JPH05300365A - Binarizing processing method - Google Patents

Abstract

PURPOSE:To provide the binarizing processing method applying binarizing processing to an original picture with higher definition without deterioration in the picture due to the effect of noise or the like onto the original picture. CONSTITUTION:In the binarizing processing method reading sequentially density data Im,n for each of consecutive picture elements forming a picture and applying the error diffusion method to the read density data to obtain binarized picture data Bm,n from the picture, a threshold level for binarization discrimination 6 with respect to density data of a noticed picture element obtained by being subjected to error diffusion processing 2 is changed based on data 9 obtained by applying smoothing 8 to the density data of the read noticed picture element.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a binarization processing method, and more particularly to a binarization processing method for processing using an error diffusion method.

[0002]

2. Description of the Related Art There is an error diffusion method as a binarizing method for reproducing a halftone in an image scanner or the like. Since this method has no periodicity, the moire phenomenon, which is a problem in other dithering methods and density pattern methods of other binarization methods, does not occur, but in an image with a small change in shading such as a photograph, an output image has a unique texture. Will occur.

In Japanese Laid-Open Patent Publication No. 63-174186, a threshold amount that serves as a reference for binarization determination in the error diffusion method is periodically changed, or based on pixel data (including differential data) of an original image. An image processing method is disclosed in which the threshold amount is automatically variable.

[0004]

In the conventional image processing method as described above, if the original image has noise or the like,
The threshold value changes based on the pixel data of noise, and the influence of moire due to the frequency component of the image is reflected in the threshold value.

The present invention has been made in order to solve the above-described problems, and the original image is enhanced without the influence of noise of the original image or moire due to frequency components of the image being reflected in the threshold value. It is an object to provide a binarization processing method capable of performing binarization processing on the quality.

[0006]

A binarization processing method according to the present invention comprises sequentially reading density data for each continuous pixel forming an image and applying an error diffusion method to the read density data. In the binarization processing method for obtaining binarized image data from an image, a threshold value for binarization determination is read for the density data of the target pixel obtained by performing the error diffusion processing. The density data of the target pixel is changed based on the data obtained by performing the smoothing process.

[0007]

According to the present invention, the threshold value for the binarization determination changes based on the density data of the smoothed target pixel.

[0008]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A first embodiment of the present invention will be described below.

FIG. 1 is a sectional view showing the basic arrangement of a digital image reading apparatus suitable for carrying out the present invention.

In the figure, a scanning optical system is constituted by the first, second and third mirrors 105, 106 and 107 and the projection lens 108. The image of the original 111 or the like placed on the original glass 102 is irradiated by the exposure lamp 104, and the reflected light is imaged on the CCD line sensor 109 via the scanning optical system. The CCD line sensor 109 has a length corresponding to the width of the document in the main scanning direction, and outputs an image signal for one line in the main scanning direction. The scanning in the sub-scanning direction can be achieved by moving the original glass 102 by a driving unit (not shown), or can be performed by a known unit such as moving the scanning optical system. Also, the image reading rate can be changed by changing the projection lens 108 and the original document 1.
This can be done by adjusting the distance between 11 and the distance between the CCD and the lens.

FIG. 2 is a block diagram of image processing of the image reading apparatus of FIG. In the figure, a clock generation circuit 21
Supplies a clock signal to the CPU 20 and CC
A sample hold signal is given to D22. CCD2
Reference numeral 2 is for receiving the reflected light from the image and converting it into an electric signal, and has pixels for one line in the main scanning direction. The A / D converter 23 converts the analog data output from the CCD 22 into digital data. The shading circuit 24 corrects unevenness in the amount of light in the main scanning direction included in the image signal output from the A / D converter 23 and variations in sensitivity between CCD pixels. In the reflectance / density conversion circuit 25, since the digital data processed by the shading circuit 24 is reflectance data, the reflectance data density is adjusted so as to have a linear characteristic with respect to the document density viewed by human eyes. It is converted into data. The binarization circuit 26 is for binarizing the multivalued density data obtained by the reflectance / density conversion circuit 25. In this embodiment, the error diffusion method is applied. .. The printer 27 is for outputting the data binarized by the binarization circuit 26.

