the CbH' block number adder by M, and the multiplier 3605 multiplies the output of the Z block adder by M ^2. The sum of the results is taken by an address adder 3606 to output a linear address of a color conversion table memory.

[0156]

(Equation 24)

[0157]

(Equation 25)

N is CrH 'of the address generator 120 and C
This is the input range of bH '. N is incremented by 1 in each of the block number generation and the block number evening processing, and after being halved, is used as the address (MAi) of the color conversion table memory.

By using such an address generation unit 102, each axis can be converted to a continuous linear address even if the number of grid points is not a power of two, thereby eliminating waste generated from discontinuity of addresses in the color conversion table. it can.

The eight parallel addresses from the address generation unit 102 are led to eight color conversion table memories 104 via the memory interface 109, respectively. Grid point output values M0-M read from color conversion table memory
3, M4 to M7 are selector 121 and selector 1 respectively.
22, the input positions a,
The grid point output values corresponding to b (d), c, e, f (h), and g are output.

FIG. 38 illustrates the types of memory output to the upper and lower surfaces when the unit interpolation section number is (100). For example, the output at the position a on the bottom surface is M1, the output at the position b is M0, the output at the position c is M3, and the output at the position d is M
It is understood that 2 should be output.

Table 14 shows the types of unit cubes (i, j,
This table summarizes the color conversion tables to be selected by the selector 121 and the selector 122 according to k).

[0163]

[Table 14]

The lower 5 bits excluding the upper bits of the output from the pixel input section 101 are subjected to triangular prism determination by the triangular prism division selecting section 103 as in the first embodiment, and the triangular prism interpolation calculating section 105 and the selector 121 are selected. , 122. The weight control unit 123 determines the least significant bit of the YH 'signal, and if the number is an odd number, sets the weight coefficient DY in the Y-axis direction to (1-DY). This operation is obtained by adding 1 after bit inversion. The weights DCr and DCb in the Cr and Cb directions are 3 as they are.
It is input to the prism interpolation calculation unit 105. The reason why DY is set to (1-DY) by the least significant bit of YH 'is that when YH' is an odd number, triangular prism interpolation in the Y-axis direction is performed from the upper surface as shown in FIG. This means that when YH 'is an odd number, a to d are on the upper surface, but the interpolation coefficient DY is set to (1−D
By adopting Y), there is no need to replace the upper surface and the lower surface, and there is an advantage that the selectors 121 and 122 can be input with four inputs.

Output 113 from triangular prism interpolation calculation unit 105
Can obtain the same output as that of the second embodiment and can effectively use the memory without causing the generation of the unused memory due to the color difference conversion of the input signal.

In the color conversion table memory, only grid point data of even-numbered blocks alone is provided. A dedicated address generator is provided independently in each memory, and the number of grid points of each axis of the three-dimensional color conversion table is not a power of two. In such a case, a linear address can be generated, the memory address can be used effectively, and as a result, the color conversion table can be designed with a small memory capacity.

FIG. 40 shows that the color conversion table memory 104
The same type is used for the bottom (A) and the top (B) of the triangular prism that is completely filled with the data of the types (black circle, black triangle, black square, and white circle, white triangle, and white square) and that is necessary when performing the triangle prism interpolation calculation. It shows that it is possible to design so that no data is used. The six types of hatched patterns represent possible types of triangular prisms.

FIG. 41 shows that it is possible to make the same pattern of the triangular prisms. In other words, all the grid points can be filled up with the shaded triangle endpoints, and at the same time, the shapes are all triangles of the same shape with the lower left being a “black circle”. In this example, the turtle-shaped hexagon is the space division unit, and the data of the end point of the lowest middle triangle among them may be held. 41A shows a bottom surface and FIG. 41B shows a top surface, as in FIG.

As described above, the embodiment shown in FIGS.
The present invention can also be realized by providing an address conversion mechanism for a three-dimensional color conversion table at a stage preceding the table.

FIG. 42 shows another embodiment in which the bottom surface of the memory is two-dimensionally addressed in the six memory configurations shown in FIG. Reference numerals 4101, 4102, and 4103 denote address lines indicating the address order of the three types of memories, and three independent types of memories on the address lines can access the triangle on the top or bottom of the target triangular prism. The address method shown in FIG. 40 can be realized by modifying the address generation unit 102 described in the present invention.

[0171]

As described above, according to the present invention, first, the interpolation in the lightness direction and the color difference direction are separated by using triangular prism type interpolation, and the two-dimensional interpolation of the chromaticity plane is most performed by using triangles. Because of the simplification, there is an advantage that, when performing color adjustment in the chromaticity plane, it is simpler than the conventional interpolation method, and there is an advantage that the hardware configuration is not so complicated.

