JP2553799B2 - Image processing device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus used in facsimiles, scanners and the like.

[0002]

2. Description of the Related Art Conventionally, as shown in FIG. 14, an apparatus of this type has an input terminal 1 for inputting an image signal, an adder 2, a threshold value generating section 3, a comparator 4 for binarizing, and a subtracter. 6
An error memory 7 for storing an error, an error filter 8 for weighted addition, and an output terminal 5.

An adder 2 is added to the image signal input from the input terminal 1.
In, the output from the error filter 8 is added, and the output of the adder 2 is binarized by the threshold value generated by the threshold value generation unit 3 in the comparator 4. In the subtractor 6, the output of the adder 2 is subtracted by the output of the comparator 4 and the error data is output.

The error data output from the subtractor 6 is stored in the error memory 7. The error data in the peripheral plural pixels which have already been binarized are weighted and added in the error filter 8 and output, and are added to the input image signal by the adder 2 as described above. The output of the comparator 4 is output from the output terminal 5. Due to the above processing, the gradation reproducibility is high,
Pseudo-halftone images with excellent resolution were obtained.

[0005]

However, such a configuration has the following problem of degrading image quality. Problem 1. If the size of the dots in the recording system that records the pseudo-halftone image is larger than the theoretical size (hereinafter referred to as "blurred") or smaller (hereinafter referred to as "blurred"), the number of reproducible gradations will decrease. The gradation reproducibility was deteriorated, for example, the continuity of gradation was weakened. FIG. 13 shows an example of dot collapse. Input image signal density 50% ((a))
On the other hand, the ideal pseudo halftone has the same black area and white area as shown in FIG. When the dots are crushed, the black area increases as shown in (c), and the density becomes 40%.

Problem 2. Due to the effect of the image outside the effective range of the input image signal, black dots may occur at the upper and left ends of the document even if the background is white. Problem 3. In the area where the change in the density level is small, a dot-continuous texture in which the dots are continuous in the main scanning direction, the sub-scanning direction, or the diagonal direction occurs.

The above-mentioned problems occur for the following reasons. Reason for occurrence of problem 1 The error diffusion process is performed by using the input image data and this input image data.
Error that is the error from the output image data when binarizing
By spreading the data to the image data of the surrounding pixels,
Sum of input image densities Σf in the local area_mnOutput image of
Total concentration Σg _mnIs a pseudo halftone process that makes
However, if the recorded dots are crushed or blurred, Σf_mn=
Σg_mn It was because that would not hold.

In order to solve this problem, as shown in FIG. 15, a gamma conversion unit 88 and an error diffusion processing unit 89 which perform density conversion.
It is conceivable that the gamma conversion unit 88 performs density conversion in advance to darken or lighten the input image, and the error diffusion processing unit 89 performs error diffusion processing. , The following problems occurred and it was not an appropriate solution.

In the gamma conversion unit 88, the gradation number of the output image signal is smaller than the gradation number of the input image signal,
Gradation reproducibility of pseudo halftone cannot be corrected sufficiently.

The problem 1 is solved by increasing the number of gradations of the image signal input to the gamma conversion unit 88 (that is, increasing the number of bits), but increasing the number of bits of the input image signal. Therefore, it is necessary to increase the number of bits of the image signal processing unit up to the stage before the gamma conversion unit 88, which causes a new problem of an increase in processing scale.

When the level of the input image signal is monotonically increased, if it is a theoretical dot, the gradation level of the output image signal will also monotonically increase, but if the dot is crushed or blurred. The gradation level of the output image signal does not monotonically increase, and fine correction by gamma conversion is difficult.

Since the degree of dot collapse and blurring varies depending on the type of recording system, a plurality of gamma conversion units 88 are required correspondingly, and the processing scale increases.

Reason for Occurrence of Problem 2 In the error diffusion processing, error data which is an error between the input image data and the output image data when the input image data is binarized is diffused to the image data of the peripheral pixels. So
This is because the error data in the area outside the effective range of the input image signal input from the sensor or from the memory is propagated and added to the pixels within the effective range.

Reason for Occurrence of Problem 3 As shown in FIG. 16, the operation of the error diffusion processing is that the main scanning direction is constant from left to right and the sub scanning direction is constant from top to bottom. This is because it is fixed to the lower right and the correlation between adjacent pixels becomes large.

This may occur due to the size of the error filter for performing weighted addition of errors and the accuracy of the coefficients being not sufficiently large.

The above can be dealt with by, for example, randomly switching the processing direction of the main scanning for each line from left to right, from right to left, or by zigzag scanning like a Peano curve. , Can be dealt with by increasing the size of the error filter and the accuracy of the coefficient, but in either case, the scale of the processing is increased, which causes a problem of cost increase. Not a solution.

The present invention has been made in view of the above problems, and an object of the present invention is to provide an image processing apparatus which corrects image deterioration due to binarization processing and reproduces high-quality pseudo halftone.

[0018]

To achieve the above object, an addition means for adding a binarized error diffusion value to input image data, a binarization means for binarizing the output of the addition means, and A difference value obtained by accumulating the output of the binarizing means and the difference means for calculating the difference value of the output of the binarizing means and the adding means, and correcting the difference value based on the data of the pixel of interest and its peripheral pixels. The first correction means for outputting correction data and the binarized error diffusion value created by weighting the difference value correction data corresponding to the peripheral pixels of the next target pixel and then integrating the difference value correction data. The error diffusion value calculating means for outputting to the adding means is provided.

