JP2004112696A - Image processing device and method therefor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To prevent inversion of color component ratio before and after quantization in quantizing processing of a multi-level color image. <P>SOLUTION: When quantizing the color image, depending on an input pixel value, threshold is varied so as to make average quantization error cancelled, using a varied threshold table 202. In the varied threshold table 202, the average value of the average quantization error for each color component, when all the values in the varied threshold table set to zeroes are registered as indices for the sum of each color component. The varied threshold table 202 outputs the registered value, depending on the sum of magenta and cyan to be input, to which a random noise is added properly. This resultant value is added to each color component of the pixel after error diffusion, and quantization is performed. <P>COPYRIGHT: (C)2004,JPO

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an image processing method and an apparatus therefor, and more particularly to image processing for quantizing input data into binary or multi-value data while storing a difference between input image density and output image density by an error diffusion method. is there.
[0002]
[Prior art]
Conventionally, an error diffusion method is known as pseudo-halftone processing for expressing multi-value data in binary (see Non-Patent Document 1). In this method, if the pixel of interest is P, the density thereof is v, and the densities of peripheral pixels P0, P1, P2, and P3 around the point P are v0, v1, v2, v3, and the threshold for binarization is T, The binarization error E at the point of interest P is distributed to the peripheral pixels P0, P1, P2, and P3 by weight coefficients W0, W1, W2, and W3 obtained empirically, and the average density is made macroscopically equal to the density of the original image. Is the method. At this time, if the output binary data is o
If v ≧ T, o = 1, E = v−Vmax;
If v <T, o = 0, E = v−Vmin;
(However, Vmax: maximum density, Vmin: minimum density)
v0 = v0 + E × W0; (Formula 2)
v1 = v1 + E × W1; (Formula 3)
v2 = v2 + E × W2; (Formula 4)
v3 = v3 + E × W3; (Formula 5)
It becomes. Here, as examples of weighting factors, W0 = 7/16, W1 = 1/16, W2 = 5/16, W3 = 3/16, and the like are set.
[0003]
With this method, when pseudo-halftone processing is performed on multiple colors, for example, cyan, magenta, yellow, and black, which are used in color ink jet printers, etc., each color is processed independently using the error diffusion method described above. In addition, when one color is viewed, even if the visual characteristics are excellent, if two or more colors overlap, good visual characteristics cannot always be obtained.
[0004]
In order to improve this defect, a pseudo halftone processing method is disclosed in which good visual characteristics can be obtained even when two or more colors overlap by using an error diffusion method combining two or more colors (Patent Document). 1 and Patent Document 2). However, in these methods, the threshold value for binarization is fixed for each color, and the case where the threshold value is different for each color cannot be handled.
[0005]
On the other hand, a method is known in which the quantization result of each color is set in advance in a quantization value table, and the quantization result of each color is obtained by referring to the quantization value table based on the input pixel value of each color in the target pixel. According to this method, an error diffusion method combining two or more colors can be performed while allowing the number of quantization levels and the threshold level to be set by data set in a quantization value table (lookup table). it can.
[0006]
On the other hand, in the error diffusion method in which the diffusion coefficient is fixed, there is a problem that dots are not uniformly dispersed in the highlight and shadow portions and are generated in a chain. On the other hand, an error diffusion method using a different diffusion coefficient depending on the input gradation has been proposed (see Patent Document 3). In this method, a first diffusion coefficient in highlights and shadows and a second diffusion coefficient in intermediate gradations are defined, and the areas from highlights to intermediate gradations and intermediate gradations to shadows are defined as the two. A linearly interpolated diffusion coefficient is used.
[0007]
Further, the conventional image processing apparatus using the error diffusion method has a problem that the dot generation at the rising portion of the low density region is significantly delayed. Furthermore, at a specific input gradation, a regular pattern is generated in the output binary image, which may deteriorate the image quality. In order to solve the above problem, an error diffusion method has been proposed in which a threshold is set based on the gradation value of multi-value image data (see Patent Document 4). An outline of this method is as follows. In an image processing apparatus that converts multi-gradation image data into a first gradation value and a second gradation value using an error diffusion method or an average error minimum method, the multi-gradation image data of a pixel of interest. A reference threshold is set on the basis of the tone value, and a random value is added to the reference threshold to set a threshold for binarization. According to this method, the dot generation delay is eliminated by optimizing the threshold value according to the gradation value of the multi-tone image data of the target pixel, and further, the threshold value is set according to the gradation value of the target pixel. Generation of the regular pattern is prevented by applying noise.
[0008]
In addition, by selecting parameters for diffusion coefficient correction and threshold correction according to the input value, even more precise dot placement and avoidance of dot generation delays at rising positions in low density areas A method for performing the above is proposed in Japanese Patent Application No. 2001-295127.
[0009]
Further, there is a method of simultaneously performing a diffusion coefficient or threshold value correction process according to an input value and an error diffusion method by combining two or more colors. For example, FIG. 8 shows a block diagram when cyan / magenta 8-bit multilevel image data is input and binary data is output by quantization processing. Note that “X” included in the character string in the text and drawings represents all the characters representing the color component (C for cyan and M for magenta). For example, inValX indicates inValC and inValM. “=” Indicates that the content of the right side is assigned to the left side.
[0010]
In the figure, inValC and inValM are cyan and magenta input pixel values sent from an input image (not shown), and are respectively supplied to adders 801, 802, error calculators 803, 804, and threshold variation setting devices 805, 806. Entered. In the adders 801 and 802, error calculation is performed on lValueC and lValueM obtained by adding the input pixel values inValC and inValM and the errors errC and errM from the quantized peripheral pixels output from the error calculators 803 and 804, respectively. The data is output to the devices 803 and 804 and the output value setting device 807.
[0011]
The threshold fluctuation amount setting devices 805 and 806 output a threshold fluctuation amount fractThX for changing the quantization threshold based on the input pixel value inValX. The output value setting device 807 compares the pixel value lValueX after error diffusion with a threshold value that is changed by the threshold value fluctuation amount fractThX, and outputs an output pixel value binDataX.
[0012]
FIG. 9 is a block diagram showing the configuration of the threshold value fluctuation amount setting devices 805 and 806. The fluctuation threshold table 901 is referred to by the value of inValX, and the reference fluctuation amount vthX registered corresponding to the value is output to the adding device 903. The contents of the fluctuation range value table are set statically in advance. The reference variation amount is set so that the variation amount of the threshold value increases because the occurrence of dots is less likely as the input value is farther than the threshold value. This is because the dot generation is not delayed.
[0013]
The noise matrix 902 is referred to by an address created based on the coordinates of the input pixel, and outputs a noise value noiseX corresponding to the address to the adder. Noise is a value used to reduce the regularity of dot arrangement that appears by error diffusion processing, and the contents of the noise matrix are given in advance for each pixel value. For example, the noise is given a completely random value or is determined for each gradation level.
[0014]
The adder 903 adds the reference fluctuation amount vthX and the noise noiseX, and outputs the threshold fluctuation amount fractThX to the output value setting device 807.
[0015]
FIG. 10 is a block diagram illustrating the configuration of the output setting device 807. The subtraction devices 1001 and 1002 output lValueC ′ and lValueM ′, which are obtained by subtracting the outputs fractThC and fractThM from the threshold value fluctuation amount setting devices 805 and 806, respectively, from the lValueC and lValueM to the output value table 1003.