Second, the color conversion characteristics, such as the so-called “clipping characteristics” and “complete through characteristics”, which were difficult to realize by the conventional interpolation calculation, have reduced accuracy in storing numerical values in the color conversion table memory. This can be realized by expanding the storable range and correcting the amount of displacement.

Third, when the present invention is used for color conversion of a color image on a display as in the case of application to a color corrector, the color conversion table memory is frequently accessed by the bus arbitration function of the host interface unit. Also, the image is not disturbed, and it is very easy to subjectively compare the original image with various color adjustments and color conversions.

Fourth, not only one color conversion is performed on the entire screen, but also different color conversions are performed by switching a color conversion table used in a specified area on the screen or for each specified pixel in real time. Can be. As described above, the color conversion device of the present invention has excellent performance as a color adjustment device and a color conversion device for a color image display device.

Fifth, even in the case of performing special color conversion such as MIN operation of three input signals, color difference conversion and triangular prism division work effectively, so that no ripple-like error occurs during interpolation.

Fifth, if the memory-address method by the first and second address generation units of the present invention is used, duplication of the color conversion table is avoided, and the unused portion of the color conversion table memory is eliminated, and the effective memory is used. Because it can be used,
There is an effect of designing a large color conversion table with a small memory capacity.

Sixth, the number of selector inputs of the color conversion table memory output can be halved by switching the interpolation between the upper surface and the lower surface by triangular prism interpolation by the method shown in the present invention. This has the effect of simplifying the structure of the part.

[Brief description of the drawings]

FIG. 1 is a block diagram of a color conversion apparatus according to a first embodiment of the present invention.

FIG. 2 is a block diagram of a triangular-prism-division-type interpolation operation unit, which is a main part of the same color conversion apparatus.

FIG. 3A is a conceptual diagram of cubic division of a YCrCb input color space of the same color conversion device in a normal mode. FIG. 3B is a conceptual diagram of cuboid division of an YCrCb input color space of the same color conversion device in an address extension mode.

FIG. 4 is a conceptual diagram of each unit cube in an input color space in the same color conversion device.

FIG. 5 is a conceptual diagram of dividing each unit cube into two triangular prisms in the same color conversion device.

FIG. 6 is a conceptual diagram showing an input point O and an interpolation weight coefficient in a triangular prism abc-efg in the same color conversion apparatus.

FIG. 7 is a view of the triangular prism abc-efg in the same color conversion device as viewed from the positive direction of the Y axis.

FIG. 8 is a conceptual diagram showing an input point O and an interpolation weight coefficient in a triangular prism add-egg in the same color conversion apparatus.

FIG. 9 is a view of a triangular prism acd-egh in the same color conversion apparatus observed from the positive direction of the Y axis.

FIG. 10 is a conceptual diagram showing an interpolation result of a lightness gradation in conversion from Lab to rgb according to the related art.

FIG. 11 shows Lab to rg according to the first embodiment of the present invention.
A conceptual diagram showing a result of interpolation of a brightness gradation in the conversion to b.

FIG. 12 is a conceptual diagram showing an interpolation result of “clipping characteristics” using the first embodiment of the present invention.

FIG. 13 is a conceptual diagram showing an interpolation result of “through characteristics” using the first embodiment of the present invention.

14A is a diagram showing a color conversion table memory configuration in a normal mode according to the first embodiment of the present invention. FIG. 14B is a diagram showing a color conversion table memory configuration in an address extension mode according to the first embodiment of the present invention. Figure

15A is a diagram showing an external address format table of a color conversion table memory in a normal mode according to the first embodiment of the present invention. FIG. 15B is a diagram showing an external address format table in an address extension mode according to the first embodiment of the present invention. The figure which showed the external address format table of the color conversion table memory

16A is a diagram showing a relationship between an external address and a color conversion RAM address in a normal mode according to the first embodiment of the present invention. FIG. 16B is a diagram showing a relationship in an address extension mode according to the first embodiment of the present invention. Diagram showing the relationship between external addresses and color conversion RAM addresses

FIG. 17A is a diagram showing a bit width configuration of a color conversion table memory when the storage bit width of the first embodiment of the present invention is 10 bits. FIG. 17B is a diagram showing the storage bit width of the first embodiment of the present invention. The figure which shows the bit width configuration of the color conversion table memory when = 8 bits

FIG. 18 is a diagram illustrating a relationship among a displacement amount, an output value on a lattice point, and a color conversion table memory set value according to the first embodiment of this invention.

