JPH11312235A - Image processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To attain a pseudo-halftone processing, which gives parallel error corrections to images of a signal raster by means of a processor that can perform parallel processings by performing quantization and calculation of error data again based on the calculated error data for a pixel whose quantization result changes due to the changes in the estimated value. SOLUTION: An arithmetic means distributes quantization errors caused when input image signals are quantized to the pixel signals of a raster next to a pixel with a quantization countermeasure, and distributes an estimated value of quantization errors to the pixels of the same raster as the quantization countermeasure error and performs quantization and calculation of the error data. Then quantization and calculation of the error data are performed again, based on the calculated error data for a pixel whose quantization result changes due to changes in the estimated value. This arithmetic unit consists of a gradation processing circuit 19 which performs a halftone processing, a binarization processing, etc., according to the set image quality. Then the images are sent to an LD control circuit 20 and printed on recording paper.

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 suitable for, for example, a digital copying machine, a facsimile, a printer, etc. for processing and outputting an input image signal.

[0002]

2. Description of the Related Art As a conventional image processing apparatus, for example, the invention disclosed in Japanese Patent Application Laid-Open No. 8-31721 is known. The present invention relates to a method for enabling pseudo halftone processing with error correction in parallel to a raster image by using a processor capable of parallel processing, so that a quantization error is not distributed to a quantization target raster. It is characterized by the following configuration.

[0003]

In the above-mentioned known system, parallel processing is certainly possible, but by not allocating an error to the same raster, a peculiar texture is generated and the image quality is degraded. There is.

The present invention has been made in view of such problems of the prior art, and has as its object to use a processor capable of parallel processing and to provide a pseudo-raster image with error correction for one raster image in parallel. An object of the present invention is to enable halftone processing and to obtain a high-speed and high-quality image at low cost.

[0005]

In order to achieve the above object, a first means is an input means for inputting an image signal of one raster, and an image signal of one raster inputted by the input means is binary or In an image processing apparatus having an arithmetic unit that quantizes a quantized signal of a multi-level level and an output unit that outputs a quantized signal obtained by the arithmetic unit, the arithmetic unit is configured to quantize an input image signal. The generated quantization error is distributed to the image signal of the pixel of the raster next to the pixel to be quantized, and the prediction value of the quantization error is distributed to the pixels of the same raster as the pixel to be quantized. And a means for determining whether or not the quantization processing result changes due to a change in the predicted value. For a pixel determined to change, an error calculated by the first processing means is calculated. The image processing apparatus characterized by and a second processing means for calculating a re-quantization processing and error data using the data. The second means is
A first means for calculating a prediction value at which a quantization result changes when the prediction value is changed from a possible minimum value to a maximum value; and a calculation result obtained by the calculation means being a preset value. And a control unit for calculating, by the second processing unit, a pixel determined to be not large by the determination unit.

[0006] In the embodiment described later, the arithmetic means is constituted by a gradation processing circuit, and each means in the arithmetic processing means corresponds to each processing function in the gradation processing circuit.

[0007]

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

[0008] 1. FIG. 1 is a configuration diagram showing a document reading device of a digital copying machine to which an embodiment of an image processing device according to the present invention is applied.

In FIG. 1, an original on a contact glass 1 is illuminated by light sources 2a and 2b, and its reflected light is reflected by a light receiving surface of a CCD image sensor 9 via mirrors 3, 4, 5, 6, 7 and a lens 8. An image is formed, thereby reading the original image. The light sources 2a and 2b and the mirror 3
Is mounted on a first traveling body 10 that moves the lower surface of the contact glass 1 in the sub-scanning direction (left-right direction in FIG. 1) in parallel with the contact glass 1, and the mirrors 4 and 5 are mounted on a similar second traveling body 11. ing.

The main scanning is performed by the solid-state scanning of the CCD image sensor 9, and the original image is scanned by the CCD image sensor 9.
And the entire document is scanned by moving the optical system as described above. In the present embodiment, the reading density is set to 16 pixels / mm, and the A3 size (29
It is configured to be able to read even a document of 7 mm × 420).