[0016]
FIG. 7 shows an output value table 1003. The output value table 1003 is referred to by lValueC ′ and lValueM ′, and the quantized values binDataC and binDataM are output. In addition, the value of each area | region has shown (binDataC, binDataM). The output value table 1003 includes a first area where the total value of magenta and cyan pixel values is 128 or less, a second area where the total value is greater than 255 + 128, a total value greater than 128 and (255 + 128) or less. And a third area where the magenta pixel value is equal to or greater than the cyan pixel value, and a fourth area where the total value is greater than 128 and equal to or less than (255 + 128), and the magenta pixel value is less than the cyan pixel value. It is divided into and. Then, (0, 0), (1, 1), (0, 1), and (1, 0) are registered in order as output values for these four areas. Here, the value on the boundary of the region may belong to any region.
[0017]
For example, when coordinates 702 (white circles in the figure) are referred to by pixel values lValueC ′ and lValueM ′ as shown in FIG. 7, 0 and 1 are output as binDataC and binDataM, respectively. Note that the coordinates 701 are coordinates when the output value table 1003 is referred to by lValueC and lValueM before the threshold is changed. Since the value of the average quantization error for each gradation level is set as the reference variation value, the value of each color component approaches the median value by adding the variation value. Note that the threshold fluctuation amount does not fluctuate the threshold value but fluctuates the pixel value. This is because the relationship between the pixel value and the threshold value is relative.
[0018]
In FIG. 10, output values binDataC and binDataM are output as quantization results to an output image (not shown) and output to error calculation devices 803 and 804, respectively.
[0019]
FIG. 11 is a block diagram illustrating the configuration of the error calculation devices 803 and 804. First, the center value table 1102 is referred to using the quantized value binDataX obtained from the output setting device 807 as an address, and the obtained center value stdX is output to the subtraction device 1103. Since the quantized value binData is binary data, when binDataX is 0, stdX is 0, and when binDataX is 1, stdX is 255. The subtractor 1103 outputs an error errorX obtained by subtracting the center value stdX from the outputs lValueX from the adders 801 and 802 to the error dispersion device 1104. On the other hand, in the distribution coefficient table 1101, distribution coefficients difX 0, difX 1, difX 2, and difX 3 obtained using the input pixel value inValX as an address are sent to the error distribution device 1104. FIG. 6 is a diagram illustrating the positional relationship of the distribution coefficients difX0, difX1, difX2, and difX3 with respect to the pixel of interest 601. The error distribution apparatus performs the following calculations in order.
[0020]
errX3 = errX2 + errorX * difX3; (Formula 6)
errX2 = errX1 + errorX * difX2; (Formula 7)
errX1 = errorX * difX1; (Formula 8)
errX0 = errorX * difX0; (Equation 9).
[0021]
Here, errX0, errX1, errX2, and errX3 are variables stored in a storage device (not shown) in the error distribution apparatus 1104, and are all initialized at the start of printing. The error dispersion device 1104 outputs errX3 obtained by Equation 6 to the error buffer 1105, and errX0 obtained by Equation 9 to the adder 1106. The error buffer 1105 stores errX3 at an address corresponding to the coordinates of the input pixel, and outputs the error errXb from the previous raster to the adder 1106.
[0022]
If the coordinates of the input pixel in the main scanning direction are hPos and the head address of the error buffer is errBufX,
errBufX [hPos-2] = errX3; (Formula 10)
errXb = errBufX [hPos]; (Formula 11)
It becomes. Here, errBufX [A] = B indicates that the value B is assigned to the Ath address from the top of the error buffer.
[0023]
The adder 1106 adds errX0 obtained from the error dispersion device 1104 and errXb obtained from the error buffer 1105, and outputs an error errX.
[0024]
The errors errC and errM output from the error calculators 803 and 804 are sent to the adders 801 and 802, respectively, and added to the next input pixel values inValC and inValM, and the process proceeds to the next pixel.
[0025]
With the above configuration, the processing for the input data for one pixel is performed, so that the quantization processing is performed on the entire image by repeating the above processing while shifting the pixel by one pixel in the main scanning direction.
[0026]
[Non-Patent Document 1]
“An Adaptive Algorithm for Spatial Gray Scale” in society for Information Display 1975, Symposium Digest of
Technical Papers, 1975, 36
[Patent Document 1]
JP-A-8-279920
[Patent Document 2]
Japanese Patent Laid-Open No. 11-10918
[Patent Document 3]
JP-A-10-31458
[Patent Document 4]
JP-A-8-307680
[Problems to be solved by the invention]
In the prior art as described above, the dot arrangement of each color in the highlight can be adjusted by determining the error distribution coefficient of each color according to the input value. However, since the threshold fluctuation amount is determined independently for each color, there is a possibility that an incorrect dot may occur when there is an extreme difference in the input pixel value of each color of the pixel of interest.
[0027]
An example of illegal dot generation when the output of both cyan and magenta is binary will be described with reference to FIG. The numbers in parentheses indicate (cyan output value, magenta output value). The thick bold line at the center is an independent threshold value for each color, and as the input value is farther than the threshold value, dots are less likely to be generated. Therefore, the variation threshold value table 902 for each color is set so that the variation amount of the threshold value increases. When the cyan / magenta error-corrected input value is coordinate 701, the cyan input value is close to the threshold value (C = 128 vertical line), so the threshold value does not vary much (the arrow to the right is short). However, since the input value of magenta is far from the threshold value (M = 128 horizontal line), the threshold value is greatly reduced (the upward arrow is long). Therefore, the coordinates 702 are referred to in the output value table at this time, the cyan output is 0, and the magenta output is 1. When this is compared with the original pixel value, it can be seen that the magnitude relationship between the input value and the output value of cyan and magenta is reversed.
[0028]
As described above, when attention is paid to a certain pixel, the ratio of the color component in the original multi-valued image may be completely different from the ratio of the color component in the quantized image. The fact that the ratio of the color components is different from the original one causes a so-called false color or false contour, which causes the image quality to deteriorate.
[0029]
The present invention has been made in view of the above-described conventional example. A multi-dimensional LUT corresponding to a multi-value image signal is used to select a threshold fluctuation amount corresponding to a combination of each color component for each pixel and combine two or more colors. In addition to the error diffusion method, the threshold fluctuation amount for preventing the delay of dot generation is determined based on the combination of the values of the respective color components, so that substantially the same hue is obtained with respect to the original multi-valued image. An object of the present invention is to provide an image processing method and apparatus capable of generating an output image.
[0030]
[Means for Solving the Problems]
In order to achieve the above object, the present invention comprises the following arrangement.
[0031]
An image processing apparatus for quantizing multi-valued color image data including a plurality of color components,
For each of a plurality of color components included in the pixel of interest, an adding means for adding a quantization error value distributed for each color component from neighboring pixels;
Variation amount determining means for determining a threshold variation amount of each color based on the sum of each color component value of the target pixel;
The sum of the values of the color components to which the error values are added by the adding means corresponds to the sum of the values after the conversion, and the sum of the pixel values after the conversion is distributed according to the ratio of the color components before the conversion. Quantizing means for determining the value of each color component after conversion,
The quantization unit varies the quantization threshold of each color component according to the threshold fluctuation amount when determining the value of each color component after conversion.
[0032]
More preferably, when the quantization unit determines the value of each color component after conversion, the quantization unit adds the threshold fluctuation amount to the value of each color component of the target pixel, thereby determining the quantization threshold of each color component. Fluctuate.
[0033]
More preferably, the number of gradation levels quantized by the quantization means is two or more.
[0034]
More preferably, the variation amount determining means has a table set to correct an average quantization error in each density gradation for each of the plurality of color components by an equal value for each of the plurality of color components. Then, a value obtained by referring to the table is determined as the threshold fluctuation amount by the sum of the values of the respective color components of the target pixel output by the adding means.