FIG. 19 is a diagram illustrating a relationship between a displacement amount, an output value on a grid point, and a color conversion table memory set value according to the first embodiment of this invention.

FIG. 20 is a diagram showing a state of area processing according to the first embodiment of the present invention.

FIG. 21 is a circuit diagram showing an area processing unit which is a main part of the color conversion apparatus according to the first embodiment of the present invention.

FIG. 22 is a diagram showing an area RAM table of the area processing unit.

FIG. 23 is a DM of the host interface unit of the color conversion apparatus.
Timing chart at A

FIG. 24 is a block diagram of a color conversion device according to a second embodiment of the present invention.

FIG. 25 is a block diagram of a color difference conversion unit which is a main part of the color conversion device.

FIG. 26 (A) is a diagram showing a constant value surface of a MIN operation and its discontinuous boundary surface inside an RGB cube in the same color conversion device. (B) MI inside an RGB cube in the same color conversion device.
Diagram showing a constant value surface of N operation and its discontinuous boundary surface

FIG. 27 (A) (RG) in the same color conversion apparatus
(BG) Conceptual diagram in which the inside of the color difference plane is divided into triangular prisms. (B) Conceptual diagram showing a discontinuous surface of a constant value of the MIN operation in the (RG) (BG) color difference plane in the same color conversion device.

FIG. 28 is a block diagram of a color conversion apparatus according to a third embodiment of the present invention.

FIG. 29 is a block diagram of an address generation unit which is a main part of the color conversion apparatus.

FIG. 30 is a diagram showing a three-dimensional relationship between a cube in the RGB orthogonal coordinate system and a parallelepiped in the YCrCb oblique coordinate system of the same color conversion apparatus.

FIG. 31 is a diagram showing a two-dimensional relationship between a cube in the RGB orthogonal coordinate system and a parallelepiped in the YCrCb oblique coordinate system of the same color conversion apparatus.

FIG. 32 is a diagram showing a two-dimensional relationship between a cube in a YCrCb rectangular coordinate system and a parallelepiped in an RGB oblique coordinate system of the same color conversion apparatus.

FIG. 33 is a conceptual diagram in which a memory space of a color conversion table corresponding to a unit cube of a color difference space of the same color conversion apparatus can be corresponded.

FIG. 34 is a conceptual diagram of a color conversion table memory arranged in a YH′CrH′CbH ′ space of the same color conversion apparatus as viewed from the CbH ′ direction.

FIG. 35 is a conceptual diagram of a color conversion table memory arranged in a YH′CrH′CbH ′ space of the same color conversion device as viewed from the YH ′ direction.

FIG. 36 is a block diagram of an address generation unit, which is a main part of the color conversion apparatus.

FIG. 37 is a block connection diagram of a block even number processing unit in the address generation unit.

FIG. 38 is a diagram showing a relationship between memories to be selected at a unit cube position of the same color conversion apparatus.

FIG. 39 is a diagram showing that YH ′ changes the interpolation direction in the YH axis direction by odd and even numbers in triangular prism interpolation in the same color conversion apparatus.

40A shows an example of bottom surface data distribution to a color conversion table memory in a main part of the same color conversion device. FIG. 40B shows an example of top surface data distribution to a color conversion table memory in a main part of the same color conversion device.

41A shows an example of bottom surface data distribution to a color conversion table memory in a main part of the same color conversion device. FIG. 41B shows an example of top surface data distribution to a color conversion table memory in a main part of the same color conversion device.

FIG. 42 is a diagram showing an example of two-dimensional data distribution to a color conversion table memory in a main part of the color conversion apparatus.

[Explanation of symbols]

 Reference Signs List 101 Pixel input unit 102 Address generation unit 103 Triangular prism division selecting unit 104 Color conversion table memory 105 Triangular prism division type interpolation calculation unit 106 Host interface unit 107 Area processing unit 108 Color conversion table compulsory setting unit 109 Memory interface unit 110 Various registers 114 Internal Bus 116 color difference conversion unit

──────────────────────────────────────────────────続 き Continuing on the front page (72) Rika Iikawa, 3-10-1, Higashi-Mita, Tama-ku, Kawasaki-shi, Kanagawa Prefecture Inside Matsushita Giken Co., Ltd. No. 10-1 Matsushita Giken Co., Ltd. (72) Inventor Hiroteru Kodera 3-10-1, Higashi Mita, Tama-ku, Kawasaki City, Kanagawa Prefecture Matsushita Giken Co., Ltd. (56) References JP-A-5-46750 (JP, A) JP-A-5-75848 (JP, A) JP-A-61-7774 (JP, A) (58) Fields investigated (Int. Cl. ⁶ , DB name) H04N 1/40-1/409 H04N 1 / 46-1/60