Adder means for adding the binarized error diffusion value to the input image data, binarizing means for binarizing the output of the adder means, and the difference value between the output of the binarizer and the adder means. And difference value correction data obtained by correcting the difference value according to the difference between the actual dot size and the theoretical dot size in the recording system when the output image of the binarizing unit is recorded. With respect to the second correction means for outputting and the difference value correction data, the binarized error diffusion value created by weighting and integrating the difference value correction data corresponding to the peripheral pixels of the next target pixel is added to the adding means. The error diffusion value calculating means for outputting is provided.

Addition means for adding the binarized error diffusion value to the input image data, binarization means for binarizing the output of the addition means, and the difference value between the output of the binarization means and the addition means. When the output of the binarizing means is recorded by accumulating the outputs of the difference means for computing and the binarizing means, selecting a predetermined correction pattern based on the data of the pixel of interest and its peripheral pixels. The correction amount of the selected correction pattern is corrected according to the difference between the actual dot size and the theoretical dot size in the recording system, and the difference value correction data in which the difference value is corrected with this corrected value is output. With respect to the third correction means and the difference value correction data, the binarized error diffusion value created by weighting and integrating the difference value correction data corresponding to the peripheral pixels of the next target pixel is output to the adding means. Error diffusion value calculation means Those were example.

Further, the error diffusion value calculation means reads the difference value correction data corresponding to the peripheral pixels of the pixel of interest from the storage means for storing the difference value correction data, and corresponds to each peripheral pixel. An error filter is provided for outputting the binarized error diffusion value created by weighting the difference value correction data and then integrating the weighted difference data.

Further, an input image valid signal indicating whether the input image is valid or invalid is input to the third correcting means, the difference value correction data is output in the case of a valid signal, and the difference value in the case of an invalid signal. The correction data is not output.

The validity of the error filter input image,
The input image valid signal indicating invalidity is input, the binary error diffusion value is output in the case of the valid signal, and the binary error diffusion signal is not output in the case of the invalid signal.

Further, the error diffusion value calculation means weights the difference value correction data and sequentially stores the weighted difference value in the memory, and sequentially reads out the difference value correction data from the memory and weights the difference value correction data corresponding to the peripheral pixels of the target pixel. The data are integrated and output as the binarized error diffusion value to the adding means.

Further, an input image valid signal indicating whether the input image is valid or invalid is input to the error diffusion value calculation means, and in the case of a valid signal, the difference value correction data is output to the error diffusion value calculation means and invalidated. In the case of a signal, the difference value correction data is not output.

Further, a plurality of sets of weighting coefficients are provided for weighting the difference value correction data, and a random number generator for selecting the plurality of sets is provided with a plurality of one-dimensional random number generators and one-dimensional random number generators. It is composed of an output selection unit for selecting an output.

[0027]

With the above configuration, the output of the binarizing means is accumulated,
The pixel of interest and its peripheral pixel data are read out from this, a correction pattern determined by the white and black distributions of this pixel of interest and its peripheral pixels that have been determined in advance is selected, and the difference value of the difference means is corrected based on this correction pattern to obtain the difference. As the value correction data, the error diffusion value calculation means weights the difference value correction data corresponding to the peripheral pixels of the next target pixel and then integrates them, and outputs this value to the addition means as the binarized error diffusion value. As a result, dot collapse and blurring are corrected.

Further, the difference value is corrected according to the difference between the actual dot size and the theoretical dot size in the recording system when the output image of the binarizing means is recorded, and the difference value correction data is obtained. The diffusion value calculation means weights the difference value correction data corresponding to the peripheral pixels of the next target pixel, accumulates the weighted difference data, and outputs this value to the addition means as a binarized error diffusion value. As a result, dot collapse and blurring are corrected.

Further, the output of the binarizing means is accumulated, the target pixel and its peripheral pixel data are read therefrom, and a predetermined correction pattern determined by the white and black distributions of the target pixel and its peripheral pixels is selected. The selected correction pattern is corrected according to the difference between the actual dot size and the theoretical dot size in the recording system when the output image of the binarization unit is recorded, and the corrected correction pattern is added. Based on this, the difference value of the difference means is corrected to obtain difference value correction data, and the error diffusion value calculation means weights the difference value correction data corresponding to the next pixel of interest and its peripheral pixels and then integrates the weighted values. The converted error diffusion value is output to the adding means. As a result, dot collapse and blurring are corrected.

Further, the storage means stores the difference value correction data, and the error filter reads the difference value correction data corresponding to the peripheral pixels of the pixel of interest from the storage means to obtain difference value correction data corresponding to each peripheral pixel. After weighting, the binarized error diffusion value calculated by integration is output to the adding means.

Further, in the above-mentioned error diffusion processing, the difference value is diffused into the image data of the peripheral pixels, but the difference value in the area outside the effective range of the input image signal input from the sensor or the memory is propagated and is within the effective range. However, black dots may be generated at the upper edge or the left edge of the document by adding to the pixels of No. 2. However, only the valid ones are output by the input image valid signal to prevent the generation of these black dots.

