JP5744461B2 - Image processing apparatus and method - Google Patents

Description

  The present invention relates to image processing for expressing an image by reducing the number of gradations of the image.

  A dither method is known as a pseudo halftone process for expressing an image by reducing the number of gradations. In the dither method, a threshold matrix prepared in advance and multi-valued input data are compared, for example, the input data is binarized, and a gradation (image) represented by the input data is expressed in a pseudo manner. The dither method is a high-speed halftone processing method capable of quantization such as binarization, ternarization, quaternarization,... By comparing a threshold matrix and multi-value input data.

  When halftone processing is performed using the dither method, good quantization is possible in a flat portion where the change in gradation is gentle. However, in an edge portion such as a fine line of characters including a high frequency wave component, edge irregularity or interruption of the fine line occurs in the image after halftone processing, and the reproducibility of the image is lowered. As a technique to reduce edge disturbance and fine line breaks when using the dither method, a technique has been proposed to reduce the threshold matrix value in areas that are determined to be characters and thin lines, thereby preventing edge disturbances and fine line breaks. (Patent Document 1).

  However, the method described in Patent Document 1 requires the determination of characters and thin lines, and if an erroneous determination occurs, the edge is disturbed or the thin lines are interrupted. Furthermore, even if characters and fine lines are correctly determined, the threshold matrix value is decreased, so that the density of characters and fine lines increases, and the density reproducibility decreases.

Japanese Patent Laid-Open No. 11-345327

  It is an object of the present invention to suppress edge disturbance and fine line interruption in halftone processing of an image.

  Another object is to improve the density reproducibility of a halftone processed image.

  The present invention has the following configuration as one means for achieving the above object.

The image processing according to the present invention generates a filtered image obtained by filtering an input image, generates a quantized image obtained by quantizing the input image, stores the quantized image in a memory, and stores the quantized image in the memory. Generating a quantized filtered image obtained by filtering the quantized image, generating a first difference image indicating a difference between the filtered image and the quantized filtered image, and generating the quantized image based on the first difference image. An updated quantized image in which pixel values of the updated quantized image are generated, an updated quantized filtered image obtained by filtering the updated quantized image is generated, and a second difference indicating the difference between the filtered image and the updated quantized filtered image is generated. generates a difference image, it calculates a second evaluation value of the first evaluation value and the second difference image of the first difference image, before If the first evaluation value is greater than the second evaluation value is determined to repeat the generation of the update quantized image, when the first evaluation value is less than the second evaluation value is the update quantum If it is determined not to repeat generation of the quantized image and it is determined to repeat generation of the updated quantized image, the quantized image stored in the memory is replaced with the updated quantized image, and the updated quantized image is If it is determined not to repeat the generation, the quantized image stored in the memory is output.

  According to the present invention, it is possible to suppress edge disturbance and fine line breaks in halftone processing of an image.

  Further, it is possible to improve the density reproducibility of the halftone processed image.

1 is a block diagram illustrating a configuration example of an image processing apparatus according to an embodiment. 6 is a flowchart for explaining halftone processing executed by the image processing apparatus. The figure explaining that a filter coefficient changes with the value of an attention pixel. The figure explaining the outline | summary of the dither process of a quantization part. Input image Gp, LPF image Gp _F, quantized image Q _Gp, diagram for explaining a quantization LPF image Q _GPF. The figure explaining the quantization image Q _Gp and the update quantization image IQ _Gp . The figure explaining the quantized image Q _Gp output from an output part.

  Hereinafter, image processing according to an embodiment of the present invention will be described in detail with reference to the drawings.

[Device configuration]
A configuration example of the image processing apparatus according to the embodiment will be described with reference to the block diagram of FIG. The image processing apparatus according to the embodiment can be realized by installing a printer driver in a general-purpose computer, for example. In this case, each configuration of the image processing apparatus described below is realized and controlled by a computer microprocessor (CPU) executing a printer driver program.

  The input unit 101 inputs an image to be subjected to halftone processing. The LPF unit 102 serving as the first filter generates a filtered image (hereinafter, LPF image) obtained by performing low-pass filter (LPF) processing on the image input by the input unit 101. The quantization unit 103 generates an image obtained by quantizing the image input by the input unit 101 (hereinafter, “quantized image”), and stores the quantized image in the pre-update memory 104. Further, when a quantized image is input from the reading unit 115, the quantizing unit 103 does not perform quantization, and replaces the quantized image stored in the pre-update memory 104 with the input quantized image.

