JP2683085B2 - Image processing device - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to an image processing device that performs a binary or multivalued quantization process on image data.

(Prior art)

2. Description of the Related Art Conventionally, error diffusion methods and average density approximation methods have been proposed as pseudo halftone processing methods in image processing apparatuses such as facsimile machines and digital copying machines.

The former error diffusion method is described in R.FLOYD & L.STEINBER
G, “AN ADAPTIVE ALGORITHMFOR SPETIAL GRAY SCA
As disclosed in LE ", SID 75 DIGEST, PP36-37,
The multi-valued image data of the pixel of interest is binarized (converted to the darkest level or the lightest level), and the error between the binarized level and the multi-valued image data before the binarization is given a predetermined weighting. This is to be added to the data of the pixels in the vicinity of the pixel of interest.

In addition, as described in Japanese Patent Application Laid-Open No. 57-104369, the latter average density approximation method converts a target pixel to black or white using binary data already binarized in the vicinity of the target pixel. Weighted average value with each neighboring pixel in the case of
The image data of the target pixel is binarized using the average of the two average values as a threshold.

[Problems to be Solved by the Invention] The above-mentioned error diffusion method is a method of correcting an error between input image data and output image data, and therefore the densities of the input image and the output image can be saved, and the resolution and the level It is possible to provide an image with excellent tonality.

However, the error diffusion method has to perform many two-dimensional operations when correcting an error between input image data and output image data, and the hardware configuration becomes very complicated due to the large amount of processing. There were drawbacks.

In addition, since the average density approximation method uses the binary data after binarization, the operation can be simplified, so that the hardware configuration can be simplified and the processing speed can be increased due to the extremely small amount of processing. It is possible.

However, the average density approximation method simply approximates a target pixel to an average value of a region including the target pixel and performs binarization, so that the number of gradations is limited and an image having a gentle density change is obtained. There is a drawback that a peculiar low-frequency texture is generated and image quality is deteriorated.

[Means and actions for solving the problem]

The present invention aims to eliminate the above-mentioned drawbacks of the prior art, and provides an image processing apparatus capable of obtaining an image excellent in both gradation and resolution in a short time with a simple hardware configuration.

That is, the image processing apparatus of the present invention includes an input means for inputting data of a pixel of interest, a calculating means for obtaining an average density value of a predetermined area, and data of the pixel of interest based on the average density value obtained by the calculating means. Quantizing means for quantizing
And a correction means for correcting the error generated when the quantization is within a predetermined range.

〔Example〕

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

Example 1 First, the principle of this system will be described.

FIG. 1 (1) is a diagram showing multi-value data for each pixel of an input image.

In FIG. 1 (1), f (i, j) indicates multi-value density data of the input image at the target pixel position to be binarized, and is a normalized value of 0 to 1. Further, the pixel position above the broken line has already been binarized,
After the value conversion, the same processing is sequentially performed as f (i, j + 1), f (i, j + 2).

FIG. 1 (2) is a diagram showing the binarized image data.
(I, j) indicates the binarized density (0 or 1) of the target pixel. The portion surrounded by the broken line is pixel data that has already been binarized when the pixel of interest is processed, and these are used in the binarization process of the pixel of interest.

FIG. 1 (3) is a diagram showing a weighting mask. R is an example of a weighting mask for obtaining an average density, and 3 ×
It is represented by a matrix of three sizes. The weight of the position corresponding to the target pixel is R (0,0), and R (0, -1) = 0 is used.

This method calculates the average density of the output image near the target pixel when the target pixel is binarized to either black or white.
Let m1 (i, j) and m0 (i, j) be the respective equations.

(However, it is assumed that B (i, j) = 1, that is, the target pixel is black) (However, it is assumed that B (i, j) = 0, that is, the target pixel is white.) Here, S is the total sum of the weights R. For example, when the weight mask of FIG. 3 is used, S = 29. .

The target pixel f (i, j) is binarized according to the following equation using the average values m1, m0 and the binarized correction value E (i, j) already assigned.

