JP2001061063A - Image processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make the structure of an image processor simpler by setting the matrix size of reference threshold array to be used within the range of a specified value by a dither processing means. SOLUTION: Inputted RGB signals are converted into CMY colors as color reproduction colors of a printer by a color conversion part 21 and supplied to a BG/UCR processing part 22. An ink component is extracted from the CMY colors, CMY colors after extraction are determined and finally converted into CMYK colors and supplied to a gamma correcting part 23 by the BG/UCR processing part 22. The CMYK colors are supplied to a pseudo gradation processing part 24 by performing density correction according to the substantial output characteristics of the printer to the CMYK colors by the gamma correcting part 23. The data of one pixel are converted into multilevel image data with the smaller number of gradation such as approximately 2 to 4 bits for each color depending on the printing capability of the printer by a multilevel dither processing for every color by the pseudo gradation processing part 24. Further the matrix size of the reference threshold array to be used is defined as K×L (K and L are integers) and (K×L)1/2 is set within the range to be expressed by an equation by the dither processing means.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a printer, a copier, a facsimile, and an MFP for dithering multivalued input image data and converting the image data into image data having a smaller number of gradations.
(Multi-Function Peripheral) and the like.

[0002]

2. Description of the Related Art Conventionally, a line LED (light emitting diode)
2. Description of the Related Art In an image forming apparatus such as a printer using a line head, such as a head, a line thermal head, and a line inkjet head, a binary image is formed by printing dots of the same size on recording paper while maintaining the resolution of the head. Was. That is, in the case of a line LED head, a plurality of LEDs, which are a plurality of recording elements arranged in a line,
In the case of a line thermal head, the same size dots are printed on the recording paper as they are in the raster direction, and in the raster direction, a plurality of heating resistors, which are a plurality of recording elements arranged in a line, are printed. Dots of the same size are printed on recording paper, and in the case of a line inkjet head, dots of the same size are recorded with the intervals in the raster direction of the plurality of ink ejection ports, which are a plurality of recording elements arranged in a line. The image was printed on paper to form a binary image. It is also common practice to scan these heads a plurality of times to accommodate resolutions greater than the element spacing.

In an image forming apparatus equipped with such a recording head, a character / line image is simply reproduced as a binary image corresponding to the resolution of the head or a scanning interval, and a graphic / photographic image is organized dithering. The image has been reproduced by a pseudo gradation process such as an error diffusion method. In the pseudo gradation processing in this case, it is very difficult to maintain high resolution and reproduce high gradation at the same time. In particular, in the systematic dither processing, the resolution and the gradation have opposite characteristics. It should be noted that the pseudo gradation process is also used for color characters, light and shade characters, and the like.

On the other hand, in an image forming apparatus provided with such a recording head, the output area within one pixel is modulated by using multi-valued image data generated by multi-valued dither processing of input image data. An image forming apparatus that can express one pixel in several gradations has also appeared. FIG. 31 shows an example of the state of a print head composed of a plurality of print elements used in these apparatuses and the output dots. In the drawing, reference numeral 1 denotes a recording head, and 2 denotes an ink ejection port.
Reference numeral 3 denotes an output dot.

FIG. 31 shows an example of dot output of an image forming apparatus capable of expressing one pixel with three values including white for simplicity. By arranging four or three of these linear printing elements in parallel, a C (cyan), M (magenta), Y (yellow), K (black) color image, or a CMY color image Can be recorded.

In an image forming apparatus capable of printing such multi-valued image data, after performing various types of image processing such as color conversion processing, UCR (under color removal) processing, or gamma correction, a printer engine specific regulation is applied. In order to reproduce the number of gradations, multivalued pseudo gradation processing such as multivalued dither processing using a screen angle for each color or multivalued error diffusion processing is performed to obtain multivalued image data of bits per pixel. I have. The amount of information is concentrated on one pixel to improve image reproducibility.

Generally, systematic dither processing is light in processing,
The degree of freedom of the configuration is high, the speed is high, and the cost can be reduced. However, it is said that the error diffusion processing is superior in image quality. The systematic dithering process discards the quantization error due to the threshold value processing as it is, while the error diffusion process stores the quantization error in peripheral pixels, which is a major algorithmic difference. As a result, in the error diffusion processing optimized from the viewpoint of output characteristics, the output pattern becomes an output pattern with high frequency characteristics that is most inconspicuous from human visual characteristics, and the large edge preservation effect is compared to systematic dither processing. This is an advantage in terms of image quality.

