JP2006203532A - Color conversion apparatus, color conversion method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a technology capable of carrying out color conversion adapted to requirements of a high image quality tendency without decreasing the processing speed and deteriorating the accuracy. <P>SOLUTION: A color conversion apparatus includes: a pixel input section 10 for receiving an input pixel value; an index conversion section 20 for converting the input pixel value into an index signal; a delta conversion section 30 for converting the input pixel value into a delta signal; an input buffer 40 for storing the index signal and the delta signal; a first conversion table storage section 50 for storing a first conversion table wherein each grating point of an input color space is cross-referenced with each interpolation coefficient; a temporary storage section 60 for caching the first conversion table; a second conversion table storage section 61 for storing a second conversion table obtained by reducing the information amount of the first conversion table; a control section 70 for acquiring the first block table or the second block table according to the discrimination on the basis of the index signal; an interpolation processing section 80 for obtaining an output pixel value on the basis of the interpolation coefficient stored in the first or second conversion table and the delta signal; and a pixel output section 90 for outputting the output pixel value. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a color conversion device that converts pixel values expressed in a predetermined color space into pixel values expressed in different color spaces.

Color reproduction devices such as displays and printers have different color reproduction principles depending on their types. Therefore, the color space for expressing a color differs depending on the type of device. Thus, in order to reproduce the color reproduced in one device on another device, it is necessary to convert the pixel value expressed in the color space of one device to the pixel value expressed in the color space of the other device Come out.
However, for conversion with strong non-linearity, such as conversion from RGB color space (RGB color space) such as a display to a printer color space (YMCK color space) It is difficult to perform the calculation by using the conversion formula. Therefore, such color conversion is usually performed by referring to a conversion table in which each lattice point in one color space is associated with a pixel value in the other color space.

On the other hand, in recent years, the demand for high-quality images has increased, and the gradation (data amount) of input data has also increased. Therefore, in order to realize color reproduction with higher accuracy, a conversion table used for color conversion is required to have a large capacity.
However, if such color conversion processing is realized by, for example, an ASIC (Application Specific Integrated Circuit) in order to improve the processing speed, a large-capacity memory is provided inside the ASIC, and the cost is also high. Will be greatly affected.
Against this background, conventionally, in data conversion processing in which data conversion processing is performed using a look-up table (LUT) and interpolation calculation means, a technique for improving calculation accuracy by switching between linear interpolation calculation and nonlinear interpolation calculation has been proposed. (See, for example, Patent Document 1). In this method, the mode when the operating frequency of the system is slow is referred to as a secondary interpolation calculation mode, and the calculation result in the nonlinear interpolation calculation unit is added to the calculation result in the linear interpolation calculation unit to obtain an output. On the other hand, the mode when the operating frequency of the system is fast is called a linear interpolation calculation mode, and the calculation result in the linear interpolation calculation unit is output as it is.

JP-A-8-137451 (page 4, FIG. 3)

However, in the invention of Patent Document 1, since only the secondary interpolation calculation mode and the linear interpolation calculation mode are switched according to the operating frequency of the system, the processing speed due to the calculation load is reduced in the secondary interpolation calculation mode. There was a problem that there was a risk of a decrease.
Further, in the linear interpolation calculation mode, there is a problem that the problem of improving the accuracy of the processing result in Patent Document 1 is still not solved.

  The present invention has been made in order to solve the technical problems as described above, and an object of the present invention is to perform color conversion that meets the demand for higher image quality without reducing the processing speed and accuracy. Is to make it.

For this purpose, in the present invention, when converting a pixel value expressed in a predetermined color space into a pixel value expressed in a different color space, the grid points are arranged sparsely with a conversion table in which the grid points are densely arranged. Switched between the conversion table. That is, the color conversion device of the present invention is a device that converts a pixel value expressed in the first color space into a pixel value expressed in the second color space, and is referred to in the conversion. Storage means for storing second information obtained by reducing the amount of information of the first information, and reference to the first information for converting the first pixel value expressed in the first color space Processing means for determining whether to refer to the second information or not based on the first pixel value.
Here, the first information is, for example, a first conversion table indicating a predetermined number of correspondences between pixel values expressed in the first color space and pixel values expressed in the second color space. Yes, the second information is a second conversion that indicates a smaller number of correspondences between the pixel value expressed in the first color space and the pixel value expressed in the second color space than the predetermined number. It is a table.

  Further, according to the present invention, when converting a pixel value expressed in a predetermined color space into a pixel value expressed in a different color space, a conversion table in which lattice points are densely arranged and a conversion table in which sparsely arranged lattice points are provided. It can also be understood as a method of switching and using. In this case, the color conversion method of the present invention is a method for converting the pixel value expressed in the first color space into the pixel value expressed in the second color space, and is expressed in the first color space. The received first pixel value, and if the received first pixel value does not satisfy the predetermined standard, the pixel value in the second color space obtained using the first information is And when the first pixel value satisfies the criterion, the second information obtained by using the second information obtained by reducing the information amount of the first information is within the second color space. Outputting a pixel value.

  On the other hand, the present invention can also be understood as a program for causing a computer to realize a predetermined function. In that case, the program of the present invention has a function of accepting the first pixel value expressed in the first color space in the computer, and if the accepted first pixel value does not satisfy the predetermined standard, When the pixel value in the second color space obtained using the information of 1 is output and the first pixel value satisfies the standard, the information amount of the first information is reduced. And a function of outputting a pixel value in the second color space obtained by using the obtained second information.