Claims (14)

(57) [Claims] 1. A pixel input unit for inputting a color image signal, and a bottom surface of a cubic or cuboid group in a color space
One of the two triangular prisms created by dividing into triangles,
Determined by the magnitude relationship between the two signals that make up the color image signal
A triangular prism division selecting section for outputting a selection signal;
The image is divided into an upper bit part and a lower bit
Color according to the selection signal from the
Address generation unit for generating a translation table memory address
And a color storing a numerical value obtained by adding or subtracting only a predetermined amount of displacement from an output value after color conversion at a vertex position of a cubic or rectangular parallelepiped group in the color space, and reducing by a predetermined number of bits. a conversion table memory, at the time of color conversion operation of a color image is operated reads the color conversion table memory at address signals from the address generator, whereas, outside the time value setting of the color conversion table memory
A memory interface unit for performing a write operation at an external address from a computer of the unit, a lower bit portion of the color image signal, a selection signal from the triangular prism division selection unit, and an output value from a color conversion table memory. performed triangular prism interpolation, a triangular prism split interpolation operation unit for generating an interpolated output end by performing correction of the displacement amount, the notes
A color conversion device comprising a host interface unit for storing values from an external computer or the like into a color conversion table memory via a re-interface unit . 2. A pixel input section for inputting a color image signal, and a bottom surface of a group of cubes or rectangular parallelepipeds in a color space,
One of the two triangular prisms created by dividing into triangles,
The magnitude relationship between the two color difference signals constituting the color image signal
And a triangular prism division selection unit that outputs a selection signal determined by
The color image signal is divided into an upper bit portion and a lower bit portion.
Select from upper bit part and triangular prism division selector
Generate color conversion table memory address by signal
From the output value after color conversion at the apex position of the cubic or rectangular parallelepiped group in the color space, an address generation unit , and a numerical value obtained by adding or subtracting only a predetermined amount of displacement and reducing by a predetermined number of bits is stored. and a color conversion table memory and, at the time of color conversion operation of a color image is operated reads the color conversion table memory at address signals from the address generator, whereas, when the color conversion table memory of numerical settings from an external computer Performing a triangular prism interpolation from a memory interface unit that performs switching to perform a write operation with an external address, a lower bit portion of the color image signal, a selection signal from the triangular prism division selection unit, and an output value from a color conversion table memory; a triangular prism split interpolation operation unit for generating an interpolated output end by performing correction of the displacement amount, the memory A host interface unit for performing storage of the values from an external computer to the color conversion table memory via the interface unit, the window area to perform a different color conversion in one image of the color conversion through the host interface An area processing unit for setting coordinates,
Output from area processing unit or input another control line
By outputting a signal to the address generation unit,
Causing the address generator to generate a different address,
A color conversion table compulsory setting unit for appropriately switching a color conversion table to be used in one image. 3. The pixel input unit according to claim 1, wherein the color image signal represented by various color signals is represented by a 10-bit digital value of Y, Cr, and Cb. The color conversion device as described in the above. 4. An address generating unit for dividing a YCrCb signal into upper 3 bits and lower 7 bits in a normal mode, and dividing only a Y signal into upper 4 bits and lower 6 bits to generate a Cr signal.
3. The color conversion apparatus according to claim 1, wherein the Cb signal is used by switching between an address extension mode in which upper 3 bits and lower 7 bits are divided. 5. The color conversion table memory section comprises six memories which can be accessed simultaneously, and each memory has an output value at each of four points among eight vertices constituting a basic solid in a color space. The memory is composed of four memories stored exclusively and two memories each of which is responsible for the remaining two points, and the latter two memories are used to select a bank according to a selection signal from the triangular prism division selection unit according to claim 1. 3. The color conversion device according to claim 1, wherein an output value at one of the remaining two points can be output by switching. 6. The color conversion table memory section calculates a dynamic range of a numerical value range from −128 to 382 as an output value after color conversion, sets a displacement amount to 128, and subtracts the displacement amount 128 from the color conversion output value. 3. A color conversion apparatus according to claim 1, wherein a signed 8-bit numerical value ignoring lower one bit is accumulated so as to obtain -256 to 254. 7. The color conversion table memory unit according to claim 1, wherein the displacement amount is set to 512 in order to set a dynamic range of a numerical value range from −512 to 1534 as an output value after the color conversion. The color conversion according to claim 1 or 2, wherein a signed 10-bit numerical value obtained by subtracting the displacement amount 512 from the value and ignoring the lower 1 bit of the numerical value from -1024 to 1022 is stored. apparatus. 8. A triangular prism division type interpolation operation unit includes: a color conversion table memory; a difference generation unit that calculates a difference value between six numerical values simultaneously output from the color conversion table memory; A difference value selection unit that selects the difference value according to the selection signal, a multiplier that calculates a product of the difference value and the lower bit part of the color image signal, an adder that adds and subtracts the multiplication result, and a subtractor, 3. The color conversion device according to claim 1, further comprising a displacement register for correcting the displacement, and an output data limiter for limiting output data. 9. The output data limiter sets the interpolation calculation result after the displacement correction to 0 as final output data.