On the other hand, in the case of multi-valued pseudo halftone processing, 2
It has also been found that the image quality does not differ as much as the value. This is because, as an effect of the multi-level quantization, as the level of the multi-level quantization is increased, the quantization error to be discarded is much smaller than in the case of the binarization. In particular, the higher the number of gradations that can be expressed by one pixel at the time of high resolution, the smaller the difference in image quality.

Further, recently, a method has been developed in which a stochastic dither or a fixed mask dither having improved clusters is used to realize an output characteristic comparable to that of the error diffusion processing at the same high-speed processing as the systematic dither processing.

[0010]

By the way, the general 2
The value dithering process basically needs to consider only the threshold arrangement of the single-plane dither matrix, and obtains a binary output image by pixel-to-pixel comparison of the input pixel with the threshold of the dither matrix at the corresponding position. ing. Figure 3 shows this situation.
It is shown in FIG. In FIG. 32, a known 4 × 4 Bayer is used.
It is a schematic diagram at the time of using a type | mold dither matrix. Here, for the sake of simplicity, the threshold of the dither matrix corresponding to the input 4 bits is compared with the input image. For example, if the input pixel value is larger than the threshold of the corresponding dither matrix, 1 (black); This example shows a case where 0 (white) is output to obtain a binary output state having a combination of 1 or 0 as a whole.

As shown in FIG. 32, the dither matrix is repeatedly used on a tile at a cycle of the size of the basic dither matrix (reference threshold array), and the above-described processing is similarly performed on all input pixels. ing. Further, in an output device such as a general printer, a pixel cannot be formed by a digital square lattice, and an output having a shape close to a circle is often generated due to a limitation in a process of each output device. The state of output in this case is shown in FIG.
In general, when a solid image is printed as shown in the dot size of FIG. 33, the shape of the printed pixel completely covers the ideal square pixel so that no gap is generated, that is, の 2 times or more the resolution pitch. Designed to be a circle with a diameter.

On the other hand, in the multi-value dither processing, in addition to the above-described basic dither matrix arrangement, consideration in the depth (pixel level) direction is required. For example, when multi-valued dither processing of D values is performed, (D-1) threshold planes are required, and the dither threshold of each threshold plane is compared with an input image to obtain an output image of D values. FIG. 34 shows a schematic diagram of the multi-value dither processing in this case, and FIG. 35 shows the output state. FIG. 34 is a schematic diagram showing an 8-level multi-value output including 0 (white).

At this time, in the dither processing, it is generally better to give some correlation between the threshold planes in terms of image quality. Therefore, based on this reference threshold array, the threshold of (D-1) dither matrices Is often calculated automatically.

The multi-value dither processing in consideration of the correlation between the respective planes is roughly divided according to a method of allocating a threshold array across the respective planes, as shown in FIGS. 36 (a) and 36 (b).
There are two sequences. In FIG. 36, for the sake of simplicity, multi-value dither processing for converting input 8-bit image data into a 4-pixel (2-bit) image using a 2 × 2 basic threshold value array is shown.

The sequence method shown in FIG. 36A is a method in which the threshold value is filled in each plane unit in ascending order of the threshold value, and is basically not affected by the appearance of dots of adjacent pixels in an ink jet printer or the like. This is a dither process used for a printer that can stably reproduce an image every time. The resolution is almost equal to the resolution performance of the engine, is very high, and the dot density is high. This is an ideal method for reproducing an image by area modulation. However, since the screen is easily filled with pixels of the same size and adjacent sizes, it is easily affected by printing accuracy.

The sequence method shown in FIG. 36 (b) is a method in which an arbitrary one pixel to be processed is sequentially filled in ascending order of the threshold value. This is a dither process that is often used in printers that are easily affected by the appearance state, difficult to form a single pixel, and unstable. This is the case where the resolution is low and the dot density is coarse. If the dither threshold arrangement is of a dot concentration type, an image called a halftone dot is formed. Since the resolution is low, minute printing accuracy unevenness in pixel units is absorbed. In both cases, all thresholds are automatically derived by defining one reference threshold plane and the order of pixel growth in the depth direction.