2. Laser Printer Section The laser printer shown in FIG. 2 generally has a laser writing system, an image reproducing system, and a paper feeding system. The laser writing system includes a laser output unit 21, an imaging lens 22, and a mirror 2
3, a laser output unit 21 is a laser diode (LD) driven by the LD control circuit 20 shown in FIG.
And a polygon mirror that rotates at high speed by a motor and scans the photosensitive drum 24 with laser light.

Around the photosensitive drum 24, a charger 25, an eraser 26, the laser writing system, and a developing unit 27 as a known electrophotographic process mechanism.
, A transfer charger 28, a separation charger 29, a separation claw 30, a cleaning unit 31, and the like, and a toner image is formed on the surface of the photosensitive drum 24 by charging, exposure, development, and transfer. Transferred to recording paper. A main scanning synchronization signal (MSYNC) is provided at a position where one end of the photosensitive drum 24 is irradiated with the laser beam.
Is provided.

The image reproducing process in the laser printer will be briefly described. The peripheral surface of the photosensitive drum 24 is uniformly charged to a high potential by a charging charger 25. When the peripheral surface is irradiated with the laser beam, the potential of the irradiated portion decreases. The laser light is turned on / off according to recording / reproducing black / white, and controls the laser irradiation energy on the surface of the photosensitive drum 24 in accordance with pulse width modulation (PWM) or power modulation (PM). As a result, the photosensitive drum 24
On the surface, potential distribution corresponding to the gradation level of the recorded image,
That is, an electrostatic latent image is formed. When the portion where the electrostatic latent image is formed passes through the developing unit 27, toner adheres according to the process of the potential, and the electrostatic latent image becomes a visualized toner image. The recording paper is conveyed from paper feed cassettes 33a and 33b by paper feed rollers 37a and 37b, respectively, stopped once in contact with the registration roller 38, and conveyed so as to match the toner image on the photosensitive drum 24. The transfer charger 28 transfers the toner image. The recording paper to which the toner image has been transferred is conveyed by a conveyance belt 34, the toner image is fixed by a fixing roller 35, and a discharge tray 36
Is discharged on top.

In this embodiment, there are two paper feed systems. One sheet feeding system, that is, the upper sheet feeding cassette 3
The recording sheet 32a in 3a is fed by a feed roller 37a. On the other hand, the recording sheet 32b in the lower paper feed cassette 33b is fed by a feed roller 37b. Then, the recording paper (recording sheet) fed from any of the paper feed rollers 37 stops once in a state of being in contact with the registration roller 38, and is sent to the photosensitive drum 24 at a timing synchronized with the progress of the recording process. It is. Although not shown, each sheet feeding system is provided with a recording sheet size detection sensor for detecting the size of the recording sheets 32a and 32b stored in the cassettes 33a and 33b, respectively.

3. Image Processing FIG. 3 is a block diagram for explaining the processing order of image data. In the figure, 12 is a sensor driver circuit,
9 is a CCD image sensor, 13 is an amplifier, 14 is A /
D conversion circuit, 15 is a shading correction circuit, 16 is a spatial filter circuit, 17 is a main scanning magnification change circuit, 18 is a gamma (γ) correction circuit, 19 is a gradation processing circuit, and 20 is a laser diode (LD) control circuit. is there.

The CCD sensor 9 is driven by a sensor driver 12 to read a document image of a maximum A3 size (297 mm × 420 mm) at a density of, for example, 16 pixels / mm in both the main scanning direction and the sub-scanning direction. The analog image signal read by the CCD sensor 9 is amplified by the amplifier 13 to a predetermined voltage amplitude, and then the A / D converter circuit 14 generates 2 n gradations per pixel (for example, 256 gradations = 8 bits).
Is converted to digital data. This digital data is corrected by the shading correction circuit 15 for uneven illumination of the light sources 2a and 2b and the sensitivity variation between the elements of the CCD sensor 9.
MTF correction for increasing the resolution of a line image or the like, signal noise removal, smoothing processing for improving the reproducibility of a photograph or the like are performed.