  According to the present invention, it is possible to perform color conversion that meets the demand for higher image quality without reducing the processing speed and accuracy.

The best mode for carrying out the present invention (hereinafter referred to as “embodiment”) will be described below in detail with reference to the accompanying drawings. In the present embodiment, a color space for expressing pixels (input pixels) input to the color conversion device is called an “input color space”, and pixels (output pixels) output from the color conversion device are expressed. A color space for this purpose is called an “output color space”.
(First embodiment)
FIG. 1 is a diagram showing a configuration of a color conversion apparatus according to the first embodiment of the present invention.
As illustrated, the color conversion apparatus according to the present embodiment includes a pixel input unit 10, an index conversion unit 20, a delta conversion unit 30, an input buffer 40, a first conversion table holding unit 50, and a temporary storage unit. 60, a second conversion table holding unit 61, a control unit 70, an interpolation processing unit 80, and a pixel output unit 90.

The pixel input unit 10 receives an input pixel value (input pixel value) in the input color space. The index conversion unit 20 acquires an input pixel value, outputs an index signal indicating a partial space to which the input pixel value belongs among the partial spaces in the input color space, the delta conversion unit 30 acquires the input pixel value, A delta signal indicating the position of the input pixel value in the subspace is output. The input buffer 40 is a buffer that reads and stores the index signal and the delta signal.
The first conversion table holding unit 50 stores the value of the corresponding output pixel in the output color space (hereinafter referred to as “interpolation coefficient”) for each grid point located at the boundary of the partial space of the input color space. (Hereinafter referred to as “first conversion table”). In the present embodiment, it is assumed that a conversion table that can be used when the input pixel value is 10 bits / pix is held as the first conversion table.
The temporary storage unit 60 is a table obtained by dividing the first conversion table (hereinafter referred to as “first block table”) and a table functioning as an index for the block table (hereinafter referred to as “first index table”). And remember).

  The second conversion table holding unit 61 stores, for a specific block in the input color space, an interpolation coefficient in the output color space for each grid point located at the boundary of the partial space in the block (hereinafter, “ Storage means for storing a second conversion table including a second block table ”and a table functioning as an index for the second block table (hereinafter referred to as“ second index table ”). Here, the second conversion table is a conversion table that manages the grid points arranged more sparsely than the first conversion table. In this embodiment, it is assumed that a conversion table that can be handled when the input pixel value is 8 bits / pix is held as the second block table (assuming that the upper 4 bits represent an index and the lower 4 bits are interpolated). And).

  If it is determined that the second conversion table should be referred to based on the index signal stored in the input buffer 40, the control unit 70 acquires the second block table from the second conversion table holding unit 61. On the other hand, if it is determined that the first conversion table should be referred to based on the index signal stored in the input buffer 40, the first block table is acquired from the temporary storage unit 60 or the first conversion table holding unit 50. The control unit 70 that performs such processing can be grasped as processing means. Further, the interpolation processing unit 80 calculates the value of the output pixel (output pixel value) to be output based on the interpolation coefficient acquired from the first or second block table and the delta signal stored in the input buffer 40. Ask. The pixel output unit 90 outputs this output pixel value.

  In this embodiment, among these functions, the index conversion unit 20, the delta conversion unit 30, the input buffer 40, the temporary storage unit 60, the second conversion table holding unit 61, and the control unit 70 The portion composed of the interpolation processing unit 80 is realized by LSI (Large Scale Integration). On the other hand, the first conversion table holding unit 50 is realized by a main memory.

The color conversion apparatus having such a configuration operates as follows. In the following description of the operation, color conversion from the RGB color space to the YMCK color space is taken as an example.
First, the pixel input unit 10 receives input pixel values (R _I , G _I , B _I ). Then, the index conversion unit 20 obtains an index signal (R _H , G _H , B _H ) from the input pixel value, and the delta conversion unit 30 obtains a delta signal (R _L , G _L , B _L ) from the input pixel value. Ask. As described above, the index signal is a signal indicating a partial space to which the input pixel value belongs in the partial space in the input color space, and the delta signal is a signal indicating the position of the input pixel value in the partial space. Therefore, the relations R _I = R _H + R _L , G _I = G _H + G _L , and B _I = B _H + B _L are established.
The signal output by the index conversion unit 20 and the delta conversion unit 30 is sequentially performed on the input pixel value, and the input buffer 40 prefetches and stores the pair of the index signal and the delta signal as long as the size allows. This is because it is assumed that reading from the first conversion table holding unit 50 to the temporary storage unit 60 and the interpolation processing unit 80 is performed at a low speed.

Next, the operations of the control unit 70 and the interpolation processing unit 80, which are the core of the color conversion processing in this embodiment, will be described. The contents stored in the second conversion table holding unit 61 will be described. In the drawings referred to in this specification, the RGB color space is shown by setting the R axis in the horizontal direction, the G axis in the vertical direction, and the B axis in the direction toward the back. The coordinates of each point in the RGB space are expressed as (R value, G value, B value).
FIG. 2 is a diagram showing the entire RGB color space. In this figure, n lattice points are provided on each of the R, G, and B axes. Therefore, this RGB color space is (0,0,0), (n-1,0,0), (n-1,0, n-1), (0,0, n-1), (0 , N−1, 0), (n−1, n−1, 0), (n−1, n−1, n−1), and a cube with (0, n−1, n−1) as vertices. Can be considered.
And the 1st conversion table hold | maintained at the 1st conversion table holding | maintenance part 50 matches an interpolation coefficient with each grid point in such RGB color space.