Considering the case of forming a color image using four standard colors of C (cyan), M (magenta), Y (yellow), and K (black), the multi-value dither processing includes a screen angle. Dot dither and Ba using
Various systems such as a dispersed dither represented by yer or an intermediate cluster dither have been already developed.

However, these dither processes have many problems. For example, when a halftone dot using a screen angle is applied to dither processing, moire such as rosette occurs due to interference between colors. In addition, conventional B
When a dispersed dither matrix such as an aerer type is used, the degree of freedom in dot arrangement is small, and a visually recognizable texture is generated in a specific gradation part. As described above, there are many problems to be solved in order to obtain optimum output characteristics over all colors and all gradations.

These phenomena are not limited to binary values, and similar phenomena occur even when applied to multiple values. In particular, the remarkable occurrence occurs in the dither processing of the sequence shown in FIG. 36B, but it does not completely disappear in the dither processing of the sequence shown in FIG. Furthermore, what can be said about these systematic dither processes in general, including the cluster type, is that the periodicity is easily noticeable visually over the entire input gradation range. In particular, in the case of an output device having a relatively low resolution such as a printer, there is a problem that the periodicity of the output device is very noticeable. As described above, the conventional fixed-period dither still has various problems at present, and the design of the reference threshold array in consideration of each output characteristic having different characteristics for various output devices is also improved. There is room.

Recently, by using stochastic dither or fixed mask dither with improved clusters,
A method for realizing output characteristics comparable to error diffusion processing by the same high-speed processing as systematic dither processing has also been developed. A good example of this is Robert Unichney's "The Void-and-
Cluster Method for Dither Array Generation ”(SPIE
/ IS & T Symposium on Electronic Imaging Science and
Technology, San Jose, CA, February, 1993). However, in these processes, since only theoretical output characteristics in an ideal system are considered, the output characteristics in a dot overlap model of a binary printer are considered at most.

Therefore, the actual accurate output characteristics of each output device are not considered, and the actual characteristics of multi-level output devices are hardly considered. In general, a stochastic dither requires a considerably large matrix size of 128 × 128 or more, and requires a large memory capacity compared to the simple configuration of the dither itself. In particular, a multivalued dither requires a memory capacity that can be said to be redundant.

Accordingly, the present invention provides an image processing apparatus having a simplified apparatus configuration.

[0023]

According to the first aspect of the present invention,
The input gradation image data of M gradations per pixel is converted into output image data of N (M>N> 2) gradations by multi-value dither processing using dither processing means using a reference threshold array, and then 2 In an image processing apparatus for outputting an image by a two-dimensional plane image output means, the dither processing means sets a matrix size of a reference threshold array to be used to K × L (K and L are integers),
(K × L) ^1/2 ,

[0024]

(Equation 3)

The range is set in the range.

According to a second aspect of the present invention, the input gradation image data of M gradations per pixel is subjected to multi-level dither processing by a dither processing means using a reference threshold array, and one pixel N (M>N> 2).
In an image processing apparatus that converts the image data into gradation output image data and then outputs an image using a two-dimensional plane image output unit, the dither processing unit sets a matrix size of a reference threshold array to be used to L × L (L is an integer); L

[0027]

(Equation 4)

The range is set in the range.

According to a third aspect of the present invention, in the image processing apparatus according to the first or second aspect, the dither processing means sets the matrix size of the reference threshold array to be used to the basic period of the systematic dither composed of the minimum number of dots. That is, it is an integral multiple.

[0030]

Embodiments of the present invention will be described with reference to the drawings. In this embodiment, an embodiment in which the present invention is applied to a color inkjet printer will be described. FIG. 1 is a block diagram showing the overall hardware configuration. The host computer 11 transfers color image data of M gradation per pixel to a printer 12 as a two-dimensional plane screen output unit. That is, the host computer 11 transfers code or raster data from the driver 111 to the printer controller 121 of the printer 12 according to the interface characteristics with the printer 12.