FIG. 3 is a block diagram showing a specific structure of the binarization circuit 26 of FIG. The image data (I _{m, n} ) is standardized on the basis of the following equation by multiplying the error data (E _{m, n} ) stored in the buffer memory 1 by the weighting count W _{k, l} in the weighting addition processing unit 2. The data is added by the adder 3 or subtracted by the subtractor 4.

[0013]

[Equation 1]

FIG. 4 is a matrix showing a specific example of the weighting coefficient W _{k, l} . In the figure, S indicates the pixel position currently being processed, and the closer to that position, the larger the value in the matrix. This is because the weighted addition processing unit 2 has error data (E _{m-2, n-2} ) and "1" in the buffer memory 1, (E _{m-1, n-2} ) and "3", (E _{m, n). -2} ) and "5", (E _{m + 1, n-2} ) and "3", and the inserted data is sent to the adder 3 or the subtracter 4. This is because the data in the buffer memory 1 near the pixel position S currently being processed is weighted.

Next, the data added by the adder 3 and
The data subtracted by the subtractor 4 is the select signal S _{m, n
Is} output by the correction data P _{m, n} .

Next, this correction data P _{m, n} is converted into a binarization circuit 6
Is compared with a threshold value T _{m, n} and binary data B _{m, n} is output.

On the other hand, the arithmetic circuit 7 causes the error data E
Output _{m, n} as follows. When P _{m, n} ≧ T _{m, n} , B _{m, n} = 1 (black dots are printed) E _{m, n} = P _{m, n} −I _max (maximum density data) (1) P _{m, n} < When T _{m, n} , B _{m, n} = 0 (no black dot is printed) E _{m, n} = P _{m, n} (2) However, the equation (1) is an error with respect to the black level and is a negative value. Equation (2) is an error with respect to the white level and has a positive value.

The result is stored in the buffer memory at the memory location corresponding to the pixel location S currently being processed. Next, the same process as described above is performed on the pixel data, but in this case, the error _{Em, n in the} buffer memory 1 is shifted to the left by one.
The binarization process can be realized by repeating a series of these operations.

Next, a method of determining the threshold value T _{m, n} for binarizing the correction data P _{m, n} will be described.

The image data I _m, the correction of _n data P _m, at the same time the image data when creating the _n I _m, smoothing processing based on the _n in the smoothing filter shown in FIG. 5 data F _m, seek _n. Then, according to a straight line shown in Figure 6, the smoothing processing data F _m, the threshold for _n T _m, seek _n.

By using a RAM as shown in FIG. 7, the threshold value can be variably set for each pixel in various matrix patterns.

Here, the reason why the smoothing process is performed to determine the threshold value will be described below with reference to FIGS.

FIG. 15 is a diagram showing the contents of binarization processing with a fixed threshold value for a solid image in which no noise is generated.

In a solid image close to a white image in which noise is not generated, as shown in (1) of FIG. 15, the image data I _{m, n} read by the image reader is substantially constant at a value lower than the threshold value. Can be obtained as the value of. On the other hand, the marginal error E in the matrix, which is also used in the error diffusion method,
_{Since m + k and n + l} fluctuate in a constant cycle, (2) in FIG.
As shown in, the error data will exceed the threshold value shown by the broken line in a certain period. Therefore, as for the signal indicating ON / OFF of the dot, the ON signal appears at a constant cycle as shown in (3) of FIG.

FIG. 16 shows a solid image close to a white image.
It is a figure which shows the content at the time of processing with a fixed threshold value when noise occurs.