[0035]
More preferably, the variation amount determining means determines a threshold variation amount of each color based on a combination of each color component value of the target pixel instead of the sum of the color component values of the target pixel.
[0036]
More preferably, the variation amount determining means has a table set so as to offset an average quantization error in each density gradation for each of the plurality of color components for each of the plurality of color components, and the addition A value obtained by referring to the table by a combination of values of each color component of the target pixel output by the means is determined as the threshold fluctuation amount.
[0037]
More preferably, the threshold variation amount so as to further cancel out the second average quantization error when the pixel of interest is quantized using the threshold variation amount determined by the variation amount determination unit for each color component. Further, a correction means for correcting the above is provided.
[0038]
More preferably, the correction means uses the table set so as to cancel out the second average quantization error in each density gradation for each of the plurality of color components for each of the plurality of color components. The threshold fluctuation amount is corrected.
[0039]
More preferably, the table is configured to be referred to by the number of gradations obtained by thinning out the number of gradations of the input pixel.
[0040]
DETAILED DESCRIPTION OF THE INVENTION
<First embodiment>
Hereinafter, a first embodiment of the present invention will be described in detail with reference to the drawings.
[0041]
<Configuration of image processing apparatus>
FIG. 1 is a block diagram illustrating the configuration of an image processing apparatus according to an embodiment of the present invention. This embodiment is an image processing apparatus that inputs 8-bit multi-value image data of cyan / magenta and outputs binary data by quantization processing. Specifically, this apparatus is realized by executing a program by a computer, for example. The computer is, for example, an embedded computer for controlling an ink jet printer. In that case, the configuration shown in FIGS. 1 to 5 is realized by the processor executing the program loaded in the memory of the computer. The procedure of the program is described below. The image data generated by this configuration is output as a hard copy by the printer engine. Hereinafter, this embodiment will be described.
[0042]
“X” included in the character string in the text and drawings represents all the characters representing the color components (C for cyan and M for magenta). For example, inValX indicates inValC and inValM. “=” Indicates that the content of the right side is assigned to the left side.
[0043]
In FIG. 1, inValC and inValM represent 8-bit multi-valued input pixel values of cyan and magenta, respectively, which are added from an input image (not shown), adders 101 and 102, error calculators 103 and 104, and threshold fluctuations, respectively. The amount is input to the amount setting unit 105.
[0044]
The addition units 101 and 102 respectively add lValueC and lValueM obtained by adding the input pixel values inValC and inValM and the errors errC and errM from the quantized peripheral pixels output from the error calculation units 103 and 104, respectively. The data are output to the error calculation units 103 and 104 and the output value setting unit 106, respectively.
[0045]
The threshold variation setting unit 105 outputs a threshold variation fractThX that varies the quantization threshold based on the input pixel value inValX.
[0046]
The output value setting unit 106 compares the pixel value lValueX after error diffusion with a threshold value that is changed by the threshold value change amount factThX, and outputs an output pixel value binDataX. However, the relationship between the threshold value and the pixel value is relative, and in this embodiment, the threshold value used by the output value setting unit 106 is not simple for each color component, and the output value is determined by two color components. In the present embodiment, the pixel value is changed without changing the threshold value. Therefore, as will be described later, the threshold value fluctuation amount flactThX is added to the pixel value to obtain an output value from the output value table. Of course, the sign is inverted when it is reflected in the pixel value compared to the case where it is reflected in the threshold value.
[0047]
<Threshold variation setting unit>
FIG. 2 is a block diagram illustrating the configuration of the threshold fluctuation amount setting unit 105. The input pixel values inValC and inValM are added by the adding unit 201 and input to the variation threshold value table 202.
[0048]
(Variation threshold table)
The variation threshold value table 202 outputs the reference variation value vthCM, which is referred to using the addition result inValC + inValM as an address, to the addition units 205 and 206. As the reference fluctuation value set in the fluctuation threshold table 202, for example, a value based on the average quantization error is set. Here, the average quantization error is a value for the addition result inValC + inValM of the input pixel value. The fluctuation threshold table 202 is set as follows, for example.
[0049]
First, the reference variation value vthCM is set to 0 with respect to all inValC + inValM, and 256 × 256 types of pixels formed by combinations of 256-tone cyan input pixel value inValC and 256-tone magenta input pixel value inValM. For a value, we try to quantize a uniform image of that value, for example 512 × 512 pixels. The output value table used for quantization at that time is shown in FIG. 4, and a first area where the sum of magenta and cyan pixel values is 128 or less, and a second area where the sum is greater than 255 + 128. A third region where the total value is greater than 128 and less than or equal to (255 + 128) and the magenta pixel value is greater than or equal to the cyan pixel value, and the total value is greater than 128 and less than (255 + 128) and magenta Are divided into a fourth area smaller than the cyan pixel value. Then, (0, 0), (1, 1), (0, 1), and (1, 0) are registered in order as output values for these four areas. Then, according to (inValC ′, inValM ′), any one of (0, 0), (1, 1), (0, 1), (1, 0) is output according to the table of FIG.
[0050]
Error diffusion is performed using this table, and an average quantization error is obtained for each value of the sum inValC + inValM of the cyan input pixel value and the magenta input pixel value. A value obtained by inverting the sign of the obtained average error value is registered in the fluctuation threshold table. The setting method of the fluctuation threshold table will be described later in detail. In this way, values that cancel the average quantization error are set in the variation threshold table. In this embodiment, the sum of the average quantization error of each color component is canceled by changing the threshold value by the same value for each color component.
[0051]
As is clear from the configuration of FIG. 2, the reference variation value vthCM output using (inValC ′ + inValM ′) as an index is commonly used for cyan and magenta. Therefore, each color component varies by the same amount. Therefore, when the variation of the pixel value is viewed on FIG. 4, the pixel value after the variation of the threshold is not considered in consideration of noise added later with respect to the original pixel value (inValC ′, inValM ′) (inValC '+ VthCM, inValM' + vthCM). In other words, by adding the reference variation value, the pixel value becomes the boundary line between the third area and the fourth area in FIG. 4 (the boundary line between the output pixel values (0, 1) and (1, 0)). Therefore, the ratio of cyan and magenta components can be prevented from being reversed before and after quantization. Thereby, generation | occurrence | production of a false color can be prevented.
[0052]
(Noise matrix)
In the noise matrices 203 and 204, noise C and noise M are referred to by addresses corresponding to coordinates of input pixels (not shown), and output to the adders 205 and 206, respectively.
[0053]
Adders 205 and 206 add noiseC and noiseM output from noise matrices 203 and 204 and vthCM output from fluctuation threshold table 202, respectively, and output fractThC and fractThM to output setting unit 106. When quantization is performed by the error diffusion method, a regular pattern tends to be easily generated at pixel values of gradations near the threshold value. Such a regular pattern generates a texture that does not exist in the original image, leading to a reduction in image quality. Therefore, in this embodiment, the amount of noise applied to the threshold for quantization is set in order to reduce the regular pattern without sacrificing the image quality of the output binary image.
[0054]
The noise matrix is given as follows, for example. That is, for a certain pixel area, for example, a 16 × 16 pixel area, a noise value is randomly set according to the pixel position in the area. As the value range, a method of selecting at random within a certain range, for example, a range of −10 to 10 or a method of selecting so that the distribution of noise values follows a normal distribution can be considered.