  The LPF unit 105, which is the second filter, generates a quantized filtered image (hereinafter referred to as a quantized LPF image) obtained by performing LPF processing on the quantized image stored in the pre-update memory 104. The adder 106 subtracts the LPF image from the quantized LPF image to generate a first difference image. Although the details will be described later, the pre-update evaluation unit 107 evaluates the first difference image output from the adder.

  Although the details will be described later, the pixel update unit 108 generates an updated quantized image in which the pixel value of the quantized image stored in the pre-update memory 104 is updated based on the evaluation value output by the pre-update evaluation unit 107. The updated quantized image is stored in the updated memory 109.

  The LPF unit 110, which is the third filter, generates an updated quantized filtering image (hereinafter referred to as an updated quantized LPF image) obtained by low-pass filtering the updated quantized image. The adder 111 subtracts the LPF image from the updated quantized LPF image to generate a second difference image. Although the details will be described later, the post-update evaluation unit 112 evaluates the second difference image output from the adder 111.

  The comparison unit 113 compares the evaluation value output from the pre-update evaluation unit 107 with the evaluation value output from the post-update evaluation unit 112. Although the details will be described later, the determination unit 114 determines whether to continue the update by the pixel update unit 108 based on the comparison result of the comparison unit 113, in other words, whether to repeat the generation of the updated quantized image. . Based on the determination result of the determination unit 114, the reading unit 115 reads either the quantized image stored in the pre-update memory 104 or the updated quantized image stored in the post-update memory 109.

  When the determination result of the determination unit 114 indicates that updating is continued, the reading unit 115 reads the updated quantized image stored in the post-update memory 109 and inputs the updated quantized image to the quantizing unit 103. When the determination result of the determination unit 114 indicates the end of update, the reading unit 115 reads the quantized image stored in the pre-update memory 104 and inputs the quantized image to the output unit 116. The output unit 116 outputs the quantized image, and a series of halftone processing ends.

[Halftone processing]
The halftone process executed by the image processing apparatus will be described with reference to the flowchart of FIG. The processing shown in FIG. 2 is realized by the CPU executing a printer driver program. In the following, an example of halftone processing of the entire image will be described. However, the image may be divided into a plurality of local areas, and these local areas may be subjected to halftone processing in parallel. In that case, the processing shown in FIG. 2 is executed for each local region. In the following, an example in which the input image is 8 bits and binarized as quantization will be described. However, the bit depth of the image and the number of gradations after quantization are not limited thereto. Further, an example in which 0 of the image data represents the minimum density (for example, white) and 255 represents the maximum density (for example, black) will be described, but the present invention is not limited to this. Furthermore, although an example in which the size of the input image is 32 × 32 pixels will be described, it goes without saying that other sizes may be used.

As processing of the LPF unit 102, the CPU performs LPF processing on the input image to generate an LPF image (S101). If the input image is Gp (x, y) (0 ≦ x ≦ 31, 0 ≦ y ≦ 31), the LPF image Gp _F is calculated by the following equation. In the LPF process, a two-dimensional Gaussian filter F (Gp, px, py) that changes according to the value of the target pixel of Gp is used.
Gp _F = Gp * F (Gp, px, py)… (1)
F (Gp, px, py) = F '(Gp, px, py) / Sum… (2)
F '(Gp, px, py) = {1 / 2πσx (Gp) σy (Gp)} exp [-{(px / σx (Gp)) ² + (py / σy (Gp)) ² } / 2]… (3)
Where * represents the convolution operation.
Sum is the sum of F 'coefficients.

  In Expression (3), the values of the weighting coefficients σx (Gp) and σy (Gp) change according to the value of the target pixel. The fact that the filter coefficient changes depending on the value of the target pixel will be described with reference to FIG. In order to simplify the explanation, FIG. 3 shows an example of a square with a filter size of 5 × 5, but a size such as 7 × 7 or 9 × 9 may be used. FIG. 3 (a) shows a filter example where σx (Gp) = 1.6 and σy (Gp) = 1.6 at Gp = 30. FIG. 3B shows a filter example in which σx (Gp) = 1.3 and σy (Gp) = 1.3 at Gp = 40. That is, the filter coefficient of the filter F changes according to the value of the target pixel.