The printer 12 is configured to drive and control a printer engine 122 by the printer controller 121. The printer controller 1
Reference numeral 21 denotes coded image data sent from the host computer 11, for example, a page description language such as PDL is developed into a bitmap, and after performing each image processing, the image data is stored in a built-in image memory. It is designed to be stored.

The printer engine 122 converts the bitmap image data from the printer controller 121 into a drive signal, and performs a printing operation by transporting paper, driving a color inkjet head, and the like.

The relationship between the host computer 11 and the printer 12 does not necessarily have to be one-to-one.
It may be used as a network printer in a network that has recently become widespread, and in this case, there is a plural-to-one relationship. Further, the interface between the printer controller 121 and the printer engine 122 basically depends on the architecture of the printer and is not specified.

FIG. 2 shows the printer controller 121.
Is a block diagram showing an example of the configuration of an image processing unit within the image processing unit. The image processing unit includes a color conversion processing unit 21, a BG / UCR processing unit 22, a gamma (γ) correction unit 23, and a pseudo gradation processing unit 24. First, a standard RGB color signal from an 8-bit monitor or the like is first converted by a color conversion processing unit 21 into CMY colors, which are color reproduction colors in a printer, and supplied to a BG / UCR processing unit 22. Note that R, G, and B represent each color of red, green, and blue, and C, M, and Y represent each color of cyan, magenta, and yellow.

The BG / UCR processing section 22 extracts the black component from the CMY colors, determines the subsequent CMY colors, finally converts the CMY colors into CMYK colors, and supplies the CMYK colors to the gamma correction section 23. Note that K indicates black.

The gamma correction unit 23 performs density correction on the CMYK colors according to the substantial output characteristics of the printer, and supplies the result to the pseudo gradation processing unit 24. The pseudo-gradation processing section 24 performs multi-value dither processing for each color to convert data of one pixel into each of colors 2 to 4 corresponding to the printing capability of the printer 12.
The data is converted into multi-valued image data having a smaller number of gradations of about a bit.

FIG. 3 is a block diagram showing a hardware configuration of the printer engine 122.
The control unit 31 uses the multi-valued image data of each color number bit to control the cyan inkjet head 32 based on the image data from the printer controller 121.
Magenta inkjet head 33, yellow inkjet head 34, black inkjet head 35
Are respectively driven and controlled, and each of the heads 32 to 32 is controlled.
A head moving device 36 for controlling the reciprocating movement of the roller 35 in the direction of the rotation axis of the rotary drum, a paper transport motor 37 for transporting the print paper to the rotary drum, a drum motor 38 for driving the rotary drum to rotate, Each of the paper fixing devices 39 provided with a charging roller for charging and fixing is controlled to be driven.

The printer engine 122 is provided with a reciprocating mechanism in which the heads 32 to 35 are mounted side by side along the rotation axis direction of the rotary drum.
The printing paper conveyed by the printing paper 7 is wound around the rotary drum, and the wound printing paper is held by the paper fixing device 3.
9, the rotating drum is rotated by a drum motor 38, and the inkjet heads 32 to 35 are driven based on print data. Further, a reciprocating mechanism is driven by a head moving device 36, When the rotary drum makes one rotation, each of the ink jet heads 32 to 35 moves by の of the interval between the ink discharge ports, and further continues.
35 is driven based on the print data, and the printing on one printing paper is completed when the rotary drum rotates twice, whereby each inkjet head 3 is printed on the printing paper.
Printing can be performed at twice the resolution of the ink ejection interval of 2 to 35.

The pseudo-gradation processing section 24 constitutes a main part of the present invention. The function of this processing section is, for example, an 8-bit, 256-gradation (0: white, 255: black) input gradation image. A case will be described as an example in which data is converted into 3 bits and 8 gradations (0: white, 7: black) for each color by pseudo halftone processing. The input and output are not limited to the above-mentioned number of gradations, and it can be easily understood from the following embodiments that the number of gradations can be changed to any number.

When a 3-bit image for each color can be handled as the capability of the printer, for example, the 3b
It is possible to obtain multi-valued image data of it. this is,
As shown in FIG. 4, a total of eight gradations including white can be reproduced in one pixel using seven types of variable dot sizes for each color per pixel. This is referred to as basic eight gradation characteristics.