[0055]
In addition, there is a method in which only a sign is randomly given according to the pixel position, an absolute value of the noise value is given according to the pixel value, and a value obtained by adding a randomly given sign to the absolute value is used as the noise value. Conceivable. In this method, the absolute value of noise is set so as to increase for density gradations where a regular pattern is likely to occur, and to give a smaller value for density gradations that do not. The ease of occurrence of a regular pattern can be evaluated by quantizing an image having a uniform density gradation with a noise value of 0 for each density gradation of 0 to 255 and using the output of the configuration shown in FIG. . The evaluation is performed by, for example, performing orthogonal transformation on the quantized image (each pixel is 0 or 255) to convert it to the frequency domain, and filtering the high frequency component that is difficult to be seen by humans (that is, less likely to be a defect) and reverses it. Convert and return to image. In this state, the dispersion of pixel values is obtained, and the value is used as an index of the texture included in the image. Then, with respect to an image having a uniform gray level of interest, the noise value is quantized while increasing from 0 to 1, and the variance of the pixel values is obtained, and the noise value at which the value becomes the smallest is obtained. Determined as the noise value for the key.
[0056]
<Output setting section>
FIG. 3 is a block diagram illustrating the configuration of the output setting unit 106. The subtraction units 301 and 302 output lValueC ′ and lValueM ′, which are obtained by subtracting the outputs fractThC and fractThM from the threshold variation setting unit 105 from the outputs lValueC and lValueM from the addition units 101 and 102, respectively, to the output value table 303.
[0057]
FIG. 4 is a diagram showing the output value table 303. The output value table 303 is referred to by lValueC ′ and lValueM ′, and the quantized values binDataC and binDataM are output. The value of each area of the output value table 303 indicates (binDataC, binDataM). For example, when ◯ 401 is referred to by lValueC ′ and lValueM ′ as shown in FIG. 4, 1 and 0 are output as binDataC and binDataM, respectively. binDataC and binDataM are output as quantized results to an output image (not shown) and also output to the error calculators 103 and 104, respectively.
[0058]
<Error diffusion part>
FIG. 5 is a block diagram for explaining the configuration of the error calculators 103 and 104. First, the center value stdX obtained by referring to the center value table 502 is output to the subtraction unit 503 using the quantized value binDataX obtained from the output setting unit 106 as an address. In this embodiment, since binarization processing is performed, when binDataX is 0, stdX is 0, and when binDataX is 1, stdX is 255. The subtraction unit 503 outputs the error errorX obtained by subtracting the center value stdX from the output lValueX from the addition units 101 and 102 to the error dispersion unit 504. On the other hand, in the distribution coefficient table 501, the distribution coefficients difX0, difX1, difX2, and difX3 obtained using the input pixel value inValX as an address are sent to the error distribution unit 504.
[0059]
FIG. 6 is a diagram illustrating the positional relationship of the distribution coefficients difX0, difX1, difX2, and difX3 with respect to the pixel of interest 601. The error distribution unit performs the following calculation.
[0060]
errX3 = errX2 + errorX * difX3 (Formula 12)
errX2 = errX1 + errorX * difX2 (Formula 13)
errX1 = errorX * difX1 (Formula 14)
errX0 = errorX * difX0 (Formula 15).
[0061]
Here, errX0, errX1, errX2, and errX3 are variables stored in a storage unit (not shown) in the error distribution unit 504, and are all initialized at the start of printing. The error variance unit 504 outputs errX3 obtained by Expression 12 to the error buffer 505 and errX0 obtained by Expression 15 to the adder 506. The error buffer 505 stores errX3 at an address corresponding to the coordinates of the input pixel and outputs the error errXb from the previous raster to the adder 506.
[0062]
In this embodiment, when the coordinate of the input pixel in the main scanning direction is hPos and the head address of the error buffer is errBufX,
errBufX [hPos-2] = errX3 (Formula 16)
errXb = errBufX [hPos] (Formula 17)
It becomes. Note that errBufX [A] = B indicates that the value B is assigned to the Ath address from the top of the error buffer.
[0063]
The adder 506 adds errX0 obtained from the error distributor 504 and errXb obtained from the error buffer 505, sends the error errX to the adders 101 and 102, and is added to the next input pixel value inValX.
[0064]
For the distribution coefficients difX0, difX1, difX2, and difX3, for example, values such as 7/16, 1/16, 5/16, and 3/16 may be fixedly used in order, but the input pixel value inValX It can also be determined according to the value. In this case, an optimal distribution coefficient is determined in advance by trial and error while changing the distribution coefficient in accordance with the input pixel value.
[0065]
With the above configuration, processing for input data for one pixel is performed. By repeating the above processing by shifting the target pixel by one in the main scanning direction, the quantization processing is performed on the entire image.
[0066]
The image obtained in this manner is obtained by using the error diffusion method by combining two or more colors by the function of the threshold value fluctuation amount setting unit which is a characteristic configuration of the present embodiment. By varying the quantization threshold with the sum, it is possible to prevent the occurrence of false colors that do not appear in the original image. Thereby, generation | occurrence | production of the pseudo contour in all the colors combined with the pseudo contour in a single color can be prevented simultaneously. Further, by determining the error distribution coefficient of each color according to the input value, it is possible to simultaneously arrange the dot arrangement in the highlight.
[0067]
(A more detailed explanation of how to set the fluctuation threshold table)
Although the method for creating the fluctuation range value table has been briefly described, it will be described in more detail here.
[0068]
When the image of uniform color of 512 pixels vertically and horizontally shown in FIG. 14 is subjected to binary quantization processing without performing threshold correction processing in this embodiment, it occurs when each pixel is calculated. The error value was measured for each pixel. Further, the cyan value was averaged for the area (black and white area of 256 pixels) displayed in black in FIG. Hereinafter, this value is referred to as an error average value. The same measurement is performed for magenta.
[0069]
FIG. 15A is a diagram illustrating a measurement result of the error average value measured under the above-described conditions while setting different values for the cyan and magenta colors for the image subjected to the binary quantization process. . In FIG. 15A, as shown in FIG. 15B, the average error value for the cyan and magenta input values is plotted with respect to the vertical axis.
[0070]
This error average value is obtained when a quantization process that does not perform a correction process (addition of a threshold fluctuation value) is performed on the threshold value, such as a dot generation delay at the rising portion of the low density area described above. The characteristics are shown.
[0071]
More specifically, when the error average value is large (plus side, that is, when plotted above in FIG. 15), it indicates that dots are stably formed in a state where the plus error is accumulated. Yes. For example, consider a case where the threshold for determining whether or not to form dots in binary quantization is 127 and the input value is 10 (the input value is 255 at the maximum). In this case, a dot is formed when a value obtained by adding an error distributed from surrounding pixels in addition to an input value exceeds a threshold value, that is, 127. That is, no dot is formed until such a positive error is accumulated. That is, a dot formation delay occurs (so-called sweeping phenomenon). In such a case, the average error value is a large number (that is, a positive value and a large absolute value).
[0072]
When the average error value is small (a negative value and a large absolute value, that is, plotted below in FIG. 15), dots are stably formed with a negative error accumulated. Yes. For example, consider a case where the threshold for determining whether or not to form dots in binary quantization is 127 and the input value is 240 (the input value is 255 at the maximum). In this case, even when there is no error distribution from the surroundings, dots are formed and an error of 10 occurs. Even if some minus error is distributed from the surrounding pixels, the input value is much larger than the threshold value, so dots are formed at a high density, and minus errors accumulate.
[0073]
Then, when the value obtained by adding a minus error from surrounding pixels to the input value 240 falls below a threshold value, that is, 127, pixels that do not form dots appear. Then, in a state where the minus error is accumulated up to a certain value, the accumulated minus error is roughly stable, and the formed dot density is constant. In this way, an image defect that can create an area with a high dot density is generally called a discharge phenomenon. In such a case, the average error value is a small number (that is, a negative value and a large absolute value).