Next, two-dimensional real-time scaling is performed by the main scanning scaling circuit 17, and thereafter, the gamma of the original image is corrected by the gamma correction circuit 18 in accordance with the set density. Next, halftone processing, binarization processing and the like are performed by the gradation processing circuit 19 according to the set image quality, and then sent to the LD control circuit 20 and printed on recording paper.

4. Gradation processing Gradation processing is processing for changing the image quality of an output image in accordance with an image mode or the like and reproducing halftones. The image mode includes a character mode mainly for a character image, a photograph mode mainly for a photographic image, a character / photo mode mainly for a mixed original of characters and photographs, and the like. Further, according to these modes, the data after the γ correction is sent to the LD control as it is 1) When multi-value processing is performed per pixel 2) The multi-value level is reduced to 3 values, 5 values, 9 values, etc. In the case of performing multi-value error confirmation processing 3) In the case of performing area gradation processing using a plurality of pixels as one unit even in multi-value processing 4) In the case of performing binarization with a fixed threshold value in the processing of binarizing one pixel 5) Binarization by error diffusion 6) Binarization by dither threshold

The present invention proposes a processing method effective for multi-value error diffusion processing or binary error diffusion processing.

4.1 Error Diffusion Method The error diffusion method, which is generally known, will be briefly described. The error diffusion method includes binary and multi-value error diffusion methods.

The error diffusion method is strictly distinguished from the average error minimization method, such as a difference in processing at the end of an image. However, basically, the input image signal (multi-valued) and the output image signal (in this case, 2
Is reallocated to the input image signal of the peripheral pixels to reduce the requantization error as a whole, and to save the grayscale information to reproduce the gradation. In many cases, these two methods are combined and called an error diffusion method.

4.1.1 Binary Error Diffusion Method In the binary error diffusion method, an error generated at the time of binarization is weighted and distributed to neighboring pixels to be binarized in the future. In other words, when a pixel is binarized, the input image signal of the pixel is binarized after being corrected by a weighted error from a neighboring pixel that has already been binarized. The error generated at this time is newly diffused to another peripheral pixel. A binary matrix in which peripheral pixel matrices are weighted with weighting coefficients is called a weight matrix.

4.1.2 Multilevel Error Diffusion Method In the same manner as the binary error diffusion method in which one pixel is binarized and the error at that time is diffused and distributed to the periphery, the multilevel error diffusion method uses Is to requantize one pixel of the signal to several to several tens of levels, and assign an error generated at that time to peripheral pixels by weighting. The circuit configuration is almost the same as that of the binary pixel diffusion. However, in the case of binarization, the error generally has a large range of values. Is in a small range. The number of operation bits in the error memory or the weight matrix may be reduced by that amount, and the multi-value does not necessarily mean that the circuit configuration is complicated or the scale is increased. The multi-level error diffusion method is not only excellent in tone reproduction and resolution, but also has an effect of suppressing the occurrence of a distinctive stripe pattern that is conspicuous in binary error diffusion.

Further, as another error diffusion method, for example, a method disclosed in Japanese Patent Application Laid-Open No. 9-65129 is also known.

4.1.3 Specific Processing of Error Diffusion Method In the error diffusion method, as shown in FIG. 4, a quantization error at the time of re-quantization (same as multi-level quantization) other than binary or binary is used. This is a method of allocating to surrounding pixels that have not been processed yet. In other words, from the point of view of the pixel of interest to be processed, it is binarized or multi-valued after adding the errors distributed from surrounding processed pixels, which were mentioned earlier.

The following equation (1): D11 '= D11 + (1 / 4.times.e00 + 1 / 4.times.e01 + 1 / 4.times.e02 + 1 / 4.times.e10) (1) represents the target pixel P as shown in FIG. Is an expression for correcting an error from a pixel around the pixel P by adding it to the data of the pixel of interest P.