In general, it is desirable that the size of each dot of each gradation is adjusted in advance for each color so as to have a linear characteristic in density as much as possible. It is almost impossible to match. For example, in the case of this ink jet printer, it is relatively easy to realize that the ink ejection volume has a linear characteristic for each drop, rather than linearly increasing the brightness and the density.

Although it is not impossible to adjust the target number by adjusting the number of ink drops and the drive waveform of each gradation, the drive waveform is complicated in this case, and redundant processing is liable to be performed. Even if the basic eight gradation characteristics are adjusted to the target characteristics, a total of 256
At the time of gradation reproduction, it is inevitable that a deviation from the target ideal gradation curve occurs. Further, these characteristics are greatly affected even if the characteristics of the paper used are slightly different.

Therefore, the simplest method is to make the configuration as simple as possible so that the gradation characteristics are not greatly distorted, and to correct the printing characteristics of the engine by processing such as gamma correction. However, it is common that at least the dot size of the maximum gradation value forms a circle with a diameter larger than that of a square pixel having a pure resolution of the engine and completely covering the pixel.

The pseudo-gradation processor 24 performs multi-value dither processing, and its basic hardware configuration will be described with reference to FIG. The configuration for implementing the multi-value dither processing is as follows.
Basically, any implementation method may be used.
Is an example.

Reference numeral 51 denotes a main counter, which periodically counts an arbitrary constant in the main scanning direction. The size here corresponds to a cycle up to a 128 pixel count in the main scanning direction. Reference numeral 52 denotes a sub-counter, which counts periodically at an arbitrary constant in the sub-scanning direction. The size here corresponds to a cycle up to a 128 pixel count in the sub-scanning direction.

Reference numeral 53 denotes an encoding unit. The encoding unit 53 converts a counter value input from the main counter 51 and the sub-counter 52 into an encoded M based on a multi-plane dither threshold array corresponding to the position.
An AX 6-bit code is output. Here, MAX6b
It means that the input image data is 8 bits, 256 bits
Assuming that the gradation and the pseudo gradation processing become 3 bits and 8 gradations, the maximum number x of thresholds that can realize the maximum number of reproduction gradations not exceeding 256 gradations by the multi-value dither processing is 25.
6 / {x * (8-1) +1} ≧ 1 Therefore, x ≦ 36
Therefore, it is based on the fact that the reproduction of the pseudo gradation processing up to the necessary and sufficient 256 gradations can be covered by the multi-value dither processing with MAX 6 bits. The basic hardware configuration can be easily realized by a RAM or the like.

Reference numeral 54 denotes an LUT (Look Up Table) unit, which also comprises a RAM or the like, and has a coded MAX of 6 bits.
The conversion result by the actual multi-valued dither processing based on the data of
t, output at 8 gradations.

The pseudo gradation processing unit 24 having such a configuration is
Pseudo gradation expression of 3 bits per pixel, 8 gradations and 256 gradations can be performed on input image data of 8 bits per pixel and 256 gradations by multi-value dither processing. Further, when the encoding unit 53 and the LUT unit 54 are configured by a RAM or the like, before the pseudo gradation processing, the dither reference threshold value array shown in FIG. 6 or the plane shown in FIG. The data calculated based on the sequence of the multi-valued threshold array spanning the interval and tabulated data is stored in each of the selectors 55, 56, 5
Initial loading via 7 enables multi-valued dithering that can be changed to any sequence.

Next, a specific configuration of the algorithm of the multi-value dither processing will be described. 6 and 7 show a configuration example of a sequence algorithm of the multi-value dither processing. In order to simplify the description, the description will be made with a dither threshold array having a very small size.

FIG. 6 shows a reference threshold value array.
It is a screw type dither matrix having a screen angle of degrees. In this case, the pseudo tone reproduction number of one pixel of eight values is
8 × (8−1) + 1 = 57 tones, which is originally small in number of tones, but for the sake of simplicity, the configuration with 57 tones will be described. Note that the basic processing configuration does not change at all even if the number of gradations increases.

In the case of the reference threshold array of FIG. 6, the number of bits of both the main counter 51 and the sub counter 52 in FIG.
The LUT unit 5 converts the 3-bit data encoded by the encoding unit 53 into the input image data.
In step 4, multi-value dither processing is performed and output as 3-bit image data.