[0074]
In view of this, referring to FIG. 15A, in a color with a small magenta input value and a large cyan input value (a condition in which the sum of the magenta input value and the cyan input value is smaller than 127), an error occurs. In the area where the average value is small, that is, the area where the image of the color is quantized, a minus error is accumulated. Further, the error average value is large for a color having a large magenta input value and a small cyan input value (condition that the sum of the magenta input value and the cyan input value is smaller than 127).
[0075]
That is, from the state in which an image having a small magenta input value and a large cyan input value (a condition that the sum of the magenta input value and the cyan input value is smaller than 127) is quantized and negative errors are accumulated. Continuously quantizing a color having a large magenta input value and a small cyan input value (condition that the sum of the magenta input value and the cyan input value is smaller than 127) causes a significant delay in dot formation. .
[0076]
In the present embodiment, a value obtained by inverting the sign of the error average value (roughly shown in FIG. 15) obtained by the method described above is used as the correction parameter of the fluctuation threshold table, that is, the fluctuation threshold table in FIG. Used for 201. More specifically, in the case where the input of the variation threshold table (here, n is assumed for the sake of explanation) is an even number, the above method is used when both the cyan input and the magenta input are n / 2. A value obtained by inverting the sign of the obtained error average value is used. Further, when the input of the variation threshold table is an odd number, the error average obtained by the above method when the input of cyan is (n + 1) / 2 and the input of magenta is (n-1) / 2. A value obtained by averaging the error and the error average value obtained by the above method when the cyan input is (n-1) / 2 and the magenta input is (n + 1) / 2 is used.
[0077]
Thereby, the absolute value of the error average value generated by the quantization process is reduced regardless of the combination of the input values of the two colors. Thereby, dot formation delay, that is, image adverse effects such as sweeping phenomenon and discharging phenomenon are avoided or reduced.
[0078]
<Modification of First Embodiment>
Hereinafter, modifications of the first embodiment of the present invention will be described with reference to the drawings. This modification is different in that the threshold fluctuation amount setting unit 105 shown in FIG. 1 is changed from the configuration shown in FIG. 2 to the configuration shown in FIG. In the present example, the description of the same configuration and processing as those in the first embodiment is omitted.
[0079]
FIG. 12 is a block diagram illustrating the configuration of the threshold fluctuation amount setting unit 105. The input pixel values inValC and inValM are added by the adding unit 1201 and input to the variation threshold value table 1202. The variation threshold value table 1202 outputs values obtained by referring to the addition result inValC + inValM as an address to the output setting unit 106 as fractThC and fractThM.
[0080]
Compared with the configuration of FIG. 2, the noise matrix is omitted. Since the noise matrix is omitted, the removal of the regular pattern has no effect as in the first embodiment, but has the same effect in that the occurrence of false color can be prevented.
[0081]
[Second Embodiment]
FIG. 13 is a block diagram illustrating the configuration of the threshold fluctuation amount setting unit 105 in the second embodiment. The present embodiment has the same configuration as that of the first embodiment, except for the configuration other than the threshold fluctuation amount setting unit 105. Therefore, the description of the same configuration is omitted.
[0082]
In FIG. 13, input pixel values inValC and inValM are input to a fluctuation threshold table 201 and a fluctuation threshold table 202, respectively. The variation threshold value table 1301 and the variation threshold value table 1302 refer to the two-dimensional lookup table using the input pixel values inValC and inValM as addresses, and output vthC and vthM to the adders 205 and 206, respectively. The noise matrices 203 and 204 are referred to as noiseC and noiseM at addresses corresponding to coordinates of input pixels (not shown), and output to the adders 205 and 206, respectively.
[0083]
In the present embodiment, the value obtained by inverting the sign of the error average value (roughly shown in FIG. 15) obtained by the method described above is used as the correction parameter of the fluctuation threshold table, that is, the fluctuation threshold in FIG. Used for the two-dimensional lookup table of the value table 1301. Similarly, the value measured for magenta is used in the two-dimensional lookup table of the fluctuation threshold table 1302 in FIG. Thereby, the absolute value of the error average value generated by the quantization process is reduced regardless of the combination of the input values of the two colors. Thereby, dot formation delay, that is, image adverse effects such as sweeping phenomenon and discharging phenomenon are avoided or reduced.
[0084]
As described above, in this embodiment, two or more colors are combined while selecting a threshold fluctuation amount corresponding to each color combination for each pixel using a multidimensional LUT corresponding to multi-value image signals of two or more colors. Perform error diffusion. In the multi-dimensional LUT for selecting the threshold fluctuation amount, a parameter reflecting a characteristic that an error occurs when an error diffusion method combining two or more colors is performed is set.
[0085]
As a result, it is possible to perform an error diffusion method combining two or more colors using the threshold fluctuation amount selected by the input value while avoiding the phenomenon that the magnitude relationship between the input value and the output value of cyan and magenta is reversed. Thus, even when each color component is observed alone, the delay of dot formation can be effectively prevented.
[0086]
(About diffusion coefficient parameters)
The correction processing of the diffusion coefficient (distribution coefficient) in this embodiment is performed independently for cyan and magenta. As described in Japanese Patent Application No. 2001-295127, the selection of the parameters takes into account the dot arrangement when a monochrome image of cyan and magenta is quantized, and the pattern of the diffusion coefficient used for each input level. Decide. Further, a diffusion coefficient may be selected by quantizing several images for each level (for example, the image of FIG. 14).
[0087]
With the parameters of the diffusion coefficient determined in this way, the dot arrangement was checked for the combination of cyan and magenta in the present embodiment, but it was confirmed in the experiment in this embodiment that there is no significant image detriment. ing.
[0088]
In this way, the selection of the diffusion coefficient for each single color means that the amount of data for selecting the diffusion coefficient is smaller than the configuration in which the diffusion coefficient is selected with a lookup table of two colors (for example, cyan and magenta). There is a merit that can be reduced.
[0089]
[Modification of Second Embodiment]
In the second embodiment, the example in which the threshold fluctuation amount is selected using a two-dimensional lookup table having the same number of gradations as the number of gradations of the input value has been described. As another example of the configuration, in order to reduce the data amount of the lookup table, the input values of the two colors are converted into representative values having a smaller number of gradations, respectively, and then 2 according to the number of gradations of the representative value. A configuration in which a threshold fluctuation amount is selected using a dimension lookup table is conceivable.
[0090]
As an example of the representative value, if the input value takes a value from 0 to 255, it may be converted by a one-dimensional lookup table to obtain a representative value that takes a value in the range from 0 to 63. . In the conversion from the input to the representative value, the input value may be simply divided by 4 in this example.
[0091]
In the data used in the two-dimensional lookup table for selecting the threshold fluctuation amount, the value obtained by averaging the error average values for the combination of input values using the data is used with the sign inverted. To do.
[0092]
The gap of the error average value seen in FIG. 15A exists in the lick direction of the space created by the cyan input value and the magenta input value (a line in which the sum of the cyan input value and the magenta input value is 255). . Of the space created by the cyan input value and the magenta input value, an area where the sum of the minimum cyan input value and the magenta input value of the gap is small is area 0, and an area where the sum of the minimum cyan input value of the gap and the magenta input value is large Will be referred to as area 1.
When the data amount of the two-dimensional lookup table is simply reduced by using the representative value, the combination of the input value of cyan and magenta covered by the representative value of one set of cyan and magenta includes the gap. It may happen that a combination of both sides is included.