In the above equation, E (i, j) is 1 of the pixel of interest (i, j).
Multi-value density data f (i, j−) of the pixel (i, j−1) before the pixel
This is an error that occurs when 1) is binarized into binary density data B (i, j-1). That is, the input pixel density data f
The binarization from (i, j-1) to 1 or 0 means that the pixel (i, j-1) is the average density in the approximation m1 (i, j-
1) or m0 (i, j-1), which means that in each case, the multi-valued density f (i, j) of the input image is
-1) and f (i, j-1) -m1 or f (i, j-1)-
An error of m0 occurs. Therefore, this binarization error E (i,
By binarizing the corrected value by adding j) to the target pixel f (i, j), the density can be completely preserved on the binarized image over the entire input image. Such 2
By carrying out the processing in consideration of the digitization error, the halftone reproduction capability is remarkably improved as compared with the above-mentioned average density approximation method.

In the equation, E (i, j + 1) is (i, j) of the pixel of interest.
The error is assigned to the pixel (i, j + 1) one pixel after j). As shown in FIG. 2, E (i, j + 1) is f (i, j,
j) + E (i, j)> (m1 + m0) / 2, then f (i, j) + E
(I, j) minus m1, f (i, j) + E (i, j) ≦
In the case of (m1 + m0) / 2, it is obtained by subtracting m0 from f (i, j) + E (i, j).

Further, although the present embodiment has an extremely small processing amount as compared with the error diffusion method, it is possible to obtain the image reproducing ability at the same time as or more than this, only by correcting the error by one adjacent pixel. Despite this, by obtaining the average density using a plurality of data after binarization, an effect equivalent to that in which an error is equivalently distributed and corrected by a plurality of pixels can be obtained.

In the present embodiment, when the error-corrected data (f (i, j) + E (i, j)) is within a predetermined range, the error (E (i, j + 1) = f (i, j). ) + E (i, j) -m0 (i,
j)) is corrected. That is, the error E (i, j + 1) when binarizing the next pixel is And in other cases, And

In other words, the processing that is characteristic of this method is, as shown in the above equation, in comparing the average values m1 and m0 at the time of binarization with the target pixel correction value, the target pixel correction value is the selected average value m1 or m0. If the difference is between the selected average value and the pixel-of-interest correction value according to the equation, if the difference is within the predetermined area (error E (i, j + 1) is within the predetermined area) taking a value close to Assign as a correction value when binarizing. On the other hand, when the difference between the target pixel correction value and m1 and m0 is outside the predetermined area, the correction value is set to 0, and the correction for binarizing the next pixel is not performed.

In other words, the former has a small change in the density of the image in the vicinity of the pixel of interest, and can therefore be judged to be an image area having halftones.
Therefore, by correcting the difference from the average density value generated by binarizing with the next pixel, the pseudo intermediate processing of the smooth density change of the image can be faithfully performed. That is, the gradation can be improved. On the other hand, in the latter case, it is possible to determine that the edge portion of the character, line drawing, or the like, that is, the pixel of interest changes drastically compared to the density of the neighboring image. Therefore, the correction value is 0 for the pixel in that case, Binary reproduction is performed while suppressing the decrease in resolution due to storage of density. Thereby, the resolution in the edge portion can be improved.

As described above, according to the characteristic processing method of the present embodiment, the halftone image area is binarized by using the binarization error according to the image density change.
The density is saved on the binarized image, and in the resolution image part such as characters, the binarization error is not corrected in order to prevent the blurring of the image due to the above density storage, and the average density values m1 and m0 are approximated. It is what makes me.

FIG. 4 is a block diagram of an image processing apparatus showing one embodiment of the present invention. The input sensor unit A includes a photoelectric conversion element such as a CCD and a driving device for scanning the same, and performs scanning of a document. The image data of the document read by the input sensor unit A is sequentially sent to the A / D converter B. Here, the data of each pixel is converted into 6-bit digital data and quantized to data having 64 levels of gradation. Next, in the correction circuit C, the shading correction for correcting the sensitivity unevenness of the CCD sensor and the illuminance unevenness due to the illumination light source is performed by digital calculation processing. Next, the corrected data is sent to the binarization circuit D. In the binarization circuit D, the input 6-bit multi-valued image data is converted into 1
Quantization processing is performed on bit binary data. The printer E is a printer configured by a laser beam or an ink jet system, and controls the dot on / off based on the binary data sent from the binarization circuit D to reproduce an image on a recording sheet.

FIG. 5 is a block diagram showing details of the binarizing circuit D in FIG.