FIGS. 7A, 7B and 7C show the sequence of the threshold value array in the depth direction, that is, the pixel level direction when FIG. 6 is used as the reference threshold value array. This threshold is not normalized by 0 to 255 and is represented by a simple serial number of the threshold. In FIG. 7, the horizontal axis item indicates a reference threshold, and the vertical axis item indicates the level number of the multi-value plane.

First, the sequence of the threshold array shown in FIG. 7A has the same threshold array configuration as that shown in FIG. 36A, and is an ideal threshold arrangement. And vertical streak problems. Also, the sequence of the threshold array of FIG. 7B has the same threshold array configuration as that of FIG. 36B, and the problems of density unevenness and vertical stripes due to engine accuracy are less noticeable, but the resolution is reduced. Occurs. The sequence of the threshold array shown in FIG. 7C is an example of a threshold array configuration showing the intermediate characteristics. FIG. 8 shows an example of printing by the above three types of multi-value dither processing when the reference threshold array of FIG. In FIG.
(a) is a print result according to (a) of FIG. 7 and (b) of FIG.
FIG. 7B shows the printing result according to FIG. 7B, and FIG.
This is the print result according to (c).

In the above-described configuration of the multi-valued dither processing, a method for realizing an optimum image reproducibility for a printer by combining the above-mentioned dither reference threshold value array and both sides of the sequence between a plurality of multi-valued dither threshold value planes. It is described below. Note that these two basic configurations have some correlation with each other from the viewpoint of image quality as shown in FIG. Further, another feature of the multi-value dither processing will be described by taking the sequence of FIG. 7A as an example for simplification of the description.

As shown in FIG. 9, in the output of the multi-value dither processing, the pixel to be printed is turned on or off (solid white) in the same manner as the output of the binary dither processing. Only the threshold plane is the low gradation portion, that is, only the highlight portion. All the threshold comparisons in the first threshold plane are turned on. In the higher gradation part, all pixels are filled with dots of some size. Output characteristics. In this case, each pixel corresponds to one halftone dot, and the image is such that each halftone dot grows gradually. Note that the number of reproduction levels of the halftone dots themselves is small.

In such a dot forming process, it is possible to obtain a much higher quality image, particularly in terms of graininess, than the dot reproduction method in which a binary on or off state is used for the same resolution. I know. Further, in a low gradation portion that is processed in the first threshold plane reproduced in the same FM modulation as the binary dither output, the normal 2
Since the dots formed on the paper with respect to the resolution pitch of the printer are much smaller than those of the printer of the value, an image with good graininess can be obtained.

In the case of multi-value dither processing, the number of tones to be reproduced for each threshold plane can be simply divided into (the number of planes minus 1) by taking the above sequence as an example. In the case of 8-value multi-value dither processing,
256 / (8-1) ≒ 36 gradations are sufficient, and this gradation reproduction is simply repeated for seven threshold planes. Therefore,
Since optimization of the pattern design for 36 gradations is sufficient, there is no periodicity in all 256 gradations, such as binary, and further, compared to the threshold design in which texture must not be generated. Optimization can be performed easily.

In the case of an output device capable of forming multi-valued pixels as in this example, the minimum drop, that is, the first basic tone dot to the number-th basic tone dot (this differs depending on the paper and printing accuracy) 4), the dot size of the printer is smaller than the resolution pitch of the printer, as shown in FIG. In such a case, it is possible to design a pattern in which the dots are dispersed as much as possible, which is visually more preferable. At this time, it is preferable to design the dither threshold value array so that the repetition period is not visually recognizable.

In the case of a printer capable of forming such a multi-valued dither image, a dither threshold array in which dots are dispersed as much as possible can be designed. It is rare that the physical accuracy is exactly the same in both scanning directions on the dimensional plane, and usually, the accuracy of one of the two is reduced depending on the architecture of the printer. In the case of an ink-jet printer, accuracy is reduced in the main scanning direction due to variations in the volume and direction of ink ejected from ink ejection ports that are recording elements.