  Next, as a process of the quantization unit 103, the CPU generates a quantized image (S102), and stores the quantized image in the pre-update memory 104 (S103). The quantization unit 103 quantizes the image by a dither method using a blue noise type threshold matrix. Of course, the image may be quantized by other halftone processing without using the dither method, or a threshold matrix other than the blue noise type, for example, a Bayer type or AM screen type threshold matrix may be used.

An outline of the dither processing of the quantization unit 103 will be described with reference to FIG. The quantization unit 103 compares each pixel value of the input image Gp with the threshold value of the corresponding threshold value matrix Th, and outputs a quantized image Q _Gp . The quantization rule at that time is expressed by the following equation.
if (Gp ≤ Th)
Q _Gp = 0;
else
Q _Gp = 255;… (4)

Next, the CPU performs LPF processing on the quantized image Q _Gp stored in the pre-update memory 104 as processing of the LPF unit 105, and generates a quantized LPF image Q _GpF (S104). Note that a filter F represented by equations (2) and (3) is used as a filter. That is, the LPF process of the LPF unit 105 is expressed by the following equation.
Q _GpF = Q _Gp * F (Gp, px, py)… (5)

The input image Gp, LPF image Gp _F , quantized image Q _Gp , and quantized LPF image Q _GpF will be described with _reference to FIG. As shown in FIG. 5, the input image Gp is an image in which halftone fine lines are drawn in a grid pattern. When the input image Gp is quantized, the thin line is interrupted as in the quantized image Q _Gp , and when the input image Gp is subjected to LPF processing, an image with a thin line as in the LPF image Gp _F is obtained. In addition, when the quantized image Q _Gp is subjected to LPF processing, an image with a broken fine line is obtained as in the quantized LPF image Q _GpF .

Then, CPU, as processing of the adder 106 calculates a difference image Gdiff obtained by subtracting the LPF image Gp _F from the quantization LPF image Q _GpF (S105).
Gdiff = Q _GpF -Gp _F … (7)

Next, the CPU evaluates the difference image Gdiff as a process of the pre-update evaluation unit 107 (S106). In this embodiment, the maximum value Max _e and the minimum value Min _e of the pixels included in the difference image Gdiff are extracted as evaluation values, and the sum Abs _e of these absolute values is calculated.
Max _e = Max (Gdiff)… (8)
Min _e = Min (Gdiff)… (9)
Abs _e = √ (Max _e ² ) + √ (Min _e ² )… (10)

Next, the CPU updates the quantized image Q _Gp stored in the pre-update memory 104 based on the coordinates of the pixels indicating the maximum value Max _e and the minimum value Min _e of the difference image Gdiff as processing of the pixel update unit 108. An updated quantized image IQ _Gp with updated pixel values is generated (S107). Then, the updated quantized image IQ _Gp is stored in the updated memory 109 (S108).

The pixel update unit 108 acquires the coordinates (x _Max , y _Max ) of the pixel P indicating Max _e and the coordinates (x _Min , y _Min ) of the pixel R indicating Min _e in the difference image Gdiff.
x _Max = x (P (Max _e ))
y _Max = y (P (Max _e ))… (11)
x _Min = x (R (Min _e ))
y _Min = y (R (Min _e ))… (12)

Subsequently, the pixel update unit 108 replaces the pixel values of the quantized image Q _Gp stored in the pre-update memory 104 based on the acquired coordinates, and generates an updated quantized image IQ _Gp .
IQ _Gp (x, y) = Q _Gp (x, y)… (13)
IQ _Gp (x _Min , y _Min ) = tmpMax… (14)
IQ _Gp (x _Max , y _Max ) = tmpMin… (15)
Where tmpMax is Q _Gp (x _Max , y _Max )
tmpMin is Q _Gp (x _Min , y _Min )

The quantized image Q _Gp and the updated quantized image IQ _Gp will be described with reference to FIG. In the quantized image Q _Gp , reference numeral 601 indicates the coordinates (x _Max , y _Max ) of the pixel indicating Max _e, and reference numeral 602 indicates the coordinates (x _Min , y _Min ) of the pixel indicating Min _e . In the updated quantized image IQ _Gp , reference numeral 603 indicates coordinates (x _Max , y _Max ), and reference numeral 604 indicates coordinates (x _Min , y _Min ). That is, the pixel value of the quantized image Q _Gp corresponding to the coordinate of the pixel indicating the maximum value Max _e in the difference image Gdiff and the quantized image Q _Gp corresponding to the coordinate of the pixel indicating the minimum value Min _e in the difference image Gdiff An updated quantized image IQ _Gp is obtained by replacing pixel values with each other.