When the noise "X" exceeding the threshold value is generated only in a certain part of the image, the peripheral error E _{m + k, n + l} in the matrix used for the error diffusion method is affected by this noise and the periodicity is generated. Lose. As shown in (2) of FIG. 16, the influence appears in a wide range from the place where the noise occurs. Therefore, the ON signals of the dots in the vicinity of the portion where noise is generated have no periodicity, and as a result, the reproducibility of a solid image containing noise is deteriorated over a wide range. ..

FIG. 17 is a diagram showing the processing contents when the threshold value is smoothed for an image containing noise similar to that shown in FIG. 16 and changed based on the image data. ..

When the smoothing processing of the threshold value is performed, FIG.
As shown in (1) and (2) of No. 7, the threshold value changes so as to be higher than the threshold values of other portions in the noise generating portion. Therefore, as shown in (2) of FIG. 17, the error data of the part where noise is generated becomes substantially equal to the value of the error data at the point where the threshold value of the other part is exceeded. Therefore, the influence of the portion where the noise is generated on the other portion of the error data is small, and as a result, as shown in (3) of FIG. 17, the area where the dot ON signal generation periodicity is lost. Can be suppressed to be much smaller than that in (3) of FIG. Therefore, the image reproducibility of a solid image including noise is significantly improved as compared with FIG.

Next, the correction data P _m, signals for selecting the _n S _{m, n} will be described. Similarly, when the threshold value T _{m, n} is created, the image data I _{m, n} is set to 2 as shown in FIG.
Edge extraction data EG _{m, n} is obtained using a second derivative filter.

Referring to FIG. 9, edge extraction data EG
_{The m and n} will be described. In this example, as shown in (1) of FIG. 9, it is assumed that a rectangular black solid image with an outline in the center is read as a document.

When such an original image is read by the image reader, the image data I _{m, n} input by the image reader becomes data as shown in (2) of FIG. Edge extraction data of this input image data is calculated using the filter shown in FIG.
The curve is as shown in (3) of FIG. That is,
It can be said that the edge extraction data EG _{m, n} has the following directions.

When the image changes from the white level to the black level
→ edge extraction data changes from "-" to "+"
(A and C parts) When the image changes from black level to white level → edge extraction
Data changes from "+" to "-" (B and D parts
Min) Next, the dot delay block 11 and the edge extraction data EG
_{m, n}For the target position of the second derivative filter
Data EG delayed by the width of the lix (2 dots) _{m-2, n}To
create. It detects the directionality of edge extraction data
This is the minimum amount of delay required to reach Then, the image data I
_{m, n}Error data generated by the weighted addition processing block 2
Data A added by the adder 3_{m, n}Correction data P
_{m, n}Or subtracted data B_{m, n}To P_{m, n}Or
Signal S for selecting_{m, n}Edge direction detection to generate
The output block 12 will be described.

FIG. 10 is a diagram showing a specific configuration of the edge direction detection block 12. Data EG with sign bit
The absolute values of _{m, n} and EG _{m-2, n} are binarized as follows by reference data Ref from the CPU.

| EG _mn | ≧ Ref → “1” | EG _{m, n} | <Ref → “0” | EG _{m-2, n} | ≧ Ref → “1” | EG _{m-2, n} | <Ref → “0” Next, this binarized data and edge extraction data EG
_A signal S _{m, n} is obtained from the sign bits of _{m, n} and EG _{m-2, n} . However, here, the sign bit is set as follows.

Positive → “1” Negative → = 0 ”Therefore, the signal S _{m, n} is − when the direction of the edge changes from negative to positive as shown in FIG.
It becomes "1" only for the side edge extraction data, and the subtraction data B _{m, n} is selected as the correction data P _{m, n} .

That is, when a negative edge is detected when the image changes from white to black, the image data I _{m, n} is subtracted by the weighted error data to obtain the correction data. As a result, a clear image can be obtained without any granular noise on the edge portion when the image changes from the white level to the black level. However, at the edge portion when the image changes from the black level to the white level, the error data of the buffer memory 1 becomes an error with respect to the black level, and takes almost a negative value. Therefore, the weighted addition data in this case has a negative value. Therefore, the data A _{m, n} added by the adder 3 is processed.
If is selected, a sharp image for the edge can be obtained as in the above.