At this time, in the dither processing for reproducing dots that are distributed isotropically in all directions as much as possible, equivalent processing is performed even though there is a bias in printing accuracy, so that printing accuracy is substantially compensated. That would not have been done. Unnecessary noise frequency components due to density unevenness and streaks in actual printing are not canceled out well. However, in the case of a low gradation portion in which adjacent dots are separated from the basic resolution pitch, the density unevenness and streaks are relatively inconspicuous in sight, and the middle to high levels of whether adjacent dots are in contact or not in contact. It becomes most noticeable in the case of the dot size of the tone portion.

Further, in such a multi-value printer, a dot pattern in which minute dots are locally non-periodically and randomly dispersed in a low gradation part in most cases, though depending on basic gradation characteristics. A dot pattern based on organized dither that is periodically and regularly distributed provides a visually pleasing smooth output. However, since human visual characteristics show strong sensitivity in the horizontal and vertical directions, higher image quality can be obtained when adjacent dots are arranged in an oblique direction.

When the largest seventh basic tone dot to be formed on the paper is set to a size that completely covers at least a square pixel of the basic resolution, the other basic tone dot characteristics generally have the characteristics shown in FIG. Become like FIG. 10 shows the results of measuring the respective densities when dots of the same size are printed on one surface for each basic tone on appropriate paper. As can be seen from this figure, the difference between the 0th basic tone density, that is, the difference between the solid white density of the paper and the first basic tone density completely filled with the first drop is the density between other adjacent basic tone levels. That is, it is larger than the difference. Therefore, in the reproduction of a low gradation part which is very important for gradation reproduction, a simple multi-valued dither threshold sequence reduces the gradation resolution in the low gradation part, the density change between the gradations is large, and the gradation jump occurs. It may be noticeable and easy to see.

In consideration of the above points, FIG.
t, a case where input gradation image data of 256 gradations (0: white, 255: black) is converted into 3-bit, 8 gradations (0: white, 7: black) for each color by pseudo halftone processing explain. FIG. 11 shows a reference threshold array, and the matrix size is 30 × 30.

Here, the maximum number x of different threshold values that can realize the maximum number of reproduction gradations not exceeding 256 gradations in the multi-value dither processing is 256 / {x * (8-1) +1} ≧
1 Therefore, x ≦ 36. This is, in other words, 7
The number of gradations assigned to each threshold plane in the threshold array with a plane is 36 gradations. That is, there are output patterns for 36 gradations in one plane. By the way, at this time, 36 * 7 + 1 =
253 gradations can be reproduced.

When these 36 gradations are constituted by a minimum unit matrix, the result is 6 × 6. By turning on arbitrary pixels one by one in each of the 6 × 6 threshold matrices, 36
The gradation of the gradation can be reproduced. Here, as shown in FIG. 11, when the 6 × 6 threshold array A1 is the minimum dither unit, 3
The total threshold array A2 of 0 × 30 has a size in which the minimum threshold array of A1 is exactly five in the main scanning direction and five in the sub-scanning direction, that is, 25 in total. If the matrix size is set to an integral multiple of the minimum dither unit in this manner, the seam of the repetitive processing by organized dither can be smoothly shifted, which is convenient.

Next, a method of arranging the reference thresholds will be described with reference to FIG.
Description will be made using the numerical values described in each threshold matrix of 1. A blank portion of the threshold matrix means that a value of 5 or more is filled. Further, the dither processing is turned on as the gradation increases in the order of smaller numbers.

In the low gradation part, as shown in FIG. 11, the thresholds are sequentially turned on from 1 to 4, but as can be seen from the threshold arrangement, the low gradation part has a local periodicity. A typical (6 × 6 unit matrix) dither threshold array is used. Furthermore, the threshold arrangement of this periodic systematic dither is designed so that it is not arranged horizontally or vertically between adjacent pixels. Thus, in the low gradation part, smooth gradation reproduction that is periodic and inconspicuous is realized. Further, the interval between adjacent pixels having a periodic dither arrangement is larger than two pixels by experiments. That is, every other arrangement in the horizontal or vertical direction is prevented. By doing so, the graininess of the low gradation portion does not decrease.