This input image valid signal is input to the third correction means to control the output of the difference value correction data, or to the error filter to control the output of the binarized error diffusion signal. Have an effect.

Further, instead of storing the difference value correction data in the storage means and then weighting it by the error filter as described above, the error diffusion value calculation means first weights it and then stores it in the memory, and the periphery of the pixel of interest. The same effect can be obtained by reading the stored data corresponding to the pixel, integrating it, and outputting it as the binarized error diffusion signal to the adding means.

At this time, by inputting the input image valid signal to the error diffusion value calculation means and controlling the output of the error diffusion value calculation means, it is possible to prevent the generation of black dots at the upper end or the left end of the document.

In order to calculate the binarized error diffusion value by weighting the difference value correction data, a plurality of weighting coefficient sets are provided and selected by a random number generated by a random number generator. By selecting using a 2-dimensional random number generator of a small size, even in an area with a small change in density level, the generation of a texture peculiar to error diffusion processing in which dots are continuous in the main scanning direction, the sub scanning direction, or the diagonal direction Can be prevented.

[0036]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a diagram showing a schematic configuration of a first embodiment of the present invention. In the figure, 1 is an input terminal for inputting an image signal, 2 is an adder for adding the output of an error filter 8a to be described later to the image signal, 3 is a threshold value generating section for generating a threshold value, and 4 is a comparator for performing binarization. 5 is an output terminal for outputting an image signal, 6 is a subtracter for subtracting the output of the adder 2 with the output of the comparator and outputting error data, 7 is a memory for error data, 8a is a two-dimensional pseudo random number generation described later. This is an error filter that performs weighted addition of error data by switching coefficients according to a pseudo-random number output from the unit 21.

9 is a recording dot correction coefficient memory for k = 0.5 (k will be described later), 10 is a recording dot correction coefficient memory for k = 0.3, and 11 is a recording dot correction coefficient for k = 0.1. A coefficient memory, 12 is a recording dot correction coefficient memory of k = 0, and 13 is an output of various recording dot correction coefficient memories 9 to 12 selected by an output of a recording dot correction coefficient selection signal input terminal 19 which will be described later. A selector for outputting to the error correction amount determination unit 14, 14 is an error correction amount determination unit for inputting a correction coefficient in advance by the selector 13 and determines an error correction amount by the output of an output image memory 18 described later, and 15 is a subtractor 6 The adder that adds the error correction amount that is the output of the error correction amount determination unit 14 to the error data that is the output of 16 is the input terminal of the input image valid signal that indicates whether the input image signal is valid, and 17 is Output and input of adder 15 If the input image signal is not valid inputs the output of the child 16, the error data correction unit for forcibly zero error.

Reference numeral 18 is an output image memory, 19 is a recording dot correction coefficient selection signal input terminal, 20 is error data E (a), E (b), E (c), E (d), 21 output from the memory. Is a two-dimensional pseudo random number generator, and 24 is a 5-bit signal of binary signals a, b, c, d and p output from the output image memory 18.

The operation will be described below. An output SE from an error filter 8a, which will be described later, is added to the image signal PIX of the pixel of interest p input from the input terminal 1 in the adder 2, and the output of the adder 2 is input to the threshold generator 3 in the comparator 4.
It is binarized with the threshold value generated in. The binarized output image signal B of the target pixel p is stored in the output image memory 18, input to the subtractor 6, and further output from the output terminal 5.
The subtractor 6 subtracts the output of the adder 2 with the output of the comparator 4, and outputs this as error data e.

The recording dot correction coefficient k corresponding to the difference between the actual dot size and the theoretical dot size of the recording system, which will be described later, is recorded at k = 0.5, 0.3, 0.1, 0. The dot correction coefficient memories are 9, 10, 11, and 12, respectively, and before the error diffusion processing is performed, the selector 13 previously selects the recording dot correction coefficient selection signal input terminal 19 according to the selection signal SLC. 9,
Select the output of 10, 11, 12 and output cd as the error correction amount determination unit
It writes in RAM table 14 mentioned later.

The recording dot correction coefficient selection signal SLC input from the input terminal 19 is 9, 10, 1 in accordance with the presence or absence and the degree of dot collapse in the recording system on the local side or the receiving side.
This is a signal for selecting the correction coefficient memory of 1 and 12, and if the type of recording system is unknown, the correction coefficient of 12 is k =
Select a correction factor of 0.

The error correction amount determining unit 14 outputs the output image memory 18
Further, the binary signal (total 5 bits) of the pixel of interest p and its neighborhoods a, b, c and d is input and the error correction amount c is generated by referring to the RAM table. The error correction amount c generated by the error correction amount determination unit 14 is added to the error data e which is the output of the subtractor 6 in the adder 15, and is output as the correction error data e1.

The correction error data e1 is input to the error data correction unit 17, and according to the input image valid signal SLE input from the input terminal 16, if the SLE indicates that the input image signal is valid, then e1 is invalid. When it indicates that, zero is forcibly output as the signal e2.

The output e2 of the error data correction unit 17 is stored in the error data memory 7. The error data 20 stored in the error data memory 7 (E (a), E (b), E (c),
E (d)) is weighted and added by switching coefficients according to the 1-bit pseudo random number QR generated by the two-dimensional pseudo random number generation unit 21 in the error filter 8a, and is output as SE. SE is added to the input image signal in the adder 2 as described above. The number of pixels that diffuse the error is 4 in this embodiment.
Although it is a pixel, it is not limited to four pixels.