Next, when binarizing the corrected data, the following equation (2) is used: D11'≥THB11 = 1; e11 = D11'-TH D11 '<THB11 = 0; e11 = D11' where TH In the case of multi-level conversion (here, quinary conversion), the following equation (3) is obtained by the equation (3): D11 '≧ TH4 B11 = 4; e11 = D11′−TH4 TH4> D11 ′ ≧ TH3 B11 = E11 = D11'-TH3 TH3>D11'≥TH2 B11 = 2; e11 = D11'-TH2 TH2>D11'≥TH1 B11 = 1; e11 = D11'-TH1 D11 '<TH1 B11 = 0; e11 = D11 'where TH1, TH2, TH3, and TH4 are constants (TH1 <TH2 <TH3 <TH4)... (3), and the binarization or requantization (Bij) and the error (eij) are calculated. .

In FIG. 4, e00, e01, and e02 are errors derived from the processed data of the previous line of the target pixel P, and the values of the errors are stored in a line memory or the like. The error e10 is one pixel before the main scanning, and the error is calculated and distributed when the processing is completed immediately before.

The weight matrix of the error shown in FIG. 4 distributes the error between the previous line and the immediately preceding pixel. The weight matrix to be distributed may be a larger matrix than the weight matrix in FIG. In such a case, it is necessary to accumulate the error between the previous line and the line before the previous line, and also in the main scanning direction, it is necessary to distribute the error from the immediately preceding pixel and the immediately preceding pixel.

4.1.4 Processing by Parallel Processing Processor Here, focusing on the distribution of errors in the main scanning direction, in such an error diffusion method, the processing result of the immediately preceding pixel is calculated by the following arithmetic processing. It can be understood that the processing method of the IIR type filter used for the above is used. In such a process, the efficiency is deteriorated in a parallel processing type circuit or processor configuration that simultaneously processes a plurality of pixels in the main scanning direction. In the parallel processing type processor 501, as shown in FIG. 5, a plurality of pixel data adjacent in the main scanning direction are simultaneously input to the processor 501, and the same processing is simultaneously performed on all the input pixels. ). As such a processor, it is also possible to calculate a target pixel including data of surrounding pixels before processing,
It is also possible to use the error data of the previous line, the line before the previous line, and the like stored in the error memory 502 for the calculation. In a system in which the data is processed while sequentially passing the processed data, there is no merit of the parallel operation. As a parallel processor, a processor as disclosed in Japanese Patent Application Laid-Open No. 8-31721 is also known. A processor that processes a plurality of data simultaneously with the same processing (instruction) is called an SIMD processor.

Such a parallel processing type processor 501
By performing the error diffusion processing, the advantage of the parallel processing can be enjoyed. The details will be described below.

For example, a weight matrix of error diffusion as shown in FIG. 6B is used as an example in an array of data as shown in FIG. 6A. Here, when D14 in FIG. 6A is set as a target pixel, the correction formula is as follows: D14 '= D14 + 1/16 × (2 × e12 + 4 × e13 + e02 + 2 × e03 + 4 × e04 + 2 × e05 + e06) (4) Assuming that the normal error diffusion process is performed, D14 '= D14 + 1/16 * (2 * e12' + 4 * e13 '+ e02 + 2 * e03 + 4 * e04 + 2 * e05 + e06) (5) However, e12 'and e13' are transformed from the equation (4) so that the prediction errors are as follows. In other words, for errors e12 and e13, which are errors from the same main-scan data, prediction errors e12 ',
e13 'is defined to use these prediction errors.

The error data eij is uniquely determined as shown in equations (2) and (3) by determining the correction data Dij '. That is, the error data eij is the correction data D
The function fe of ij ′ can be defined as eij = fe (Dij ′) (6). Therefore, the prediction of the error data starts from the prediction of the correction data.

<Prediction Error Calculation Example 1> To calculate the prediction error, for example, the correction data D13 'and D12' are respectively applied to the equation (4), and the adjacent error in the main scanning direction at that time is calculated. , Cn (n: 0, 1, 2, 3) as constants, D13 '= D13 + 1/16 × (2 × C0 + 4 × C1 + e01 + 2 × e02 + 4 × e03 + 2 × e04 + e05) (7) D12 ′ = D12 + 1/16 × (2 × C2 + 4 × C3 + e00 + 2 × e01 + 4 × e02 + 2 × e04 + e04) (8) From the calculation result, e13 ′ = fe (D13 ′) (9) e12 ′ = fe (D12 ') ... A predicted value of the error data is calculated as shown in (10).