9. The color conversion apparatus according to claim 8, wherein the range is limited to a numerical range from 0 to 255 or from 0 to 1023. 10. The host interface unit reads a color conversion table memory by a signal from an address generation unit during image scanning when converting a video image of one frame in synchronization with a pixel clock, and performs image reading. 3. The color conversion apparatus according to claim 1, wherein the memory interface unit is controlled so that the host can access the color conversion table memory by an external address only during a blanking time period. 11. The color conversion table compulsory setting unit switches the color conversion table in units of a minimum of one pixel by inputting a pixel position signal during the color conversion operation for each pixel. Color conversion device. 12. One of three signals constituting a color signal
One of the select signal as a color difference reference signal, a signal corresponding to the difference between the remaining two signals and chrominance reference signal and color difference signals, said color
A color difference conversion unit for generating a difference signal and linear independent signal as lightness signal, a cube or bottom surface of the rectangular groups in between color space
One of the two triangular prisms formed by dividing into two triangles
Is larger than the two signals of the color difference signals constituting the color image signal.
Triangular prism division selection that outputs a selection signal determined by a small relation
And the color difference signal is divided into an upper bit portion and a lower bit portion.
Select from upper bit part and triangular prism division selector
Generate color conversion table memory address by signal
From the output value after color conversion at the vertex position of the cubic or rectangular parallelepiped group in the color space, an address generator , and a numerical value obtained by adding or subtracting only a predetermined amount of displacement and reducing by a predetermined number of bits is stored. and a color conversion table memory and, at the time of color conversion operation of a color image is operated reads the color conversion table memory at address signals from the address generator, whereas, when the color conversion table memory of numerical settings from an external computer Performing a triangular prism interpolation from a memory interface unit that performs switching to perform a writing operation with an external address, a lower bit portion of the color image signal, a selection signal from the triangular prism division selecting unit, and an output value from a color conversion table memory; a triangular prism split interpolation operation unit for generating an interpolated output end by performing correction of the displacement amount, the memory Color conversion device comprising a host interface unit for performing storage of the values from an external computer to the color conversion table memory via the interface unit. 13. When obtaining an output color signal for an input color signal by using a color conversion table memory storing values of an output color signal for the input color signal, a domain of the input color signal is defined as a unit. Divide into hexahedrons, set a new coordinate system with the diagonal line of the unit hexahedron as the main axis, and make the other two axes the same as the input color signal, and triangular prisms in the new coordinate system created using the vertices of the adjacent hexahedron And determining which triangular prism the input color is in, and interpolating the output color signal for the input color signal using the color conversion table memory accumulation values corresponding to the vertices forming the triangular prism. A color conversion method characterized by the following. 14. RGB, YMC, XYZ, YCrCb
Dividing the color space into 2 ^3N number of unit cubes by equally dividing each axis of the three-dimensional color space configured 2 ^N digital signals of the three W bit width equal, for the unit cube side the length remembers the output value independently for 8 vertices of a unit cube lattice point position falls even numbered all the axes in 3-space that extends axis outside one unit or more input 3-dimensional space as a unit, select two unit cube in contact Ritonari by the upper N bits of each input signal, the unit cube selected
Depending on whether the upper N bits of each input signal are even or odd, if all are even, one of the stored even-numbered unit cubes is selected, and each of the input signals of the selected unit cube is selected.
If at least one of the upper N bits is odd, an adjacent even-numbered unit cube is selected, and eight addresses at which output values corresponding to the selected unit cube are stored are calculated. eight of the indicated by address calculation
Reading the output value for the 8 vertices of a unit cube in parallel,
Outputs for the 8 vertices of the 8 unit cubes read out
By arranging force values according to a certain rule, the input signal
Corresponding to the output value at the vertex position of the unit cube containing
And the output value at the vertex position of the unit cube is
Lower (W-N) bits of each axis excluding the bit and the weight
Color conversion method you characterized in that the three-dimensional interpolation calculation Te.