Next, as processing of the LPF unit 110, the CPU performs LPF processing on the updated quantized image IQ _Gp to generate an updated quantized LPF image IQ _GpF (S109). Note that a filter F represented by equations (2) and (3) is used as a filter. That is, the LPF process of the LPF unit 110 is expressed by the following equation.
IQ _GpF = IQ _Gp * F (Gp, px, py)… (16)

Then, CPU, as processing of the adder 111 calculates a difference image IGdiff obtained by subtracting the LPF image Gp _F from updated quantization LPF image IQ _GpF (S110).
IGdiff = IQ _GpF -Gp _F … (17)

Next, the CPU evaluates the difference image IGdiff as a process of the post-update evaluation unit 112 (S111). In the present embodiment, as the evaluation value, and extracted maximum value IMax _e of pixels including the difference image IGdiff, the minimum value Imin _e, calculates their sum of absolute values IABS _e.
IMax _e = Max (IGdiff)… (18)
IMin _e = Min (IGdiff)… (19)
IAbs _e = √ (IMax _e ² ) + √ (IMin _e ² )… (20)

Then, CPU as processing of the comparison section 113 compares the magnitude relationship between the evaluation value IABS _e evaluation value Abs _e with updated evaluation unit 112 of the update before evaluating unit 107 (S112). Then, as the processing of the determination unit 114, if the evaluation value before the update exceeds the evaluation value after the update (Abs _e > IAbs _e ), it is determined that the update is continued, and the evaluation value before the update is equal to or less than the evaluation value after the update ( for abs _e ≦ IAbs _e) determines that the update end (S113).

When updating is continued, the CPU reads the updated quantized image IQ _Gp stored in the post-update memory 109 as processing of the reading unit 115 (S114). Then, the process returns to step S103, and the quantized image Q _Gp stored in the pre-update memory 104 is replaced with the updated quantized image IQ _Gp . Further, when the update is completed, as the processing of the reading unit 115 and the output unit 116, the quantized image Q _Gp stored in the pre-update memory 104 is read, the quantized image Q _Gp is output (S115), and the processing is terminated. To do.

The quantized image Q _Gp output from the output unit 116 will be described with reference to FIG. FIG. 7A shows a quantized image Q _Gp that has been subjected to dither processing (quantization) by the quantization unit 103 (S102). FIG. 7B shows a quantized image Q _Gp output from the output unit 116. From these figures, it can be seen that the black pixel and the white pixel are exchanged by the pixel value exchange by the pixel updating unit 108 (S107), and the reproducibility of the thin line is improved.

The replacement (update) of the pixel values is repeated until the evaluation value before the update becomes equal to or lower than the evaluation value after the update (Abs _e ≦ IAbs _e ). That is, if the evaluation value becomes smaller due to the replacement of the pixel value, the replacement of the pixel value (generation of the updated quantized image) is continued, and if the evaluation value does not become smaller, the replacement of the pixel value (generation of the updated quantized image) ) Ends. Then, as the dithering of the quantized image Q _Gp immediately before the evaluation value before updating is below evaluation value updated is in the best condition, and outputs the data stored in the pre-update memory 104 quantized image Q _Gp .

  In this way, by changing the pixel values (generation of updated quantized images) so as to improve the edges, fine lines, and reproducibility of the halftone processed image, it is possible to suppress edge disturbances and fine line interruptions, A halftone processed image with high reproducibility can be generated.

[Modification]
In the above description, the example in which the maximum value and the minimum value of the pixels included in the first and second difference images are extracted and the sum of their absolute values is used as the evaluation value has been described. This is an example in which an evaluation value is calculated based on the value of a pixel having a large value in both positive and negative directions because each pixel of the difference image takes a negative value from a positive value. The evaluation value is not limited to the sum of absolute values, and an absolute value of the difference between the maximum value and the minimum value can also be used.