In this case, since the value of the variation threshold value to be set differs greatly on both sides of the gap, there is a possibility that the correction effect due to the variation threshold value is greatly reduced or an image defect is caused depending on the value to be set.
[0093]
In order to solve this problem, in this embodiment, two types of lookup table data for area 0 and area 1 are prepared, and whether the input value belongs to either area 0 or area 1 is determined. A configuration is adopted in which a determination is made and the LUT to be used is selected. For a combination of cyan and magenta representative values including a combination of cyan and magenta input values on both sides of the gap, the error average value on the area 0 side of the gap is included in the area 0 two-dimensional lookup table. And a value obtained by averaging and sign-inverting the error average value of the gap on the area 1 side is set in the two-dimensional lookup table for the area 1.
For a combination of cyan and magenta representative values including a combination of cyan and magenta input values on only one side of the gap, the average difference and average of the difference average values are inverted for all combinations of input values included in the representative value. The value is set as the value of the two-dimensional lookup table for both area 0 and area 1.
[0094]
With this configuration, it is possible to select the threshold fluctuation amount with high accuracy even in the combination of input values covered by the combination of the representative values of cyan and magenta, including the combination of the input values of cyan and magenta on both sides of the gap. Can do.
[0095]
When the input value of cyan or magenta is 0 or the maximum, the quantization error average value is zero as can be seen from FIG. That is, in this case, it is desirable that the threshold fluctuation amount is zero. Therefore, in this modification, it is determined whether the input value of cyan or magenta is 0 or the maximum, and in this case, 0 is selected as the threshold fluctuation amount.
[0096]
FIG. 16 shows a flowchart of threshold variation selection in this modification. First, it is determined whether the input value of any color component is 0 or 255 (S1601). If it is 0 or 255, 0 is selected as the threshold fluctuation amount (S1605). Otherwise, it is determined whether the position plotted in FIG. 15 is area 0 across the gap (S1602). If it is area 0, the input value is converted into a representative value with a small number of gradations (see FIG. 15). S1603). For example, 256 gradations are once converted into 64 gradations. Then, using the converted value as an index, a threshold fluctuation amount is obtained from the two-dimensional LUT for area 0 (S1604). On the other hand, if it is area 1, each input value is converted into a representative value with a small number of gradations (S1606). Then, using the converted value as an index, a threshold fluctuation amount is obtained from the two-dimensional LUT for area 1 (S1607). In the LUT for each area, values of the average quantization error shown in FIG. 15A corresponding to area 0 and area 1 are registered.
[0097]
With this configuration, it is possible to select a threshold fluctuation amount with the highest possible accuracy while reducing the data amount of the two-dimensional lookup table.
[0098]
According to the configuration of this modification, an error measured with respect to an image obtained by combining the two colors with a configuration in which the threshold variation amount of each color is set using a multidimensional LUT corresponding to multi-value image signals of two or more colors. The threshold fluctuation amount is set using parameters reflecting the characteristics of occurrence. Accordingly, it is possible to perform a quantization process combining input values of two or more colors while eliminating an adverse effect of an image whose output values are reversed, which is given as a problem to be solved by the invention.
[0099]
[Third Embodiment]
FIG. 17 is a block diagram illustrating the configuration of the threshold fluctuation amount setting unit 105 in the second embodiment. The present embodiment has the same configuration as that of the first embodiment, except for the configuration other than the threshold fluctuation amount setting unit 105. Therefore, the description of the same configuration is omitted.
[0100]
FIG. 17 is a block diagram illustrating the configuration of the threshold fluctuation amount setting unit 105. The input pixel values inValC and inValM are input to the adding unit 201 and the sum of them is output. The signal is input to the fluctuation threshold table 1702, and vthCM is output with reference to the one-dimensional lookup table. Further, the input pixel values inValC and inValM are input to the fluctuation threshold table 1701 and the fluctuation threshold table 1703.
[0101]
In the variation threshold table 1701 and the variation threshold table 1703, first, the number of gradations of the input pixel values inValC and inValM is converted into a quarter value by a one-dimensional lookup table, and the two values are used as an address for two-dimensional lookup. Each of the tables is referred to, and the obtained first threshold fluctuation amounts vthC and vthM are output to the adding units 205 and 206, respectively. The data of each lookup table will be described in detail later. Note that the variation threshold value table 1701 and the variation threshold value table 1703 have the same configuration as that described as the modification of the second embodiment.
[0102]
The fluctuation threshold table 1702 outputs a threshold fluctuation amount based on the sum of cyan and magenta values output from the adding unit 201. Note that the fluctuation threshold table 1702 is the same as the fluctuation threshold table 202 in the first embodiment.
[0103]
The noise matrices 203 and 204 are referred to as noiseC and noiseM at addresses corresponding to coordinates of input pixels (not shown), and output to the adders 205 and 206, respectively.
[0104]
Adders 205 and 206 add noiseC and noiseM output from noise matrices 203 and 204, vthCM output from fluctuation threshold table 1702, and further vthCM and vthM output from fluctuation threshold table 1701 and fluctuation threshold table 1703, respectively. And output to the output setting device 106 as fractThC and fractTM.
[0105]
(Threshold correction characteristics based only on the fluctuation threshold table 1702)
FIG. 18 shows the case where only the threshold correction of the variation threshold table 1702 described above is performed and the threshold correction by the variation threshold table 1701 and the variation threshold table 1703 is not performed (that is, from the variation threshold table 1701 and the variation threshold table 1703). The characteristic of the error average value measured under the condition that the output value is always 0) is shown.
[0106]
FIG. 18A shows the average error value of cyan measured under the above-mentioned conditions while setting different values for cyan and magenta for the color of the image subjected to binary quantization processing. Yes. Further, as shown in FIG. 18B, the error average value in FIG. 18A is plotted with the axes of cyan and magenta input values, and the error average value is plotted along the vertical axis.
[0107]
In the data of FIG. 18A, compared to the characteristics (FIG. 19A) when the threshold value of the fluctuation threshold table 1702 is not corrected (FIG. 19A), the diagonal direction of the space based on the cyan input value and the magenta input value. The gap (what is visible in the center in FIG. 15A) is gone.
[0108]
Further, in the characteristics of FIG. 18A, since the quantization error still slightly varies according to the input pixel value of each color component, an error average value gradient remains. In the present embodiment, it is corrected with correction values based on the fluctuation threshold table 1701 and the fluctuation threshold table 1703. Since the gap shown in FIG. 15A is eliminated, the variation threshold value table 1701 and the variation threshold value table 1703 can easily reduce the number of levels of the input values of the lookup table, that is, make the table small. .
[0109]
This is the merit of the configuration of the present invention. By using the one-dimensional lookup table and the two-dimensional lookup table in combination, the data amount of the two-dimensional lookup table can be reduced.
[0110]
(Setting of fluctuation threshold table 1701 and fluctuation threshold table 1703)
The fluctuation threshold table 1701 uses a value obtained by reversing the sign of the error average value (value shown in FIG. 18A) measured only by correcting the threshold of the fluctuation threshold table 1702. That is, first, as described in the first embodiment, an average quantization error is obtained using an image as shown in FIG. 14 in which each density gradation pixel is uniformly distributed as a sample. In that case, the fluctuation range value table 1702 is used, but the fluctuation range value tables 1701 and 1703 are not used. That is, the image is quantized using the image processing apparatus described in the first embodiment, and an average quantization error for each color component is obtained in relation to the values of the color components cyan and magenta. This is FIG. 19 (A). Then, the sign of this value is inverted to obtain the fluctuation range value tables 1701 and 1703. However, the input values for cyan and magenta are each ¼. For this purpose, for example, for each of cyan and magenta colors, an average quantization error is obtained for each 4i (i is an integer from 0 to 63), and this is used as the fluctuation threshold tables 1701 and 1703. In order to draw this table, the input value is converted. For example, a method of adding a certain value (for example, 0 or 2) to the input value and dividing by 4 and using the quotient can be considered.