As a result, the predicted value of the error data can be calculated even if there is no processing result of the previous pixel. Therefore, the data before processing and the accumulated error data of the previous line are simultaneously used between adjacent pixels. Processing can be performed.

<Example 2 of Predicted Error Calculation> It is also possible to perform processing by a different processing method as described below. That is, it is conceivable that the constant Cn is set to "0" when the total error of the previous line is smaller than a preset threshold value (T0, T1), and a constant other than that. This is to predict that a state where there is no error in the previous line is a place where there is no data change such as a white background of an image.

In this processing, in the equation (7), C0 and C1 are expressed as follows: when e01 + e02 + e03 + e04 + e05 <T0, C0 = 0, C1 = 0 e01 + e02 + e03 + e04 + e05 ≧ T0 When C0 = K0, C1 = K1, T0, K0, K1 Constant ... Determined under the condition of (11).

Similarly, in the equation (8), C2 and C3 are expressed as follows. When e00 + e01 + e02 + e03 + e04 <T1, C2 = 0, C3 = 0 e00 + e01 + e02 + e03 + e04 ≧ T1 C2 = K2, C3 = K3 where T1, K2, and K3 are constants. ··· Determined under the condition of (12).

<Prediction Error Calculation Example 3> In addition, for example,
When the sum of the previous line errors is equal to or larger than the threshold value, the constant may be changed by comparing the data before correction with another threshold value (T2, T3). This process is a process for reducing the possibility that the prediction error deviates from the actual one.

In this processing, in the equation (7), C0 and C1 are set as follows: if e01 + e02 + e03 + e04 + e05 <T0, C0 = 0, C1 = 0 e01 + e02 + e03 + e04 + e05 ≧ T0 If D13 ≧ T2, then C0 = K00 and C1 = K10 D13 < If T2, C0 = K01, C1 = K11. However, T0, T2, K00, K01, K01, and K11 are determined under the condition of a constant (13).

Similarly, in the equation (8), C2 and C3 are expressed as follows. When e00 + e01 + e02 + e03 + e04 <T1, C2 = 0, C3 = 0 e00 + e01 + e02 + e03 + e04 ≧ T1 If D12 ≧ T3, then C2 = K20 and C3 = K30 D12 <T3. If there is, C2 = K21, C3 = K31, where T1, T3, K20, K21, K30, and K31 are determined under the condition of a constant (14).

<Prediction Error Calculation Example 4> In these predictions, when the sum of the errors of the previous line is small, it is possible to predict the error data by adding a certain value instead of setting the constant to 0. This is to prevent the image from becoming unnatural at the boundary due to the smooth rise of the error distribution when entering the image portion from a white background.

In this processing, equations (11), (12),
In (13) and (14), e00 + e01 + e02 + e03
When + e04 <T0 or T1, C0, C1, C2, C3
A constant other than "0". However, these constants are K00, K01, K10, K11, K20, K21, K30, K31
Is a constant different from.

The error is predicted by each of these methods, and the following description will be made using equations (7) and (8).
This error diffusion process is executed according to a processing procedure as shown in the flowchart of FIG.

In this processing, equations (5), (7),
According to (8), (9), and (10), the parallel processor 501 simultaneously corrects the correction data D1 for all pixels.
At the same time as calculating the error-corrected data corresponding to 4 '(steps 701, 702, 703), the error eij uniquely obtained and the binarized or multi-valued quantized value Bij are calculated. (Step 704). In this calculation, equation (2) is used for binarization, and equation (3) is used for quinary. At this time, it is determined whether or not the quantized value of Bij changes when the estimated values of the errors e12 'and e13' are changed from the minimum value to the maximum value in equation (5). Specifically, Bij when the minimum value is entered in both e12 'and e13' is compared with Bij when the maximum value is entered in both, and it is determined whether they are the same (steps 705 and 706). .