By the way, in this embodiment, the RGB color space is divided into blocks, and the first block table corresponding to each block is stored in the temporary storage unit 60.
The division of the RGB color space into blocks is performed, for example, as shown in FIG. That is, as shown by a thick line in FIG. 3A, a plurality of partial spaces surrounded by lattice points are collected to form one block.
FIG. 3B shows an enlarged view of one block. Of the eight vertices of each block, the vertex closest to the origin of the RGB coordinate system is defined as a block index, and is represented by (R _B , G _B , B _B ). In the present embodiment, in each block, m lattice points are provided on each axis (m is a divisor of n). Therefore, each block has (R _B , G _B , B _B ), (R _B + m−1, G _B , B _B ), (R _B + m−1, G _B , B _B + m−1), (R _B , G _B , B _B + m−1), (R _B , G _B + m−1, B _B ), (R _B + m−1, G _B + m−1, B _B ), (R _B + m−1, G _B + m−1, B _B + m−1) and (R _B , G _B + m−1, B _B + m−1) can be regarded as cubes having apexes.

In FIG. 3B, it is assumed that a point P is an input pixel value. In this case, if the coordinates of the vertex closest to the origin of the RGB coordinate system among the eight vertices of the partial space to which the point P belongs are (R _H , G _H , B _H ), the index signal indicating this partial space is ( R _H , G _H , B _H ). When the coordinates of the point P in the RGB space are (R _H + R _L , G _H + G _L , B _H + B _L ), the delta signal indicating the position of the point P in the subspace is (R _L , G _L , B _L ). As shown in FIG. 3B, one block is a collection of a plurality of such partial spaces.
And the 1st block table memorize | stored in the temporary memory part 60 matches an interpolation coefficient with each lattice point in such a block.

In the present embodiment, the grid points of the specific block in the RGB color space in FIG. 2 are arranged more sparse than in FIG. 2, and the second block table corresponding to such a block is stored in the second conversion table holding unit. 61 is stored.
For example, color conversion from the RGB color space to the YMCK color space has high nonlinearity. Specifically, when the interpolation coefficients for the grid points at regular intervals in the RGB color space hardly change in the YMCK color space, they may change drastically if they change in the same way. In subspaces that hardly change or change extremely, the higher the ratio of linear interpolation, the farther away from accurate color conversion. On the other hand, in a subspace that changes in the same way, although depending on the spacing of the grid points in the RGB color space, pixel values with a certain degree of accuracy can be obtained even if linear interpolation is performed. Therefore, in such a partial space, even if the degree of dependence on linear interpolation is increased, it is possible to perform conversion with a certain degree of accuracy. Therefore, in the present embodiment, a second conversion table is provided in which the lattice point intervals are sparse for the portion of the first conversion table in which pixel values with a certain degree of accuracy can be obtained even if linear interpolation is performed.

In the present embodiment, the above-described property in the color conversion from the RGB color space to the YMCK color space is utilized, and even if a pixel value is input at 10 bits / pix in a partial space that changes in the same manner, 2. Color conversion is performed using a conversion table. Specifically, of the 10 bits / pix, the lower 2 bits are ignored and recognized as 8 bits / pix, or interpolation is performed with the lower 6 bits. This is because such a process does not cause a large error in a subspace that changes in the same manner.
On the other hand, the second conversion table cannot be used to perform accurate color conversion for a subspace that hardly changes or a subspace that changes extremely. Therefore, color conversion is performed using the first conversion table.

FIG. 4 shows a block in which the arrangement of lattice points is sparse. Of the eight vertices of the block, the vertex closest to the origin of the RGB coordinate system is defined as a block index, and is represented by (R _B , G _B , B _B ). In this embodiment, the grid points are sparsely arranged by providing one grid point for k (k> 1) grid points in the RGB color space of FIG. Note that t lattice points are provided on each axis in one block. Therefore, each _{block, (R B, G B,} B B), (R B + (t-1) k, G B, B B), (R B + (t-1) k, G B, B _{B + (t-1) k} ), (R B, G B, B B + (t-1) k), (R B, G B + (t-1) k, B B), (R B + _{(t-1) k, G} B + (t-1) k, B B), (R B + (t-1) k, G B + (t-1) k, B B + (t-1) k), (R _B , G _B + (t−1) k, B _B + (t−1) k) can be regarded as a cube having vertices.

In FIG. 4, it is assumed that a point P is an input pixel value. In this case, if the coordinates of the vertex closest to the origin of the RGB coordinate system among the eight vertices of the partial space to which the point P belongs are (R _H , G _H , B _H ), the index signal indicating this partial space is ( R _H , G _H , B _H ). When the coordinates of the point P in the RGB space are (R _H + R _L , G _H + G _L , B _H + B _L ), the delta signal indicating the position of the point P in the subspace is (R _L , G _L , B _L ). As shown in FIG. 4, one block is a collection of a plurality of such partial spaces.
And the 2nd block table memorize | stored in the 2nd conversion table holding | maintenance part 61 matches an interpolation coefficient with each lattice point in such a block.