FIG. 12 is a diagram showing a specific example in which the above image processing is performed. As shown in (1) of FIG. 12, a black solid square image of a black level is formed on the white level image as in the case of FIG. 18, and when the image is read by an image reader, It shows the processing contents by the handling in. As shown in (2) of FIG. 12, the image data read by the image reader is the same as that of FIG. Then, based on this image data,
When the edge extraction data is obtained, the edge portion is extracted as shown in (3) of FIG. Based on this edge extraction data, the select signal changes as shown in (4) of FIG. Therefore, since the error data of the dot when the select signal is turned on is treated as subtraction, the error data around the original image by the error diffusion method changes as shown in (5) of FIG. As is apparent from the figure, the error data for the pixel at the left end of the original image, that is, the pixel at the edge portion from the white level to the black level is lower than the threshold value TH indicated by the broken line. As a result, the dot ON signal is output by the width corresponding to the length of the solid black image in the scanning direction as shown in (6) of FIG. 12, and the edge portion of the solid black image is as shown in FIG. A clear image can be obtained without becoming thick.

FIG. 13 shows a second embodiment of the present invention.
3 is a block diagram of the binarization circuit 26. FIG. This embodiment is different from FIG. 3 in the first embodiment in that there are two types of weighted addition processing circuits (2a, 2b) and a selector 4 for selecting one of the processings is provided. ,
Further, the data subjected to the weighted addition processing is always added to the image data, and the subtraction processing as in the previous embodiment is not performed. Therefore, a selector for selecting the operation of the addition or subtraction processing is provided. That is not the point. The configuration and operation of the different points will be described below.

First, the weighting coefficient in the weighting addition process 2a is W _{k, l} . Then, the weighting coefficient in the weighting addition processing 2b, when W _'k, and _l, data X _m was normalized by weighting addition processing 2a and weighted addition processing _{2b, n} and Y _{m, n} are as follows.

[0040]

[Equation 2]

FIG. 14 is a diagram showing a specific example of the weighting factors W _{k, l} and _{W'k, l} . Selector 4 based selection signal S _m, a _n, the data X _{m, n,} or Y _m, selected by selecting the _n data Z _m, and outputs the _n. Selection data Z
_{The m, n} and the input image data I _{m, n} are added by the adder 3 to generate error correction data P _{m, n} .

Next, the weighting coefficient matrices W _{k, l} and W ′ _{k, l} used in the above-mentioned normalization processing will be described based on the buffer memory 1.

The position S of the buffer memory 1 is currently being processed.
The pixel position of
The value of the queue is large for both weighting factors.
It In the weighted addition processing unit 2a, this is
Error data of the far memory 1 (E _{m-2, n-2}) And “1”,
(E_{m-1, n-2}) And “3”, (E_{m, n-2}) And “5”, (E
_{m + 1, n-2}) And “3”, ...
In the weighting addition processing unit 2b,
Error data of 1 (E_{m-1, n-1}) And “1”,
(E_{m, n- 1}) And “3”, (E_{m-1, n}) And "3"
Combined total data X_{m, n}Or Y_{m, n}Is the selector
Select data Z via 4_{m, n}Is sent to the adder 3 as
It This is close to the pixel position S currently being processed
Weighting processing is performed by the data in the buffer memory 1.
This is because.

Next, since the edge direction detection block for producing the signal S _{m, n} for selecting the selection data Z _{m, n} to be added to the image data I _{m, n} is the same as that of the first embodiment, The description will not be repeated. In this example, the signal S _{m, n} output from the edge direction detection block 12 is “1”.
Is added, that is, only when the − side edge extraction data is obtained as in the portions A and C of FIG.
_The data Y _{m, n in} which _{m, n} has been processed by the weighting process is selected.