Next, a second embodiment will be described with reference to FIG.
2, the same reference numerals as those in FIG. 1 represent the same members. First
The difference from the embodiment is that the error data correction unit 17 in the first embodiment does not perform the error data correction processing based on the input image valid signal SLE, but the error filter 8b that performs error weighting addition.

An error filter 8b performs weighted addition of error data by switching coefficients according to the pseudo random number output from the two-dimensional pseudo random number generator 21 and the input image valid signal output from the input image valid signal input terminal 16. Is.

In the error filter 8b, the error data 20 stored in the error data memory 7 and the two-dimensional pseudo random number generator
1-bit pseudo-random number QR generated in 21 and the input image valid signal SLE input from the input image valid signal input terminal 16
When the SLE indicates that the input image signal is valid, the pseudo-random number QR is used, and when QR is 1, the error filter coefficient is set to a coefficient sequence (ka1, kb1, kc1, k) described later.
d1), when QR is 0, the coefficient sequence (ka2, kb2, kc2, kd2) described later is selected as the error filter coefficient, and when the SLE indicates that the input image signal is invalid, the error filter coefficient is set to (0, 0, 0, 0) is selected to perform weighted addition processing of the error data 20.

The SLE is the same as in the first embodiment. In the present embodiment, the pseudo random number is 1 bit, and the error filter coefficients to be switched are two sets, but the number of bits of the pseudo random number may be increased to increase the number of error filter coefficient sets to be switched.

Next, a third embodiment will be described with reference to FIG.
3, the same reference numerals as those in FIG. 1 represent the same members.

Reference numeral 22 is an error filter for switching the coefficients according to the pseudo-random number output from the two-dimensional pseudo-random number generator 21 to perform weighted addition of error data, and reference numeral 23 is an error integrated operation unit consisting of the error filter 22 and the error data memory 7. Is.

The difference of the present embodiment from the first embodiment is that the error data is accumulated by the error filter 22. In the present embodiment, the error data memory 7 is read and written repeatedly to perform the error data integration calculation. The detailed operation of the error accumulation calculator 23 will be described below with reference to FIG.

FIG. 9 is a detailed configuration diagram of the error integrated calculation unit 23. 76 is a pseudo random number input terminal, 77 is an error data input terminal,
78 is B memory, 79 is B + C memory, 80 is B + C + D
Memory, 81 is an A + B + C + D memory, and 82 is an output terminal of the error integrated calculation unit.

A 1-bit pseudo random number QR is input from the pseudo random number input terminal 76. In the selectors 49, 50, 51 and 52, the error filter coefficients are (ka1, kb1, kc1, kd1) and (ka2, kb) in accordance with QR as in the second embodiment.
2, kc2, kd2) and output.

The corrected error data is input from the input terminal 77, multiplied by the coefficient kb (kb1 or kb2) selected by the selector 49 by the multiplier 53a, and the address of E (B) in the memory 78 is stored. Written in.

Further, the corrected error data is the multiplier 54a.
Is multiplied by the coefficient kc (kc1 or kc2) selected by the selector 50, and (B) E (B + of the memory 78
The data read from the address C) is added by the adder 57a and written to the (B + C) address of the (B + C) memory 79.

Further, the corrected error data is multiplied by the coefficient kd (kd1 or kd2) selected by the selector 51 by the multiplier 55a, and E (B + C) of the (B + C) memory 79 is obtained.
The data read from the address of (C + D) is added by the adder 58.
a is added, and (B + C + D) of the memory 80 (B + C +)
D) Written to address.

Further, the corrected error data is multiplied by the coefficient ka (ka1 or ka2) selected by the selector 52 by the multiplier 56a, and E of the (B + C + D) memory 80 is obtained.
The data read from the address (A + B + C + D) is added by the adder 59a, and the (A + B + C + D) memory 81
(A + B + C + D) address of

The data read from the (A + B + C + D) memory is output from the output terminal 82 as a weighted average integrated error SE.

Next, the recording dot correction coefficient memory 9 to the adder
15. The correction of error data for correcting the collapse of the recording dots composed of the output image memory 18 will be described in detail.

First, the principle will be described. As a first example, FIG. 11 shows a case where the peripheral pixels are all white and the target pixel is black. The pixel of interest is p, and the peripheral pixels are a, b, c, and d. In the shaded area, the actual black dots are the surrounding pixels a,
It is a region that is a margin for b, c, and d. The correction of the area of the dotted line is not performed on the pixel of interest, but the pixels e, f, and
Perform at g and h.

Assuming that the width of the region that is surrounded as shown in FIG. 11 is k (correction coefficient) and the theoretical dot width is 1, the area of the shaded area is S = 2k + 2k ² . Further, when the white image signal level of the input image signal is W, the correction amount C is as follows.

C = (2k + 2k ² ) × W

The corrected error data e1 for the error data e is as follows. e1 = e + (2k + 2k ² ) × W

As a second example, FIG. 12 shows a case where all the peripheral pixels are black and the target pixel is white. The hatched area is an area in which the actual white dots are eroded from the peripheral pixels a, b, c, and d to become black. Similar to the first example, the shaded area is as follows.