  In the above, the pixel value of the quantized image corresponding to the coordinate of the pixel indicating the maximum value of the difference image Gdff and the pixel value of the quantized image corresponding to the coordinate of the pixel indicating the minimum value of the difference image Gdff are interchanged and updated. An example of generating a quantized image has been described. However, the pixel value of the quantized image corresponding to the coordinates of the pixel indicating the maximum value of the first, second, third,... And the coordinate of the pixel indicating the minimum value of the first, second, third,. The updated quantized image may be generated by exchanging the pixel values of the corresponding quantized images.

[Other Examples]
The present invention can also be realized by executing the following processing. That is, software (program) that realizes the functions of the above-described embodiments is supplied to a system or apparatus via a network or various storage media, and a computer (or CPU, MPU, etc.) of the system or apparatus reads the program. It is a process to be executed.

Claims (6)

First filter means for generating a filtered image obtained by filtering the input image;
Quantization means for generating a quantized image obtained by quantizing the input image and storing the quantized image in a memory;
Second filter means for generating a quantized filtered image obtained by filtering the quantized image;
A first difference image generating means for generating a first difference image indicating a difference between the quantized filtering image and the filtering image;
Updating means for generating an updated quantized image in which pixel values of the quantized image are updated based on the first difference image;
Third filter means for generating an updated quantized filtered image obtained by filtering the updated quantized image;
A second difference image generating means for generating a second difference image indicating a difference between the updated quantized filtering image and the filtering image;
Calculating a second evaluation value of the first evaluation value and the second difference image of the first difference image, when the first evaluation value is greater than the second evaluation value is the update quantum A determination unit that determines that the generation of the quantized image is repeated, and determines that the generation of the updated quantized image is not repeated when the first evaluation value is equal to or less than the second evaluation value;
When it is determined that the generation of the updated quantized image is repeated, replacement means for replacing the quantized image stored in the memory with the updated quantized image;
An image processing apparatus comprising: output means for outputting the quantized image stored in the memory when it is determined not to repeat generation of the updated quantized image. The first evaluation value is the sum of the absolute value of the maximum value of the pixel and the absolute value of the minimum value of the pixel in the first difference image,
2. The image processing apparatus according to claim 1, wherein the second evaluation value is a sum of an absolute value of a maximum value of a pixel and an absolute value of a minimum value of the pixel in the second difference image. Until the determination means determines not to repeat the generation of the updated quantized image, the processing of the second and third filter means, the first and second difference image generation means, and the update means is performed. 3. The image processing device according to claim 1, wherein the image processing device is repeated. The update means includes the pixel value of the quantized image corresponding to the coordinate of the pixel indicating the maximum value of the first difference image and the quantum value corresponding to the coordinate of the pixel indicating the minimum value of the first difference image. of pixel values of an image by the replaced each other, image processing apparatus according to any one of claims 1 to 3, characterized in that generating the update quantized image. An image processing method of an image processing apparatus having first to third filter means, quantization means, first and second difference image generation means, update means, determination means, replacement means, and output means,
The first filter means generates a filtered image obtained by filtering the input image;
The quantization means generates a quantized image obtained by quantizing the input image, and stores the quantized image in a memory;
The second filter means generates a quantized filtered image obtained by performing the filtering process on the quantized image stored in the memory;
The first difference image generating means generates a first difference image indicating a difference between the quantized filtered image and the filtered image,
The update means generates an updated quantized image in which pixel values of the quantized image are updated based on the first difference image,
The third filter means generates an updated quantized filtered image obtained by performing the filtering process on the updated quantized image;
The second difference image generating means generates a second difference image indicating a difference between the updated quantized filtering image and the filtering image,
Said determining means, said calculating a first second evaluation value of the first evaluation value and the second difference image of the difference image, the first evaluation value is greater than the second evaluation value If it is determined that the generation of the updated quantized image is repeated, and if the first evaluation value is equal to or less than the second evaluation value, it is determined that the generation of the updated quantized image is not repeated ,
If it is determined that the replacement unit repeats generation of the updated quantized image, the quantized image stored in the memory is replaced with the updated quantized image,
An image processing method comprising: outputting the quantized image stored in the memory when the output unit determines that the generation of the updated quantized image is not repeated. A program for causing a computer to function as each unit of the image processing apparatus according to any one of claims 1 to 4.