Alternatively, instead of determining whether or not the quantized value of Bij changes, it is possible to determine the possibility of change. In this case, the same value is put in e12 'and e13', and a point at which Bij changes when gradually increasing from the minimum value is searched. Whether the values of e12 'and e13' at that time are larger than a certain threshold value is determined. The possibility of the change is determined. In addition,
When it is larger than the threshold value, the possibility that Bij changes is small. In the determination as to whether or not to change, and in the determination of the possibility, a determination result is issued at the same time as the calculation of D14 ', and Bij and eij of each pixel after error diffusion processing are calculated.

If it is determined in step 706 that the quantized value of Bij does not change or is not likely to change, error diffusion processing is executed using eij and Bij calculated in step 704. On the other hand, step 706
In the determination of Bij, if Bij changes or is likely to change, in other words, for the pixel that is determined to change Bij or is likely to change, the line including the target pixel calculated in step 701 Using the error data e ij, D 14 ′ is recalculated based on the equation (4), and the newly quantized value of B ij and the error data e ij are corrected and calculated, and the calculation result in step 701 is corrected. The error diffusion processing is completed.

[0048]

As described above, according to the first aspect of the present invention, it is possible to use a processor capable of parallel processing and perform pseudo halftone processing with error correction on an image of one raster in parallel. Thus, a high-speed, low-cost, high-quality image can be obtained. Further, even when the parallel processor is not used, the processing can be simplified. Further, the influence of the prediction error of the prediction error can be minimized.

According to the second aspect of the present invention, in addition to the effect of the first aspect of the present invention, the possibility of misprediction of the error is determined, and only the one with the highest possibility is subjected to the restoration processing, so that reprocessing is possible. , While providing high quality image quality.

[Brief description of the drawings]

FIG. 1 is a schematic configuration diagram illustrating a configuration of an image reading device according to an embodiment of the present invention.

FIG. 2 is a schematic configuration diagram illustrating a configuration of a laser printer according to an embodiment of the present invention.

FIG. 3 is a block diagram illustrating a processing order of image data from image reading to image output in the embodiment.

FIG. 4 is an explanatory diagram illustrating a processing method of an error diffusion process.

FIG. 5 is a block diagram illustrating a configuration of a parallel processor.

FIG. 6 is a diagram illustrating a processing data array and a weight matrix when an error diffusion process is performed using a parallel processor.

FIG. 7 is a flowchart illustrating a procedure of an error diffusion process according to the embodiment.

[Explanation of symbols]

 Reference Signs List 9 CCD image sensor 12 Sensor driver circuit 13 Amplifier 14 A / D conversion circuit 15 Shading correction circuit 16 Spatial filter circuit 17 Main scanning magnification change circuit 18 Gamma (γ) correction circuit 19 Gradation processing circuit 20 Laser diode (LD) control circuit

Claims (2)

[Claims] 1. An input means for inputting an image signal of one raster, an operation means for quantizing an image signal of one raster input by the input means into a binary or multi-level quantized signal, and this operation Output means for outputting a quantized signal obtained by the means, wherein the calculating means calculates a quantization error generated when the input image signal is quantized by a pixel of a raster next to the pixel to be quantized. First processing means for allocating a prediction value of a quantization error to pixels on the same raster as the pixel to be quantized, and for quantizing and calculating error data; Means for determining whether or not the processing result changes; and for the pixel determined to change, the quantization processing and error data are calculated again using the error data calculated by the first processing means. The image processing apparatus characterized in that it comprises a second processing means for, the. 2. A calculation means for calculating a prediction value at which a quantization result changes when the prediction value is changed from a possible minimum value to a maximum value, and a calculation result by the calculation means is calculated from a preset value. A determination unit for determining whether the pixel value is large, and a control unit for calculating, by the second processing unit, a pixel determined to be not large by the determination unit. The image processing apparatus according to any one of the preceding claims.