Next, an example of the first conversion table held in the first conversion table holding unit 50 will be described with reference to FIGS.
FIG. 5 shows an example of ordering for arranging the grid points in the RGB color space in one dimension. In this way, adjacent lattice points are arranged so that addresses are as continuous as possible on the conversion table. As a result, memory characteristics such as burst transfer can be utilized, and the reading speed can be improved.
The ordering in FIG. 5 will be described in more detail. First, lattice points on the foremost plane (plane B = 0) in FIG. 5 are (0, 0, 0), (1, 0, 0),..., (N−1, 0, 0), ( (0, 1, 0), (1, 1, 0), ..., (n-1, 1, 0), ..., (0, n-1, 0), (1, n-1, 0), ... , (N-1, n-1, 0) in this order. Next, the lattice points on the innermost surface (plane B = 1) are also arranged in the order of (0, 0, 1),..., (N−1, n−1, 1). This is performed up to the lattice points on the innermost surface (plane B = n−1), and n ³ lattice points arranged three-dimensionally are arranged one-dimensionally.

Then, each grid point in the RGB color space arranged in one dimension is mapped to an address on the memory as shown in FIG. Here, for the sake of simplicity, it is assumed that the head address is “0” and the next grid point is mapped to an address counted up by one.
In addition, when each grid point in the RGB color space is continuously picked up and mapped in this way, generally, the address A is
A = αR + βG + γB + δ
It can be expressed as. FIG. 6 shows an example in which α = 1, β = n, γ = n ² , and δ = 0.

An example of the first index table and the first block table stored in the temporary storage unit 60 will be described with reference to FIGS.
FIG. 7 shows an example of ordering for arranging the grid points in the block cut out from the RGB color space in a one-dimensional manner. First, lattice points on the foremost surface (plane B = B _B ) in FIG. 7 are (R _B , G _B , B _B ), (R _B +1, G _B , B _B ) _{,. + m-1, G B,} B B), (R B, G B + 1, B B), (R B + 1, G B + 1, B B), ..., (R B + m-1, G B + 1, B _{B), ..., (R B} , G B + m-1, B B), (R B + 1, G B + m-1, B B), ..., (R B + m-1, G B + m-1, B Arrange in the order of _B ). Next, the lattice points on the innermost plane (plane B = B _B +1) are similarly (R _B , G _B , B _B +1),..., (R _B + m−1, G _B + m−. 1, B _B +1). This is performed up to the lattice points on the innermost surface (plane B = B _B + m−1), and m ³ lattice points arranged in three dimensions are arranged one-dimensionally.
Then, each lattice point in this one-dimensionally arranged block is mapped to an address on the memory.

The temporary storage unit 60 does not necessarily store only one block table. Accordingly, a first index table indicating which block table is stored from which address is provided.
FIG. 8A shows an example of the first index table. The first index table associates the block index indicating the position of each block, the address of the first block table corresponding to each block, and the reference count (hit count) of the first block table corresponding to each block. Is. For example, for the block index (R ₀ , G ₀ , B ₀ ), the address “a ₀ ” and the hit count “N ₀ ” are for the block index (R ₁ , G ₁ , B ₁ ). The address “a ₁ ” and the hit count “N ₁ ” are stored.

Further, each lattice point in the one-dimensionally arranged block in FIG. 7 is mapped to an address on the memory as shown in FIG. 8B. The mapping method in the block is the same as in FIG. However, the head address of the first block table corresponding to each block index is the address associated with the block index in the first index table. Specifically, the first address of the first block table corresponding to the block index (R ₀ , G ₀ , B ₀ ) is “a ₀ ” and corresponds to the block index (R ₁ , G ₁ , B ₁ ). The first address of the first block table is “a ₁ ”. The next grid point is assumed to be mapped to an address counted up by 1 as in FIG.

In addition, when each grid point in the block is continuously picked up and mapped in this way, generally, the address A is
A = αR + βG + γB + δ
It can be expressed as. In the first block table for the block index (R ₀ , G ₀ , B ₀ ) in FIG. 8B, α = 1, β = m, γ = m ² , δ = a ₀ −R ₀ −mG ₀ −. m ² B ₀ . In the first block table for the block index (R ₁ , G ₁ , B ₁ ), α = 1, β = m, γ = m ² , δ = a ₁ −R ₁ −mG ₁ −m ² B ₁ It is.
Here, the number of first block tables in FIG. 8B is determined according to the capacity of the temporary storage unit 60. Then, as many entries as the first block table are provided in the first index table of FIG.

Further, an example of the second index table and the second block table stored in the second conversion table holding unit 61 will be described with reference to FIGS.
FIG. 9 shows an example of ordering for arranging grid points in a specific block in the RGB color space in a one-dimensional manner. First, lattice points on the foremost surface (plane B = B _B ) in FIG. 9 are (R _B , G _B , B _B ), (R _B + k, G _B , B _B ) _{,. + (t-1) k,} G B, B B), (R B, G B + k, B B), (R B + k, G B + k, B B), ..., (R B + (t-1 _{) k, G B + k,} B B), ..., (R B, G B + (t-1) k, B B), (R B + k, G B + (t-1) k, B B), ..., (R _B + (t-1) k, G _B + (t-1) k, B _B ). Next, the lattice points on the innermost surface (plane B = B _B + k) are similarly (R _B , G _B , B _B + k),..., (R _B + (t−1) k, Arrange in the order of G _B + (t−1) k, B _B + k). This is performed up to the lattice points on the innermost surface (plane B = B _B + (t−1) k), and t ³ lattice points arranged in three dimensions are arranged one-dimensionally.
Then, each lattice point in this one-dimensionally arranged block is mapped to an address on the memory.