[0111]
In this way, the data amount is reduced by setting the number of gradations to ¼ of the input value as described above.
[0112]
This is the same for cyan and magenta, and both measured values are used for the two-dimensional lookup tables of the fluctuation threshold tables 1701 and 1703 in FIG.
[0113]
Thereby, the absolute value of the error average value generated by the quantization process is reduced regardless of the combination of the input values of the two colors. Thereby, dot formation delay, that is, image adverse effects such as sweeping phenomenon and discharging phenomenon are avoided or reduced.
[0114]
(About diffusion coefficient parameters)
The diffusion coefficient correction processing in this embodiment is performed independently for cyan and magenta. As described in Japanese Patent Application No. 2001-295127, the parameters are selected by considering the dot arrangement when a monochrome image of each of cyan and magenta is quantized, and the diffusion coefficient used for each input level. / Decide on a turn. Further, a diffusion coefficient may be selected by quantizing several images for each level (for example, the image of FIG. 7).
[0115]
With the diffusion coefficient parameters determined in this way, the dot arrangement was checked with the combination of the two colors cyan and magenta in the present embodiment. However, in the experiment in the present embodiment, there is no significant image problem. It has been confirmed.
[0116]
In this way, the selection of the diffusion coefficient for each single color means that the amount of data for selecting the diffusion coefficient is smaller than the configuration in which the diffusion coefficient is selected with a lookup table of two colors (for example, cyan and magenta). There is a merit that can be reduced.
[0117]
As described above, according to the configuration of the present embodiment, the threshold variation amount of each color is set using a multidimensional LUT corresponding to multi-value image signals of two or more colors, and the above two colors are combined. The threshold fluctuation amount is set using a parameter reflecting the error occurrence characteristic measured for the image. Accordingly, it is possible to perform a quantization process combining input values of two or more colors while eliminating an adverse effect of an image whose output values are reversed, which is given as a problem to be solved by the invention.
[0118]
<Application of each embodiment>
In each of the above embodiments, the configuration in which the diffusion coefficient is corrected for each of cyan and magenta is shown. However, the present invention is not limited to the configuration. For example, for the input values of cyan and magenta, a diffusion coefficient may be selected using a multidimensional lookup table. In the case of this configuration, it is possible to make the arrangement of dots formed more uniform. However, the method of selecting the diffusion coefficient for each single color in the embodiment has an advantage that the amount of data can be relatively reduced.
[0119]
Although the binary example has been described in the above embodiment, the present invention is not limited to the configuration. The same effect can be obtained when multi-level quantization processing is performed.
[0120]
In the above embodiment, an example has been described in which noise matrix data is referred to by an address corresponding to the coordinates of input pixels of two colors, and noise characteristics are added to the variation amount of the threshold. However, the present invention is not limited to this configuration. The reason why the noise characteristic is added to the fluctuation amount of the threshold value is to avoid or reduce the deterioration of the image quality due to the regular dot arrangement appearing in the output image as a result of the quantization process. The present invention is not limited to this configuration, and the same effect is obtained when no noise characteristic is added to the amount of variation in threshold.
[0121]
In the above-described embodiment, an example in which quantization processing is performed on input pixel values of two colors of cyan and magenta in combination has been described. However, the present invention is not limited to this configuration. For example, it is also effective when processing is performed for combinations of inks of other colors. In addition, there is an effect on a combination of differently conspicuous combinations such as a printing system using relatively dark dots and thin dots, a printing system using relatively large dot diameters, and small dots.
[0122]
Further, combinations of more than two colors, that is, cyan, magenta, black, and the like are also effective. In this case, a three-dimensional lookup table is used instead of the two-dimensional lookup table.
[0123]
In the above-described embodiment, an example in which the output determination unit determines an output value with reference to a multidimensional lookup table has been described. However, the present invention is not limited to this configuration, and is effective in an error diffusion method in which output determination is performed with a combination of two or more colors.
[0124]
Further, in the above-described embodiment, the configuration in which the error variation amount is selected using the input value of the pixel that is subjected to quantization processing by the error diffusion method has been described. However, if the image to be quantized has a characteristic that the input value hardly changes greatly for each pixel, the input value of a neighboring pixel of the pixel to be quantized or It is also effective to select the threshold fluctuation amount using the average value of the input values of these pixels.
[0125]
For example, in order to form an image of a photographic image or the like with a high-resolution recording apparatus, there is a case where quantization processing is performed at a higher resolution than the resolution of the optical system of the camera that inputs the image.
[0126]
Further, by applying the image processing apparatus configured as described above to a printer that quantizes a multi-valued image and forms an image, it is possible to print a high-quality image. In particular, in an ink jet printer that forms an image by placing dots of a certain size for each color component on a medium, the quality of the image formed varies greatly depending on how the dots are arranged. The application of the present invention, which is applied to prevent the delay of dot generation and prevents the occurrence of false colors due to the reversal of the color component ratio of the input pixel and the color component ratio of the output pixel, has a great effect on improving the image quality.
[0127]
【The invention's effect】
As described above, according to the present invention, in an image processing apparatus that uses an error diffusion method by combining two or more colors, the quantization threshold is changed by the sum of the input values of each color in the pixel of interest, so While simultaneously preventing the generation of pseudo contours in all colors combined with the contours, it is possible to simultaneously adjust the dot arrangement in the highlight by determining the error distribution coefficient of each color according to the input value.
[0128]
In the present invention, an error diffusion method in which two or more colors are combined while selecting a threshold fluctuation amount corresponding to each color combination for each pixel using a multidimensional LUT corresponding to multi-value image signals of two or more colors. The multi-dimensional LUT for selecting the threshold fluctuation amount is set with a parameter that reflects a characteristic that causes an error when an error diffusion method combining two or more colors is performed.
[0129]
Accordingly, an error diffusion method combining two or more colors using a threshold fluctuation amount selected according to an input value while avoiding a phenomenon in which the magnitude relationship of the color components such as cyan and magenta is reversed between the input value and the output value. It became possible to do.
[0130]
Further, in the present invention, an error diffusion method combining two or more colors while selecting a threshold fluctuation amount corresponding to each color combination for each pixel using a multidimensional LUT corresponding to multi-value image signals of two or more colors. And a multi-dimensional LUT having parameters reflecting characteristics that cause errors when an error diffusion method combining two or more colors is performed and the quantization threshold is changed by the sum of input values of each color in the target pixel. Further, the fluctuation of the quantization threshold is corrected.
[0131]
As a result, it is possible to perform an error diffusion method combining two or more colors using the threshold fluctuation amount selected by the input value while avoiding the phenomenon that the magnitude relationship between the input value and the output value of cyan and magenta is reversed. It became. Since the threshold fluctuation amount is corrected in two stages, the average quantization error that causes the dot generation delay during the quantization process can be sufficiently reduced, and the image quality can be further improved.
[Brief description of the drawings]
FIG. 1 is a block diagram illustrating a configuration of an image processing apparatus according to an embodiment.
FIG. 2 is a block diagram illustrating a configuration of a threshold fluctuation amount setting unit in the image processing apparatus according to the first embodiment.
FIG. 3 is a block diagram illustrating a configuration of an output setting unit in the image processing apparatus according to the embodiment.