That is, when an edge in the negative direction when the image data changes from the white level to the black level is detected _{, W'k, l} with a small weighted addition matrix is added to the image data I _{m, n.} Weighted using Y
_{m, n-adds} as Z _{m, n.} As a result, at the edge portion where the image changes from the white level to the black level, the matrix that creates the error becomes smaller, and it becomes difficult to add the error that has occurred around the pixel of interest. Hitting is less likely to occur, making it easier to reproduce clear images.

However, the image changes from the black level to the white level
At the edge part where
The error data for the
Easier to record, in this case a small matrix
Coefficient W ′_{k, l}Data processed by_{m, n}Select to reverse
It is easier to hit extra dots on the edges,
The image may become thick, making it difficult to reproduce a clear image.
Become Therefore, for such an edge,
Matrix coefficient W_{k, l}Data processed by _{m, n}For
Correction data P_{m, n}Creates a clearer image
You can

[0047]

As described above, according to the present invention, since the threshold value for the binarization determination changes based on the density data of the smoothed target pixel, the influence of noise or the like contained in the image is small. It is possible to obtain high-quality binary data.

[Brief description of drawings]

FIG. 1 is a sectional view showing a basic configuration of a digital image reading apparatus according to a first embodiment of the present invention.

FIG. 2 is a system block diagram of image processing of the image reading apparatus of FIG.

3 is a block diagram showing a specific configuration of the binarization circuit of FIG.

FIG. 4 is a matrix showing a specific example of weighting coefficients used in the first embodiment of the present invention.

FIG. 5 is a diagram showing the contents of a smoothing filter used in the first embodiment of the present invention.

FIG. 6 is a graph showing a relationship between smoothed data and a threshold value according to the first embodiment of the present invention.

7 is a diagram showing a configuration of a ROM for generating a threshold value based on the graph of FIG.

FIG. 8 is a diagram showing the contents of a secondary differential filter for obtaining edge extraction data according to the first embodiment of the present invention.

FIG. 9 is a diagram showing specific contents of edge extraction processing according to the first embodiment of the present invention.

10 is a diagram showing a specific configuration of the edge direction detection block of FIG.

FIG. 11 is a diagram showing the contents of processing from the reading of a solid image to the generation of dot signals according to the first embodiment of the present invention.

FIG. 12 is a diagram showing the specific content of edge extraction processing for a black solid image according to the first embodiment of the present invention.

FIG. 13 is a block diagram showing a specific configuration of a binarizing circuit according to a second embodiment of the present invention.

FIG. 14 is a diagram showing specific contents of weighting coefficients according to the second embodiment of the present invention.

FIG. 15 is a diagram showing the contents of a binarization processing method for a solid image close to a white level by a general error diffusion method.

FIG. 16 is a diagram showing a problem in processing by a conventional error diffusion method for an image containing noise.

FIG. 17 is a diagram showing an improvement content of the processing of the error diffusion method according to the present invention for an image including noise.

FIG. 18 is a diagram showing a problem with a conventional error diffusion method for a black solid image.

[Explanation of symbols]

1 Buffer Memory 2 Weighted Addition Processing Unit 3 Adder 4 Subtractor 5 Selector 6 Binarization Circuit 7 Operation Circuit 8 Smoothing Filter 9 Conversion Table 10 Secondary Derivative Filter 11 Dot Delay Circuit 12 Edge Direction Detection Block 2a Weighted Addition Processing Unit 2b Weighted addition processing unit In the drawings, the same reference numerals indicate the same or corresponding portions.

Claims (1)

[Claims] 1. A binarization process for obtaining binarized image data from an image by sequentially reading density data of continuous pixels forming an image and applying an error diffusion method to the read density data. In the method, a threshold value for binarizing the density data of the pixel of interest obtained by performing the error diffusion process is obtained by performing a smoothing process on the density data of the read pixel of interest. A binarization processing method that changes based on the obtained data.