The second conversion table holding unit 61 does not necessarily store only one block table. Therefore, a second index table indicating which block table is stored from which address is provided.
FIG. 10A shows an example of the second index table. The second index table associates a block index indicating the position of each block, the number t of grid points provided on one axis of each block, and the address of the second block table corresponding to each block. It is. For example, for the block index (R ′ ₀ , G ′ ₀ , B ′ ₀ ), the number of grid points “t ₀ ” and the address “a ′ ₀ ” are changed to the block index (R ′ ₁ , G ′ ₁ , For B ′ ₁ ), the number of lattice points “t ₁ ” and the address “a ′ ₁ ” are stored.

Further, each lattice point in the block arranged in one dimension in FIG. 9 is mapped to an address on the memory as shown in FIG. The mapping method in the block is the same as in FIG. However, the start address of the second block table corresponding to each block index is the address associated with that block index in the second index table. Specifically, the start address of the second block table corresponding to the block index (R ′ ₀ , G ′ ₀ , B ′ ₀ ) is “a ′ ₀ ”, and the block index (R ′ ₁ , G ′ ₁ , The leading address of the second block table corresponding to B ′ ₁ ) is “a ′ ₁ ”. The next grid point is assumed to be mapped to an address counted up by 1 as in FIG.

In addition, when each grid point in the block is continuously picked up and mapped in this way, generally, the address A is
A = αR + βG + γB + δ
It can be expressed as. In the second block table for the block indexes (R ′ ₀ , G ′ ₀ , B ′ ₀ ) in FIG. 10B, α = 1 / k, β = t ₀ / k, γ = (t ₀ / k) ² , δ = a ′ ₀ − (1 / k) R ′ ₀ − (t ₀ / k) G ′ ₀ − (t ₀ / k) ² B ′ ₀ . In the second block table for the block index (R ′ ₁ , G ′ ₁ , B ′ ₁ ), α = 1 / k, β = t ₁ / k, γ = (t ₁ / k) ² , δ = a ′ ₁ − (1 / k) R ′ ₁ − (t ₁ / k) G ′ ₁ − (t ₁ / k) ² B ′ ₁
Here, the number of second block tables in FIG. 10B is determined according to the capacity of the second conversion table holding unit 61. Then, as many entries as the second block tables are provided in the second index table of FIG.

In this way, the operation of the control unit 70 starts with information stored in the first conversion table holding unit 50, the temporary storage unit 60, and the second conversion table holding unit 61.
FIG. 11 is a flowchart showing the operation of the control unit 70.
First, the control unit 70 extracts index signals (R _H , G _H , B _H ) from the input buffer 40 (step 101).
Then, it is determined whether or not block information including a grid point corresponding to the index signal in the RGB color space exists in the second conversion table (step 102). The block in this case is a specific block in the RGB color space, and is a block obtained by arranging the grid point intervals managed by the first conversion table holding unit 50 more sparsely.
Specifically, referring to the second index table,
R ′ _j ≦ R _H <R ′ _j + t _j k
G ′ _j ≦ _GH <G ′ _j + t _j k
B ′ _j ≦ B _H <B ′ _j + t _j k
(j = 0, 1, 2, ...)
It is determined whether or not j satisfying the above condition exists.

  As a result, when it is determined that such block information exists in the second conversion table, the block index of the block is extracted from the second index table (step 103). Then, the address of the second block table corresponding to the block index is obtained (step 104).

On the other hand, when it is determined that such block information does not exist in the second conversion table, a block index of a block including a lattice point corresponding to the index signal in the RGB space is obtained (step 105). The block in this case is a block obtained by dividing the RGB color space.
In particular,
R _B ≦ R _H <R _B + m (R _B = 0, m, 2 m,..., N−m)
_{G B ≦ G H <G B} + m (G B = 0, m, 2m, ..., n-m)
B _B ≦ B _H <B _B + m (B _B = 0, m, 2 m,..., N−m)
Lattice points (R _B , G _B , B _B ) that satisfy the above are obtained.

Next, the control unit 70 determines whether or not the block index (R _B , G _B , B _B ) exists in the first index table of the temporary storage unit 60 (step 106).
As a result, when it is determined that the block index exists in the first index table, 1 is added to the number of hits corresponding to the block index (step 107). Then, the address of the first block table corresponding to the block index is obtained (step 108).

On the other hand, when it is determined that the block index does not exist in the first index table, the operation proceeds to an operation of reading the block table from the first conversion table holding unit 50. However, at that time, there may be a case where the temporary storage unit 60 does not have a capacity for newly adding a block table. Therefore, first, it is determined whether or not there is a margin (empty area) for writing the block table in the temporary storage unit 60 (step 109).
As a result, when it is determined that there is no room for writing, the number of hits for each block stored in the first index table is referred to, and the first block table of the block with the smallest number of hits is deleted from the temporary storage unit 60. (Step 110), go to Step 111. On the other hand, if it is determined that there is room for writing, the process proceeds to step 111 as it is.

And the control part 70 cuts out the block table corresponding to the block index calculated | required by step 105 from a 1st conversion table, and writes it in the temporary memory part 60 (step 111). Also, the head address of the written block table is acquired, and the block index and its address are written into the first index table (step 112). In this case, the hit count is set to “0”.
Finally, the control unit 70 outputs the address and block index of the first block table or the second block table (hereinafter simply referred to as “block table”) and the block index to the interpolation processing unit 80 (step 113).

As a result, the operation of the interpolation processing unit 80 starts.
FIG. 12 is a flowchart showing the operation of the interpolation processing unit 80.
First, the interpolation processing unit 80 acquires an index signal and a delta signal from the input buffer 40, and acquires an address of the block table and a block index from the control unit 70 (step 201). The index signal may be acquired from the control unit 70.