In FIG. 5, reference numerals 1 and 2 denote a delay RAM for storing one line of binary data subjected to the binarization processing, reference numerals 3 to 7, 11 denote DF / Fs (flip flops) for delaying the binary data by one pixel,
8 is an average density calculation ROM for calculating the average density around the target pixel and outputting a threshold value; 9 is a subtractor for calculating the difference between the input multi-value data of the target pixel and the threshold value; 10 is output from the ROM 8 11 is a DF / F, 12 is a ROM for calculating error data to be added to the multi-value data input next to the pixel of interest, and 13 is input data. And an adder for adding the error data output from the ROM 12.

In the above configuration, the comparator 10 outputs 1-bit data B (i, j) binarized based on the equation to the DF / F7 and the printer E. The binary data is stored in a RAM for delaying each line.
2, Binary data B (i-1, j + 1) input to RAM1 and delayed by one line by RAM2, and binary data B (i-2, j + 1) delayed by two lines by RAM1 are output to ROM8. .

Further, DF / F3 is B (i-2, j), and DF / F4 is (i-2, j-
1), DF / F5 is (i-1, j), DF / F6 is (i-1, j-1),
DF / F7 outputs (i, j-1) to ROM8.

The binary data is input image f as shown in FIG.
(I, j) is a binarized image of peripheral pixels, and if these are connected to the input address of the ROM 8, the ROM 8 previously stores the binarized threshold (m1 (i ,
Since j) + m0 (i, j)) / 2 is stored, the binarization threshold value can be obtained at high speed.

This threshold is input to the subtractor 9 and the comparator 10. On the other hand, the subtractor 9 and the comparator 10 provide f (i, j) +
E (i, j) is input.

Based on these two inputs, the subtractor 9 calculates the difference f (i, j) + E (i, j)-(m1 (i, j) + m0 (i, j)) / 2 between the two sides of the inequality in the equation. . If the above is transformed using the equation, Becomes

On the other hand, the comparator 10 calculates f (i,
j) + E (i, j) is compared with (m1 (i, j) + m0 (i, j)) / 2 to output binary data B (i, j).

That is, in the ROM 12, the value of B (i, j) from the comparator 10 and f (i, j) -E (i, j)-(m1 (i, j) + m0 from the subtractor 9
E (i, j + 1) expressed by the equation is calculated from (i, j)) / 2.

In the above formula, since the weights R (0,0) and S are known, the binarization error E (i, j + 1) according to the formula is calculated in advance and stored in the error calculation ROM 12, so that the binarization is performed. These are the data B (i, j) and the output of the subtractor 9.

If f (i, j) + E (i, j)-(m1 (i, j) + m0 (i, j)) / 2 is input to the ROM 12, E (i, j + 1) is obtained by table conversion.

FIG. 6 shows an example of a table stored in the ROM 12.

In this embodiment, the weight mask 1 shown in FIG.
In order to normalize to the actually input 6-bit image density level (0 to 63), the average density calculation ROM table is converted into a 6-bit value by multiplying the value obtained by the formula by 63. Store it. In this case, the weight mask 1 is as shown in FIG.

FIG. 6 shows a table when α = 1 in the equation, and when E (i, j + 1) is larger than R (0,0) = 18, E (i, j + 1) is set to 0.

The output f (i, j) + E (i, j)-(m1 (i,
j) + m0 (i, j)) / 2 is input to ROM12 as an absolute value,
Positive or negative is determined according to the value of B (i, j).

The error E (i, j + 1) obtained by the ROM 12 is added to the input image data f (i, j + 1) by the adder 13. The DF / F11 delays the added value by one data period.

As described above, this embodiment can be easily realized by adding a few chips of the operation IC as compared with the average density approximation method.

As described above, according to the first embodiment of the present invention, the average density is calculated based on the already binarized data, and the binarization processing is performed based on the average density. The processing amount for can be significantly reduced. Moreover, when the difference between the average density generated when binarizing and the input multi-valued data is within the predetermined range, the difference is corrected, so that the halftone processing excellent in gradation can be performed.

Further, in the present embodiment, when the difference between the average density and the input multi-valued data is larger than the predetermined value, the difference is not corrected, so that the resolution is prevented from being lowered by storing the density, and the edge portion is clearly reproduced. be able to.

In the present embodiment, the binarization error E is corrected by allocating it to only the next pixel in the formula, but it may be allocated two-dimensionally to a plurality of neighboring pixels at a predetermined allocation rate. In this case, although the hardware configuration is somewhat complicated, a uniform image can be obtained in the main scanning direction as well as the sub scanning direction, and reproducibility is improved.