S = 2k-k ²

The correction amount C is as follows. C = (2k−k ² ) × W Therefore, the correction error data e1 is as follows for the error data e.

[0068] e1 = e + (2k-k 2) × W

Similarly, the target pixel p, the peripheral pixel a,
When b and c are other than that, the correction amount C is determined, and each C
On the other hand, the correction error data e1 is as follows.

E1 = e + C

These are summarized as shown in Tables 1 and 2.

[0072]

[Table 1]

[0073]

[Table 2]

Based on the above principle, the operation of the error data correction unit based on recording dots will be described. First, k = 0.5,
The data values corresponding to Table 1 and Table 2 at the time of 0.3, 0.1, 0 are recorded dot correction coefficient memories 9, 10, 11, 12 respectively.
Stored in.

Next, a recording dot correction coefficient selection signal input terminal
The recording dot correction coefficient selection signal SLC is input from 19
Depending on the presence or absence of dot spread in the local copy system or the recording system on the receiving side, if the dot spread is large, the correction coefficient memory 9 with k = 0.5 is used. If the dot spread is slightly large, k = 0.3 is used. The correction coefficient memory 10 is a correction coefficient memory 11 with k = 0.1 when the dot spread is small.
If there is no dot spread or it is unknown, the selector 13 selects the correction coefficient memory 12 with k = 0 and outputs it to the error correction amount determination unit 14 as the signal cd.

Next, the error correction amount determining unit 14 will be described with reference to FIG. In FIG. 4, 25 is a correction coefficient input terminal,
26 is a selection signal input terminal for a RAM address signal described later, 27 is a binary signal p input terminal, 28 is a binary signal a input terminal, 29 is a binary signal b input terminal, 30 is a binary signal c input terminal, 31 is a binary signal d input terminal, 32a is a write pulse generator for RAM described later, 32b is a write address generator for RAM, 33 is the output of 32b and the 5-bit output of 27, 28, 29, 30, 31. A selector for selecting and outputting as an address of a RAM 34, which will be described later, stores a correction coefficient from the correction coefficient input terminal 25, and outputs a binary signal a, b, c, d, or p.
RAM for outputting correction amount for bit input, 35 for RA
This is the output terminal of M34.

Next, the operation will be described. The correction data is written in the RAM 34 from the input terminal 25. At this time, a write pulse signal is input from the write pulse generator 32a to the RAM 34, a write address is output from the write address generator 32b, and a signal for selecting a write address is generated from the address selection signal generator 26. The output of the write address generator 32b is selected by the selector 33 and input to the RAM 34 as an address signal.

During the error diffusion processing operation, the output image memory 18
A total 5 bit signal of the binarized signal of the pixel of interest p and its neighboring 4 pixels a, b, c and d is input from the input terminals 27 to 31 of FIG. 4 and the correction amount C of the already written RAM table is shown. Output from the output terminal 35 of 4. The correction amount C is added to the error data e which is the output of the subtractor 6 in the adder 15 to become the correction error data e1.

In the above embodiment, k = 0, 0.1, 0.
Although the correction amounts of 3, 0.5 are stored, selected, and processed, the correction coefficient memory to be selected may be increased or decreased. Further, although the crushed dots are modeled as squares, the correction tables (Tables 1 and 2) may be determined and implemented as circles instead of squares.

Further, since the peripheral pixels in the oblique direction (b, d in FIG. 11) have a small effect on the target pixel, only the peripheral pixels (a, c in FIG. 11) adjacent in the main scanning direction and the sub-scanning direction are affected. The scale of the process may be reduced with reference to the above. Further, in the present embodiment, the correction is made for the one in which the recording dots are crushed, but the same can be applied to the recording system in which the recording dots are faint.

Further, although the present embodiment is directed to the recording system in which the recording dots are crushed in the two-dimensional direction, the same concept is applied to the recording system in which the recording dots are crushed only in one direction such as the main scanning direction or the sub-scanning direction. Can be corrected with.
The above is the operation of correcting the error data for the collapse of the recording dots.

Next, the detailed operation of the error data correction unit 17 will be described. The error data correction unit 17 inputs the input image valid signal SLE from the input terminal 16 and the corrected error data e1 from the adder 15 and outputs e2 according to the logic of Table 3. That is, when the input image signal is invalid and SLE = 0, the output e2 = 0 regardless of the value of e1. When the input image signal is valid and SLE = 1, the output is e2 = e + c. Here, e is input error data.

[0083]

[Table 3]

The case where the input image signal is invalid means that the leading edge of the document does not reach the reading position of the sensor when the image signal is input to the image processing apparatus by the document reading by the sensor, or the image of the sensor is read. This is the case when the data of the signal invalid area is output, or when the output of the image memory when the image memory is input to the image processing apparatus is not valid.

The above is the operation of the error data correction unit 17.

Next, the operation of the error filter 8a will be described in detail. FIG. 5 shows an example of the configuration of the error filter 8a. In the figure, 36 is a pseudo random number input terminal, 37 is an error data E (a) input terminal, 38 is an error data E (b) input terminal, 39 is an error data E (c) input terminal, and 40 is an error data E ( d) Input terminals, 41, 42, 43, 44, 45, 46, 47, 48 are error filter coefficient setting units, and 49, 50, 51, 52 are 41, 4 described above.
Selectors for selecting outputs of 2, 43, 44, 45, 46, 47, 48, 53a, 54a, 55a, 56a are multipliers, 57a, 58a, 59
a is an adder and 60a is an error filter output terminal.