Next, the interpolation processing unit 80 obtains an interpolation coefficient for the input pixel value and performs an interpolation process. Here, as an example of an interpolation method, four-point interpolation is employed in which interpolation is performed using interpolation coefficients for four vertices of a tetrahedron including input pixels. Of course, other known interpolation methods may be employed.
First, the interpolation processing unit 80 obtains a tetrahedron including input pixel values (R _I , G _I , B _I ) (step 202). Specifically, in the cube as shown in FIG. 13, it is determined whether the input pixel value is included in tetrahedron ABCG, tetrahedron ABFG, tetrahedron ACDG, tetrahedron ADGH, tetrahedron AEFG, or tetrahedron AEGH. To do. This can be easily determined based on the magnitude relationship between R _L , G _L , and B _{L in} the delta signal (R _L , G _L , B _L ).

Thereafter, the interpolation processing unit 80 obtains addresses A ₀ , A ₁ , A ₂ , and A _{3 in} which the interpolation coefficients for the four vertices of the tetrahedron obtained here are stored (step 203).
Specifically, it calculates | requires as follows. First, an address corresponding to each vertex in FIG. 13 is obtained. Addresses at which interpolation coefficients for vertices A, B, C, D, E, F, G, and H are stored are respectively A _A , A _B , A _C , A _D , A _E , A _F , A _G , A _{If H} , then: Here, the address of the block table obtained by the control unit 70 is “a”.
_{A A = a + (R H} -R B) + m (G H -G B) + m 2 (B H -B B)
_{A B = a + (R H} + 1-R B) + m (G H -G B) + m 2 (B H -B B)
_{A C = a + (R H} + 1-R B) + m (G H -G B) + m 2 (B H + 1-B B)
_{A D = a + (R H} -R B) + m (G H -G B) + m 2 (B H + 1-B B)
_{A E = a + (R H} -R B) + m (G H + 1-G B) + m 2 (B H -B B)
_{A F = a + (R H} + 1-R B) + m (G H + 1-G B) + m 2 (B H -B B)
A _G = a + (R _H + 1−R _B ) + m (G _H + 1−G _B ) + m ² (B _H + 1−B _B )
_{A H = a + (R H} -R B) + m (G H + 1-G B) + m 2 (B H + 1-B B)

The relationship between the tetrahedron including the input pixel value and the addresses A ₀ , A ₁ , A ₂ , A ₃ is as follows.
For tetrahedron ABCG, A ₀ = A _A , A ₁ = A _B , A ₂ = A _C , A ₃ = A _G
For tetrahedron ABFG, A ₀ = A _A , A ₁ = A _B , A ₂ = A _F , A ₃ = A _G
In the case of tetrahedron ACDG, A ₀ = A _A , A ₁ = A _C , A ₂ = A _D , A ₃ = _AG
In the case of tetrahedron ADGH, A ₀ = A _A , A ₁ = A _D , A ₂ = A _G , A ₃ = A _H
In the case of tetrahedron AEFG, A ₀ = A _A , A ₁ = A _E , A ₂ = A _F , A ₃ = A _G
In the case of tetrahedron AEGH, A ₀ = A _A , A ₁ = A _E , A ₂ = A _G , A ₃ = A _H

Then, the interpolation processing unit 80, with reference to the block table, the address _{A 0} from the interpolation coefficients _{Y 0, M 0, C 0} , the _{K 0,} interpolated from the address _{A 1} coefficients _{Y 1, M 1, C 1} , K _1, take out the interpolation coefficients _{Y 2, M 2, C 2} , K 2 from the address _{a 2,} the address _{a 3} from the interpolation coefficients _{Y 3, M 3, C 3} , K 3 , respectively (step 204).

Further, the interpolation processing unit 80 performs multiplication coefficients Wy ₀ , Wm ₀ , Wc ₀ , Wk ₀ , and interpolation coefficients Y ₁ , M ₁ , C ₁ , K ₁ for the interpolation coefficients Y ₀ , M ₀ , C ₀ , K ₀ . Multiplication coefficients Wy ₁ , Wm ₁ , Wc ₁ , Wk ₁ , interpolation coefficients Y ₂ , M ₂ , C ₂ , K ₂ , multiplication coefficients Wy ₂ , Wm ₂ , Wc ₂ , Wk ₂ , interpolation coefficients Y ₃ , M ₃ , Multiplication coefficients Wy ₃ , Wm ₃ , Wc ₃ , and Wk ₃ for C ₃ and K ₃ are obtained (step 205).

These multiplication coefficients are obtained as follows, for example. That is, first, the point P is set as the input pixel value, and among the vertices of the tetrahedron including the point P, the vertices corresponding to the interpolation coefficients Y ₀ , M ₀ , C ₀ , K ₀ are P ₀ , the interpolation coefficients Y ₁ , M ₁ , C ₁ , K ₁ vertices corresponding to P ₁ , interpolation coefficients Y ₂ , M ₂ , C ₂ , K ₂ vertices corresponding to P ₂ , interpolation coefficients Y ₃ , M ₃ , C ₃ , K ₃ the corresponding vertex and _{P 3.} Then, _{P 0} and P the point at which the straight line intersects the surface facing the _{P 0} connecting the _{P 0} ', _{P 1} a point where the straight line intersects the surface facing the _{P 1} connecting the _{P 1} and P', _{P 2} P ₂ straight line connecting the P is a point of intersection with the surface facing the P ₂ and 'a point at which a straight line connecting the P ₃ and P intersects the surface facing the P ₃ P _3' as the calculation as follows I do.