FIG. 4 is a schematic diagram illustrating an output value table in the image processing apparatus according to the embodiment.
FIG. 5 is a block diagram illustrating a configuration of an error calculation unit in the image processing apparatus according to the embodiment.
FIG. 6 is a schematic diagram showing an error distribution window.
FIG. 7 is a schematic diagram illustrating an output value table in a conventional image processing apparatus.
FIG. 8 is a block diagram illustrating a configuration of a conventional image processing apparatus.
FIG. 9 is a block diagram illustrating a threshold fluctuation amount setting unit of a conventional image processing apparatus.
FIG. 10 is a block diagram illustrating a configuration of an output setting unit in a conventional image processing apparatus.
FIG. 11 is a block diagram illustrating a configuration of an error calculation unit in a conventional image processing apparatus.
FIG. 12 is a block diagram showing another configuration of the threshold fluctuation amount setting unit in the first embodiment.
FIG. 13 is a block diagram illustrating a configuration of a threshold fluctuation amount setting unit in the image processing apparatus according to the second embodiment.
FIG. 14 is an explanatory diagram of an image used for measurement of an error average value.
FIG. 15 is an explanatory diagram of an error average value characteristic when threshold correction is not performed.
FIG. 16 is a flowchart of processing in a threshold fluctuation amount setting unit according to a modification of the second embodiment.
FIG. 17 is a block diagram showing a configuration of a threshold fluctuation amount setting unit in the third embodiment.
FIG. 18 is an explanatory diagram of an error average value characteristic when only correction for selecting a cabinet price change is performed by the sum of input values of two colors.
[Explanation of symbols]
101, 102, 201, 205, 206, 506, 801, 802, 903, 1106, 1201 adder
103, 104, 803, 804 Error calculator
105, 805, 806 Threshold fluctuation amount setting unit
106,807 Output value setting section
202, 902, 1202 Fluctuation threshold table
203, 204, 901 Noise matrix
301, 302, 503, 1001, 1002, 1103 Subtraction unit
303,1003 Output value table
401, 702 Error-corrected input pixel
501, 1101 distribution coefficient table
502, 1102 Central value table
504, 1104 Error dispersion unit
505, 1105 Error buffer
701 Input pixel

Claims (23)

An image processing apparatus for quantizing multi-valued color image data including a plurality of color components,
For each of a plurality of color components included in the pixel of interest, an adding means for adding a quantization error value distributed for each color component from neighboring pixels;
Variation amount determining means for determining a threshold variation amount of each color based on the sum of each color component value of the target pixel;
Quantization means for outputting a quantized value of each color using the error-corrected value of each color component, and calculating an error of each color from the error-corrected value of each color component and the quantized value of each color to the surrounding pixels. An error calculating means for distributing,
The image processing apparatus, wherein the quantizing unit varies a quantization threshold of each color component according to the threshold fluctuation amount when determining a value of each color component after conversion. The quantization means associates the sum of the values of the color components to which the error value has been added by the addition means with the sum of the values after the conversion, and the sum of the pixel values after the conversion, The image processing apparatus according to claim 1, wherein the value of each color component after distribution is determined according to a ratio and is determined. The quantization means, when determining the value of each color component after conversion, adds the threshold fluctuation amount to the value of each color component of the pixel of interest, thereby changing the quantization threshold of each color component. The image processing apparatus according to claim 1, wherein the image processing apparatus is characterized. The quantization means determines the quantization value of each color by referring to a look-up table in which the quantization value of each color is stored in advance with the value of each color component added with the error value by the addition means. The image processing apparatus according to claim 3. The image processing apparatus according to claim 1, wherein the number of gradation levels quantized by the quantization unit is two or more. The variation determining means has a table set so as to correct an average quantization error in each density gradation for each of a plurality of color components by an equal value for each of the plurality of color components, and the addition The value obtained by referring to the table by the sum of the values of the respective color components of the target pixel output by the means is determined as the threshold fluctuation amount. Image processing device. 7. The variation amount determining unit determines a threshold variation amount of each color based on a combination of color component values of the target pixel instead of a sum of color component values of the target pixel. The image processing apparatus according to any one of the above. The variation amount determining means has a table set so as to cancel the average quantization error in each density gradation for each of a plurality of color components for each of the plurality of color components, and is output by the adding means. The image processing apparatus according to claim 1, wherein a value obtained by referring to the table based on a combination of values of each color component of the target pixel is determined as the threshold fluctuation amount. Correction for correcting the threshold fluctuation amount so that the second average quantization error when the pixel of interest is quantized using the threshold fluctuation amount determined by the fluctuation amount determination means is further canceled for each color component. The image processing apparatus according to claim 1, further comprising a unit. The correction means calculates the threshold fluctuation amount using a table set so as to cancel out the second average quantization error in each density gradation for each of the plurality of color components for each of the plurality of color components. The image processing apparatus according to claim 1, wherein correction is performed. The image processing apparatus according to claim 1, wherein the table is configured to be referred to by a gradation number obtained by thinning a gradation number of an input pixel. An image processing method for quantizing multi-valued color image data including a plurality of color components,
For each of a plurality of color components included in the pixel of interest, an addition step of adding a quantization error value distributed for each color component from neighboring pixels;
A variation determination step for determining a threshold variation of each color based on the sum of the color component values of the pixel of interest;
A quantization step of outputting the quantized value of each color using the value of each color component corrected for error,
In the quantization step, when determining the value of each color component after conversion, the quantization threshold of each color component is changed according to the threshold fluctuation amount. In the quantization step, the sum of the values of the color components to which the error values are added in the addition step corresponds to the sum of the values after the conversion, and the sum of the pixel values after the conversion, The image processing method according to claim 11, wherein the value of each color component after conversion is determined by distribution according to a ratio. In the quantization step, when determining the value of each color component after conversion, the quantization threshold of each color component is changed by adding the threshold fluctuation amount to the value of each color component of the target pixel. The image processing method according to claim 12 or 13, wherein the image processing method is characterized by the following. The quantization step is to determine the quantization value of each color by referring to a lookup table in which the quantization value of each color is stored in advance with the value of each color component added with the error value in the addition step The image processing method according to claim 14. 16. The image processing method according to claim 12, wherein the number of gradation levels quantized in the quantization step is two or more. The variation determination step includes a table set to correct an average quantization error in each density gradation for each of a plurality of color components by an equal value for each of the plurality of color components, and the addition The value obtained by referring to the table by the sum of the values of each color component of the pixel of interest output in the process is determined as the threshold fluctuation amount. Image processing method. 18. The variation amount determining step determines a threshold variation amount of each color based on a combination of color component values of the target pixel instead of a sum of color component values of the target pixel. The image processing method according to any one of the above. The variation amount determining step has a table set so as to cancel the average quantization error in each density gradation for each of the plurality of color components for each of the plurality of color components, and is output by the adding step. 19. The image processing method according to claim 12, wherein a value obtained by referring to the table based on a combination of values of each color component of the target pixel is determined as the threshold fluctuation amount. Correction for correcting the threshold variation amount so that the second average quantization error when the pixel of interest is quantized using the threshold variation amount determined in the variation amount determination step is further canceled for each color component. The image processing method according to claim 12, further comprising a step. In the correcting step, the threshold fluctuation amount is calculated using a table set so as to cancel out the second average quantization error in each density gradation for each of the plurality of color components for each of the plurality of color components. The image processing method according to claim 12, wherein correction is performed. A program for causing a computer to execute the method according to any one of claims 12 to 21. A computer-readable recording medium on which the program according to claim 22 is recorded.

Images