Next, the operation will be described. Input 1-bit pseudo-random number QR from 36 and input terminals 37, 38, 39, 40
From the error data E (a), E (b), E (c), E
Input (d) respectively. 41 to 48 are error filter coefficients ka1, ka2, kb1, kb2, kc1, respectively.
kc2, kd1 and kd2 are stored. Selector 4
9, 50, 51, 52 are QR filter with QR = 1, the error filter coefficient is (ka1, kb1, kc1, kd1), QR
When = 0, it is selected and output from (ka2, kb2, kc2, kd2).

Table 4 shows the above contents.

[0089]

[Table 4]

The error data E (a) is the coefficient ka (ka1 or ka selected by the selector 49 by the multiplier 53a).
It is multiplied by 2) and input to the adder 57a.

The error data E (b) is the coefficient kb (kb1 or kb selected by the selector 50 by the multiplier 54a.
It is multiplied by 2) and input to the adder 57a.

The error data E (c) is the coefficient kc (kc1 or kc selected by the selector 51 by the multiplier 55a).
2) is multiplied and input to the adder 58a.

The error data E (d) is the coefficient kd (kd1 or kd selected by the selector 52 by the multiplier 56a.
2) is multiplied and input to the adder 58a. The output of 57a and the output of 58a are added by the adder 59a, and output terminal 60
The signal SE is output from a.

Next, another configuration example of the error filter 8a is shown in FIG.
Will be explained. The symbols in the figure have the same meanings as in FIG. Incidentally, 53b, 54b, 55b, 56b are multipliers, and 57b, 58.
Reference numerals b and 59b are adders, and reference numeral 60b is a selector that switches the outputs of 59a and 59b by the random number input from 36.

A 1-bit pseudo random number QR is input from 36, and error data E (a), E is input from input terminals 37, 38, 39, 40.
Input (b), E (c) and E (d) respectively. 41 ~
48 are error filter coefficients ka1, ka2, kb, respectively.
1, kb2, kc1, kc2, kd1, kd2 are stored.

The error data E (a) is multiplied by the coefficient ka1 by the multiplier 53a and input to the adder 57a. Further, the error data E (a) is calculated by the coefficient ka2 by the multiplier 53b.
Is multiplied by and input to the adder 57b. Error data E
(B) is multiplied by the coefficient kb1 by the multiplier 54a,
It is input to the adder 57a. Further, the error data E (b) is multiplied by the coefficient kb2 by the multiplier 54b, and the adder 57
Input to b. The error data E (c) is multiplied by the coefficient kc1 by the multiplier 55a and input to the adder 58a.

The error data E (c) is multiplied by the coefficient kc2 by the multiplier 55b and input to the adder 58b. The error data E (d) is converted into a coefficient kd by the multiplier 56a.
It is multiplied by 1 and input to the adder 58a. The error data E (d) is multiplied by the coefficient kd2 by the multiplier 56b and input to the adder 58b. The output of 57a and the output of 58a are added by the adder 59a and input to the selector 60b,
The output of 57b and the output of 58b are added by the adder 59b and input to the selector 60b.

The selector 60a is an adder 59a when QR = 1.
Is selected, and when QR = 0, the output of the adder 59b is selected and output as a signal SE. The above is the error filter 8
This is the operation of a.

Next, the error filter 8b shown in FIG. 2 representing the second embodiment will be described.

The difference from the error filter 8a is that the input image effective signal SLE is input and the coefficient of the error filter is switched by the pseudo random numbers QR and SLE.

FIG. 10 is a block diagram of the error filter 8b.
In the figure, reference numerals 83, 84, 85, 86 denote zero error filter coefficient setting units, and 87 denotes an input image valid signal input terminal.

Input image valid signal SLE from 36-bit 1-bit pseudo-random number QR, 87, and input terminals 37, 38, 39, 40.
From the error data E (a), E (b), E (c), E
Input (d) respectively. 41 to 48 are error filter coefficients ka1, ka2, kb1, kb2, kc1, kc2.
kd1 and kd2 are stored. Error filter coefficient 0 is stored in 83 to 86.

The selectors 49, 50, 51 and 52 follow the pseudo random number QR when the SLE indicates that the input image signal is valid (SLE = 1), and when the QR is 1, the error filter coefficient is set to the coefficient sequence (ka1, kb1, kc1, kd1), and when QR is 0, the error filter coefficient is set to a coefficient sequence (ka2, kb2, k).
c2, kd2) and output. Also, selector 4
When the SLE indicates that the input image signal is invalid (SLE = 0), reference numerals 9, 50, 51 and 52 select and output a coefficient string (0, 0, 0, 0) as an error filter coefficient.

The error data E (a) is the coefficient ka (ka1 or ka2) selected by the selector 49 by the multiplier 53a.
Or is multiplied by zero) and input to the adder 57a.

The error data E (b) is the coefficient kb (kb1 or kb2 selected by the selector 50 by the multiplier 54a.
Or is multiplied by zero) and input to the adder 57a.