Then, the interpolation processing unit 80 converts Y from the interpolation coefficients Y ₀ , Y ₁ , Y ₂ , Y ₃ and the multiplication coefficients Wy ₀ , Wy ₁ , Wy ₂ , Wy ₃ to the interpolation coefficients M ₀ , M ₁ , M _2. , M ₃ and multiplication coefficients Wm ₀ , Wm ₁ , Wm ₂ , Wm ₃ and M to interpolation coefficients C ₀ , C ₁ , C ₂ , C ₃ and multiplication coefficients Wc ₀ , Wc ₁ , Wc ₂ , Wc ₃ To C and K from the interpolation coefficients K ₀ , K ₁ , K ₂ , K ₃ and the multiplication coefficients Wk ₀ , Wk ₁ , Wk ₂ , Wk ₃ , respectively, are obtained as follows.

Finally, the interpolation processing unit 80 outputs the obtained Y, M, C, and K to the pixel output unit 90 (step 207).
Thereby, the pixel output unit 90 outputs an actual output pixel.
Thus, the operation of the first embodiment is completed.

  Thus, in this embodiment, a specific portion of the first conversion table for performing color conversion from a predetermined color space to a different color space is stored as a second conversion table with sparsely arranged grid points. I tried to do it. This makes it possible to perform color conversion in response to a request for high image quality without reducing the processing speed and accuracy. In particular, if the second conversion table holding unit 61 is provided inside the ASIC and the first conversion table holding unit 50 is realized by a main memory or the like outside the ASIC, a great effect in terms of cost and speed can be obtained.

(Second embodiment)
The configuration of the color conversion apparatus in the second embodiment of the present invention is substantially the same as the configuration in the first embodiment shown in FIG. The only difference is the contents of the second index table stored in the second conversion table holding unit 61 and the operation of the control unit 70 corresponding thereto.
FIG. 14 shows an example of the second index table in the present embodiment. In the second index table in the present embodiment, in addition to the information in the second index table in the first embodiment, the position information of the input pixel on the input image is referred to when the pixel value of the input pixel is converted. Stored for the block index of the two-block table.

  For example, it is assumed that the pixels on the input image are input in the direction from left to right (main scanning direction) in each line, and the line is input in the direction from top to bottom (sub scanning direction). In this case, the pixel value of the input pixel on a certain line and the pixel value of the pixel at the same main scanning position on the line immediately above are highly likely to approximate. In the present embodiment, by utilizing this tendency, information on the main scanning position X is added to the second index table of the second conversion table holding unit 61 and the processing when referring to the second conversion table holding unit 61 is made efficient. It has become.

Specifically, in step 102 of FIG. 11, the control unit 70 performs the following processing.
That is, first, the main scanning position X of the input pixel that is the source of the input pixel value to be converted is acquired. This is assumed to be transferred to the control unit 70 via the pixel input unit 10 and the like together with the input pixel value. Then, among the entries of the second index table shown in FIG. 14, the entry in which the main scanning position closest to the acquired main scanning position X is preferentially referenced.

Incidentally, in FIG. 14, a plurality of main scanning positions of the second index table are stored in one entry. Of course, there may be a case where there is only one main scanning position, but normally, it is considered that colors that refer to the same block table often appear several times in the same image. By the way, as a method of determining a priority entry when a plurality of main scanning positions are stored in this way, the acquired main scanning position X and the main scanning positions stored in the second index table (for example, X ₀₀ , X ₀₁ ,...) Is calculated as an average value and an entry having the smallest average value is preferentially referred to.

  As described above, in the present embodiment, the position of the pixel on the input image is also managed in the second conversion table obtained by sparsely arranging the lattice points of specific portions of the first conversion table. This makes it possible to efficiently determine whether to use the first conversion table or the second conversion table.

In the first and second embodiments described so far, the second conversion is performed by sparsely arranging the lattice points for a specific portion of the first conversion table held in the first conversion table holding unit 50. It was held in the second conversion table holding unit 61 as a table. However, if the capacity of the second conversion table holding unit 61 allows, the entire first conversion table may be stored as a second conversion table in which grid points are sparsely arranged.
Further, the first conversion table holding unit 50 may hold only information on a portion that cannot be supported by the second conversion table, instead of holding information on the entire predetermined color space.
In the above description, the sparsely arranged grid points in the first conversion table are used as the second conversion table. However, information obtained by reducing the amount of information in the first conversion table by various methods is used as the second conversion table. It can also be captured.

By the way, in the first and second embodiments, whether the first conversion table should be used with reference to the second index table stored in the second conversion table holding unit 61 as a part of the second conversion table. Although it is determined whether or not the two conversion table should be used, various other determination methods are conceivable.
For example, the information corresponding to the second index table may be held by the control unit 70 and determined based on the information. Also, whether to refer to the first conversion table or the second conversion table may be input from the outside, stored in a register, for example, and determined based on this. Alternatively, based on the input pixel value, it is possible to determine whether the pixel value exists in a subspace having a linear tendency or in a subspace having a non-linear tendency by calculation, and can determine based on the result. . Furthermore, a method may be considered in which a tag is used for the input pixel value and the first conversion table is determined to be referred to if it is “0” and the second conversion table is referred to if it is “1”.
In the end, whether to use the first conversion table or the second conversion table depends on the first conversion table holding unit 50 (large capacity, low speed operation, low cost) and the second conversion table holding unit 61 (small capacity, high speed operation, It is a trade-off of the nature of (high cost) and may be determined arbitrarily.