Further, the binarization error E is, for example, 3E (i, j + 1) / 4 to pixel (i, j + 1) and E to pixel (i, j + 2) in FIG.
If the pixels are distributed to a plurality of pixels such as (i, j + 1) / 4, the gradation reproducibility is improved even if the average processing mask is small.

Further, although the weight mask is increased as approaching the pixel of interest, the inclination and distribution are not limited, and pixels at discrete positions that are not adjacent may be used.

Further, the present invention can be widely used in image processing apparatuses such as facsimile machines and copying machines.

Second Embodiment In the first embodiment, the binarization error E is divided into cases represented by the equation, and when the error E is a certain value or more using the constant α, the error E is set to 0 and is assigned to the next pixel. Although it is not covered, the value of the constant α can be changed according to the average density value or the target pixel density value.

For example, as shown in FIG. 8, if the average density is set to be smaller as it gets closer to 0 or 1, it is possible to obtain an edge portion such as a black character in a white background or a white character in a black background.
It is possible to perform binarization with higher precision.

Example 3 The following formula is used in place of the formula of Example 1.

Here, K is a constant, and good results can be obtained by setting K = 0,1.

According to the present embodiment, the error E is set to 0 when the average density value approaches 0 or 1, so that the same as in the second embodiment.
The character portion can be binarized with high precision.

In the second and third embodiments, as shown in FIG. 5, the binary data B (i,
Instead of j), the average density calculation ROM8 output (m1 (i, j) +
By inputting m0 (i, j) / 2, it is possible to easily carry out the table conversion processing based on the prewritten data as in the first embodiment.

Fourth Embodiment In the first embodiment, the weight mask of 3 × 3 matrix shown in FIG. 3 is used. Generally, in order to smoothly binarize the halftone part, the weight R (0, It is desirable to set 0) small. Further, the edge portion shown by the equation is also detected by the weight R for the density change of the binarized data.
The smaller (0,0), the more accurate the operation. Therefore,
When the 4 × 5 weight mask shown in FIG. 9 is used, R (0,0) in FIG. 3 of the first embodiment is 8/28 = 0.29, whereas that in FIG. Since /96=0.11, the halftone part can be smoothly binarized, and the edge part such as the character part can be binarized more finely.

In the previous embodiment, the calculation of the correction value E for storing the in-process density is performed using the average values m0 and m1. However, for example, it is publicly known whether or not E is set to 0 in the edge portion. Technology of
For example, a two-dimensional Laplacian is obtained from the binarized image data, the edge portion is determined based on the determination result obtained by thresholding the value, and the same result can be obtained even when E is 0 in the edge portion. Further, based on the instruction obtained by the operator's area specifying operation, the edge portion may be specified in a wide area instead of switching the processing for each pixel, and E may be set to 0 in that area.

In the above embodiment, an example in which the input multi-valued data is quantized into binary data has been described, but the present invention is ternary or quaternary.
It can also be used when quantizing to a value.

In the above embodiment, the type of input data is one (one color).
However, the present invention can also be applied to a color image by making the input data R, G, and B colors.

〔The invention's effect〕

As described above, according to the present invention, an image having excellent gradation and resolution can be obtained in a short time with a simple hardware configuration.

[Brief description of the drawings]

FIG. 1 is a diagram showing a multi-valued image for each pixel, a binarized image, and a weighting mask, FIG. 2 is a diagram showing errors that occur during binarization processing, and FIG. 3 is an example of the weighting mask. FIG. 4, FIG. 4 is a block diagram showing the configuration of the image processing apparatus in this embodiment, FIG. 5 is a block diagram showing the details of the binarization circuit D of FIG. 4, and FIG. FIG. 7 is a diagram showing an example of a stored table, FIG. 7 is a diagram showing a case where the weight mask 1 is converted into 6-bit data, and FIG. 8 is a diagram for explaining a second embodiment of the present invention. FIG. 9 is a diagram showing another example of the weight mask. In the figure, 1 and 2 are delay RAM, 3 to 7 are DF / F, and 8 is average density calculation.
ROM, 9 is subtractor, 10 is comparator, 11 is DF / F, 12 is ROM, 13
Is an adder.

Claims (1)

(57) [Claims] 1. An input means for inputting data of a pixel of interest, a calculating means for obtaining an average density value of a predetermined area, and a quantum for quantizing the data of the pixel of interest based on the average density value obtained by the calculating means. An image processing apparatus, comprising: a converting unit; and a correcting unit that corrects the error generated when the quantization is within a predetermined range.