The error data E (c) is the coefficient kc (kc1 or kc2 selected by the selector 51 by the multiplier 55a).
Or it is multiplied by zero) and input to the adder 58a.

The error data E (d) is the coefficient kd (kd1 or kd2 selected by the selector 52 by the multiplier 56a).
Or it is multiplied by zero) and input to the adder 58a.

The outputs of 57a and 58a are added by the adder 59a and output from the output terminal 60a as the signal SE.

The above is the operation of the error filter 8b, which is summarized in Table 5.

[0110]

[Table 5]

Next, the operation of the two-dimensional pseudo random number generator 21 will be described. A suitable method is already known as a method for generating a one-dimensional pseudo random number. For example, FIG. 8 shows an example thereof (digital IC practical manual: Radio Engineering Co., Ltd.).

In the figure, 70 is an image signal clock input terminal, 71 is an A bit for inputting a clock and an output A from 74, which will be described later, and outputting a signal of a total of A bits from 1 bit shift to A bit shift. A shift register, 72 is a NOR gate for inputting the A-bit signal of 71 to obtain an inversion OR, and 73 is an output of 71 for the A-bit shift signal and A-
An exclusive OR gate for inputting a 1-bit shift signal for exclusive OR, 74 is an exclusive OR gate for inputting the output of 72 and the output of 73 for exclusive OR, and 75 is A This is the output terminal of the bit pseudo-random number generator.

The A-bit shift register 71 inputs the output of the exclusive OR gate 74 as a signal for shifting, and also inputs the image signal clock from the input terminal 70 to obtain a total A from the 1-bit shift signal to the A-bit shift signal. Outputs a bit shift signal. The NOR gate 72 inputs the output of A bit of the A bit shift register 71 and performs an inversion OR.

The exclusive OR gate 73 inputs the A bit shift signal and the A-1 bit shift signal of the output of the 71, and performs an exclusive OR. The exclusive OR gate 74 inputs the output of 73 and the output of 72 and outputs the output.
Enter in 71. By the above operation, a periodic signal of 2 ^A -1 period is generated, and by making A sufficiently large, it functions as a pseudo random number generator. However, since this pseudo random number is one-dimensional, it becomes a stripe pattern when used as a random number of an image, and it did not function sufficiently as a pseudo random number for an image.

In the first to third embodiments, the two-dimensional pseudo random number generation is realized by the following method. FIG. 7 shows the configuration of the two-dimensional pseudo random number generator. In the figure, 61 is an image signal clock generator that generates an image signal clock, 62 is an A-bit pseudo random number generator, 63 is a B bit pseudo random number generator, 64 is a C bit pseudo random number generator, and 65 is a Z bit pseudo random number generator. Random number generator, 66 is 62
Output QRA, 63 output QRB, 64 output QRC, ...
A selector that selects and outputs from the output QRZ of 65 by a selection signal of a pseudo random number selection unit 68 described later, 67 is a line synchronization signal generation unit, and 68 is a pseudo random number selection unit that inputs the output of 67 and outputs a selection signal, 69 is an output terminal of the selector 66.

The image signal clock generated from the image signal clock generator 61 is an A bit pseudo random number generator 62, a B bit pseudo random number generator 63, a C bit pseudo random number generator 64, ..., A Z bit pseudo random number generator 65. Entered in. 62, 63, 64, 65,
Then pseudo random numbers QRA, QRB, Q
It outputs RC, ..., QRZ.

In the pseudo random number selection unit 68, for each line to which the line synchronization signal is input from the line synchronization signal generation unit 67,
Output the selection signal of the output of 65. In the selector 66, the pseudo random numbers QRA, QRB, QRC, ..., QRZ are input, selected by the selection signal of 68, and output from the output terminal 69. The pseudo-random number generator of FIG. 8 may be used for 68 as well.

With the above configuration, a random pseudo random number is generated in the two-dimensional direction.

[0119]

As is apparent from the above description, the present invention has the following effects. Increasing the number of input bits of the image processing unit by correcting the difference value of input / output data of the binarization comparator in consideration of the difference between the size of the recording dot of the binarized output image and the theoretical dot size. Even if the recorded dots are crushed or faded, it is possible to reproduce smooth and high-gradation pseudo halftones that can be optimally corrected for the crushed and blurred print. , High image quality can be achieved. The input image valid signal is used to prevent errors in the area outside the valid range of the input image signal from being added to the image signal. For example, a floor where a noise-like black dot is generated at the upper and left edges of a white original document. It is possible to eliminate the deterioration of tonality. Since the error diffusion direction is fixed by using the two-dimensional pseudo-random number generating means, it is possible to increase the accuracy of the error filter coefficient or increase the processing scale by making the processing direction irregular. In a region where the change in density level is small, it is possible to suppress the texture in which dots are continuous in the main scanning direction, the sub scanning direction, and the oblique direction, and it is possible to improve the gradation reproducibility.

[Brief description of drawings]

FIG. 1 is a configuration diagram of a first embodiment of the present invention.

FIG. 2 is a configuration diagram of a second embodiment of the present invention.

FIG. 3 is a configuration diagram of a third embodiment of the present invention.

FIG. 4 is a detailed configuration diagram of an error correction amount determination unit.

FIG. 5 is a detailed configuration diagram of an error filter 8a.