In the present embodiment, the temporary storage unit 60 is provided to cache the information once referred to in the first conversion table. However, a part of the temporary storage unit 60 is used to cache the information once referred to in the second conversion table. May be assigned.
Further, if the reading speed of information from the first conversion table holding unit 50 is not a problem, the temporary storage unit 60 may be omitted.

Furthermore, in the above description, information in a table format is assumed as information to be referred to when performing color conversion. However, the present invention can be applied to color conversion based on any other information such as a neural network. It is.
Furthermore, in the above description, the core part of the present embodiment is realized by an LSI, but the same function can also be realized by a computer program. That is, the CPU (not shown) of the computer reads each program that realizes the functions of the index conversion unit 20, the delta conversion unit 30, the control unit 70, the interpolation processing unit 80, etc. from the magnetic disk device (not shown) into the main memory. May be realized.

It is the block diagram which showed the structure of the color conversion apparatus in embodiment of this invention. It is a figure for demonstrating the division | segmentation of the color space in embodiment of this invention. It is a figure for demonstrating the division | segmentation of the color space in embodiment of this invention. It is a figure for demonstrating the division | segmentation of the color space in embodiment of this invention. It is the figure which showed an example of the method of arranging each lattice point of RGB color space in one dimension in embodiment of this invention. It is the figure which showed an example of the response | compatibility with the content of the 1st conversion table in embodiment of this invention, and the lattice point in RGB color space. It is the figure which showed an example of the method of arranging each lattice point of the block cut out from the RGB color space in one embodiment in one dimension. It is the figure which showed an example of the content of the 1st index table in embodiment of this invention, and an example of a response | compatibility with the content of the 1st block table, and the grid point in RGB color space. It is the figure which showed an example of the method which arranges each lattice point of the specific block in RGB color space in embodiment of this invention in one dimension. It is the figure which showed an example of the content of the 2nd index table in embodiment of this invention, and an example of a response | compatibility with the content of the 2nd block table, and the grid point in RGB color space. It is the flowchart which showed the operation | movement of the control part in embodiment of this invention. It is the flowchart which showed the operation | movement of the interpolation process part in embodiment of this invention. It is a figure for demonstrating specifically the interpolation process in embodiment of this invention. It is the figure which showed an example of the content of the 2nd index table in the 2nd Embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 ... Pixel input part, 20 ... Index conversion part, 30 ... Delta conversion part, 40 ... Input buffer, 50 ... 1st conversion table holding part, 60 ... Temporary storage part, 61 ... 2nd conversion table holding part, 70 ... Control , 80... Interpolation processing unit, 90... Pixel output unit

Claims (9)

A color conversion device that performs conversion from a pixel value expressed in a first color space to a pixel value expressed in a second color space,
Storage means for storing second information obtained by reducing the amount of information of the first information referred to in the conversion;
Whether to refer to the first information or the second information to convert the first pixel value expressed in the first color space is determined based on the first pixel value. And a color conversion device. The first information is a first conversion table indicating a predetermined number of correspondences between pixel values expressed in the first color space and pixel values expressed in the second color space;
The second information is a second transformation that indicates a correspondence relationship between the pixel value expressed in the first color space and the pixel value expressed in the second color space that is smaller than the predetermined number. The color conversion apparatus according to claim 1, wherein the color conversion apparatus is a table.   The processing means determines whether to refer to the first information or the second information based on a position of the first pixel value in the first color space. The color conversion device according to claim 1.   The processing means determines to refer to the first information when the tendency of the change in the corresponding pixel value in the second color space with respect to the change in the first pixel value is nonlinear, and the tendency 2. The color conversion apparatus according to claim 1, wherein the second information is determined to be referred to when the second information is linear.   When the processing means converts the pixel value of a specific pixel in a predetermined image with reference to the second information, the processing means is located in the vicinity of the specific pixel on the image in the second information. 2. The color conversion apparatus according to claim 1, wherein a part necessary for conversion of a pixel value of another pixel is preferentially referred to. A color conversion method for performing conversion from a pixel value expressed in a first color space to a pixel value expressed in a second color space,
Receiving a first pixel value expressed in the first color space;
When the received first pixel value does not satisfy a predetermined criterion, the pixel value in the second color space obtained using the first information is output, and the first pixel value is Outputting the pixel value in the second color space obtained by using the second information obtained by reducing the information amount of the first information if the criterion is satisfied. A color conversion method characterized by that.   In the outputting step, the pixel value obtained by using the first information when the tendency of the change of the corresponding pixel value in the second color space with respect to the change of the first pixel value is nonlinear 7. The color conversion method according to claim 6, further comprising: outputting a pixel value obtained by using the second information when the tendency is linear. On the computer,
A function of accepting a first pixel value expressed in a first color space;
When the received first pixel value does not satisfy a predetermined criterion, the pixel value in the second color space obtained using the first information is output, and the first pixel value is When the criterion is satisfied, a function of outputting the pixel value in the second color space obtained by using the second information obtained by reducing the information amount of the first information is realized. Program to let you.   In the output function, the pixel value obtained using the first information when the tendency of the change of the corresponding pixel value in the second color space with respect to the change of the first pixel value is nonlinear The program according to claim 8, wherein when the tendency is linear, the pixel value obtained by using the second information is output.

Images