FIG. 6 is another detailed configuration diagram of the error filter 8a.

FIG. 7 is a detailed configuration diagram of a pseudo random number generator.

FIG. 8 is a detailed configuration diagram of an A-bit pseudo random number generator.

FIG. 9 is a detailed configuration diagram of an error integrated calculation unit.

FIG. 10 is a detailed configuration diagram of an error filter 8b.

FIG. 11 is an explanatory diagram of a state in which a pixel of interest extends beyond peripheral pixels.

FIG. 12 is an explanatory diagram of a state in which a pixel of interest extends beyond a peripheral pixel.

FIG. 13 is a diagram showing an example in which dots are collapsed.

FIG. 14 is a diagram showing a configuration example of a binarizing device used in a conventional facsimile, scanner, or the like.

FIG. 15 is a diagram showing a configuration example of a conventional device for preventing dot collapse.

FIG. 16 is an explanatory diagram of a main scanning direction and a sub scanning direction.

[Explanation of symbols]

 1,16,19 Input terminal 2,15 Adder 3 Threshold generator 4 Comparator 5 Output terminal 6 Subtractor 7 Error data memory 8a, 8b, 22 Error filter 9, 10, 11, 12 Recording dot correction memory 13 Selector 14 Error correction amount determination unit 17 Error data correction unit 18 Output image memory 21 Two-dimensional pseudo-random number generation unit 23 Error integrated calculation unit

Claims (9)

(57) [Claims] 1. An adding unit for adding a binarized error diffusion value to input image data, and a binarizing unit 2 for binarizing an output of the adding unit.
The binarizing means, the subtracting means for calculating the difference between the outputs of the binarizing means and the adding means, and the output of the binarizing means are accumulated, and based on the data of the target pixel and its peripheral pixels. The difference value is corrected and the difference value corresponding to the black and white type of the pixel of interest.
A correction means for outputting correction data and the binarization error created by weighting the difference value correction data corresponding to the peripheral pixels of the next pixel of interest and then integrating the difference value correction data. An image processing apparatus comprising: an error diffusion value calculation means for outputting a diffusion value to the addition means. 2. Addition means for adding a binarized error diffusion value to input image data, and 2 for binarizing the output of this addition means.
The binarizing means, the subtracting means for calculating the difference value between the output of the binarizing means and the adding means, the actual dot size in the recording system when recording the output image of the binarizing means, and the theoretical value. Correct the difference value according to the difference with the dot size, and
Compensation data output corresponding to the black and white type of eye pixel
Positive means and the binarized error diffusion value created by weighting and adding the difference value correction data corresponding to the peripheral pixels of the next target pixel to the adding means. An image processing apparatus comprising an error diffusion value calculation means for outputting. 3. Addition means for adding a binarized error diffusion value to input image data, and 2 for binarizing the output of this addition means.
The binarizing means, the subtracting means for calculating the difference between the outputs of the binarizing means and the adding means, and the output of the binarizing means are accumulated, and based on the data of the target pixel and its peripheral pixels. A predetermined correction pattern is selected, and the difference value is corrected according to the difference between the actual dot size and the theoretical dot size in the recording system when the output image of the binarizing means is recorded ,
Correction means for outputting difference value correction data corresponding to the black-and-white type of the target pixel, and the difference value correction data is weighted to the difference value correction data corresponding to the peripheral pixels of the next target pixel. An image processing apparatus comprising: an error diffusion value calculation unit that outputs the binarized error diffusion value created by integrating after that. 4. The error diffusion value calculation means reads out difference value correction data corresponding to pixels around a pixel of interest from the storage means for storing the difference value correction data,
4. The error filter according to claim 1, further comprising an error filter for outputting the binarized error diffusion value created by weighting and integrating the difference value correction data corresponding to each surrounding pixel to the adding means. The image processing device according to claim 1. 5. An input image valid signal indicating whether the input image is valid or invalid is input to the third correcting means, the difference value correction data is output in the case of a valid signal, and the difference value in the case of an invalid signal. The image processing apparatus according to claim 3, wherein the correction data is not output. 6. An input image valid signal indicating whether the input image is valid or invalid is input to the error filter, the binary error diffusion value is output in the case of a valid signal, and the binary error diffusion value is output in the case of an invalid signal.
The image processing apparatus according to claim 4, wherein the value error diffusion signal is not output. 7. The error diffusion value calculation means weights the difference value correction data and sequentially stores the weighted difference value in a memory, and sequentially reads out the difference value correction data from the memory and weights the difference value correction data corresponding to the peripheral pixels of the target pixel. The image processing apparatus according to claim 1, wherein the image processing apparatus accumulates the binary error diffusion values and outputs the binary error diffusion values to the adding unit. 8. An input image valid signal indicating whether the input image is valid or invalid is input to the error diffusion value calculation means, and in the case of a valid signal, the difference value correction data is output to the error diffusion value calculation means and invalidated. The image processing apparatus according to claim 7, wherein the difference value correction data is not output in the case of a signal. 9. A plurality of sets of weighting coefficients are provided for weighting the difference value correction data, and a random number generator for selecting the plurality of sets is provided with a plurality of one-dimensional random number generators and one of the one-dimensional random number generators. The image processing apparatus according to claim 1, comprising an output selection unit that selects an output.