Next, a process of filling a threshold value into a blank portion is performed. Basically, if one pixel in each minimum dither unit is turned on one by one for each input gradation, a total of 36 levels will be used. The key can be reproduced. From the threshold arrangement of the periodic systematic dither in the low gradation part to the next relatively middle gradation part to the high gradation part (in this example, the threshold value of 5 to 36), the distance between the adjacent minimum dither blocks is small. To realize a threshold configuration that results in an output pattern that is locally aperiodic.

The simplest method for realizing this is a method in which the range of thresholds 5 to 36, which is the remaining unthresholded portion, is determined by random numbers. That is, the next tone value portion is randomly selected for each minimum dither unit, and the threshold is assigned to the selected portion in ascending order. As a result, threshold values of 1 to 36 can be allocated over the entire matrix size.

In general, it has been found that when a uniform gradation surface is processed, a threshold pattern determined by random numbers causes a trouble due to formation of a continuous surface which is visually unpleasant, and becomes noisy. The systematic threshold value array which is distributed most uniformly in the low gradation part is used. Further, in a multi-level printer, adjacent dots do not contact each other in a low basic gradation dot, so that compared to the binary case. Black lumps that are visually unpleasant are difficult to recognize. Therefore,
Even if the threshold value is generated by using a random number, no obvious output pattern that is visually uncomfortable is generated.

On the other hand, a more preferable method of obtaining a threshold value is a method of calculating a portion having the best dispersibility in each basic dither unit by a convolution filter process with reference to the surrounding minimum dither unit. . This process is shown in the flowchart of FIG.

First, in step S1, the numerical value 1 shown in FIG.
Assuming a state where the thresholds indicated by .about.4 are turned on, a pattern having a size of 30.times.30 in which the on portion is 1 and the remaining portion is 0 is set as an initial pattern. Next, in step S2, a convolution filter process is performed on the initial pattern, and the least sparse part of the pattern at the position where the value is 0, that is, the minimum value as a result of the filter operation is calculated. Detect the part to take. At this time, it was found that an excellent output pattern could be obtained by using a filter having the following formula as an example of a suitable filter.

[0073]

(Equation 5)

Here, i is a convolution variable in the main scanning direction, j is a convolution variable in the sub-scanning direction, ki, kj, n
Is an arbitrary constant.

The optimum values of ki and kj are determined by the actually printed dot diameter (minimum dot diameter) and the pitch interval, and the exponent n is determined by the dot shape, especially the dot edge shape. It has become. This is an approximate calculation model that is a shape obtained by patterning ink dots printed on paper.

At this time, it is assumed that there are a plurality of positions having the same value. This is likely to occur because the initial pattern is a periodic systematic dither. At this time, even if the position to be selected is selected at random, a suitable result can be obtained as the first obtained position.

Subsequently, in step S3, the order of the pixel at the detected position is stored, and a pattern in which the bit at the position is changed from 0 to 1 is generated.
In this case, the target is not the minimum dither cycle unit,
A series of ranks at all matrix sizes is determined.

This is repeated until there is no bit pattern of 0, and the priority order of all 30 × 30 pixels is determined. In the assignment of priorities, if a portion (25 × 4 = 100 pieces) that is already reproduced in a periodic pattern is assigned the first order in advance, the priorities 101 to 900 are obtained in the calculation process. Is assigned to a threshold value of 5 to 36.

A threshold is assigned according to the priority. In the case of a matrix having a size of 30 × 30, the allocation is such that the threshold value is increased one by one for every 25 pixels. As a result, the positions of the pixels having the remaining thresholds of 5 to 36 are determined, and finally all the thresholds in the 30 × 30 matrix are filled. This is a dither reference threshold array of 30 × 30 size.

At this time, the weight of the filter operation is relatively changed in the main scanning direction and the sub-scanning direction so that the threshold value is generated so that the dots continue in a direction in which the printing accuracy is relatively lower. That is, weights are given to the values of ki and kj in the above equation (5). More specifically, by setting ki <kj, it is possible to generate a pattern having low printing accuracy and easy to connect in the main scanning direction. It is desirable that the connection strength be set optimally by changing the ratio of ki and kj in accordance with the printing accuracy as shown in FIG.

Thus, in the case of a multi-valued printer, the reference threshold array generated anisotropically does not have a large correction effect on the printing accuracy by itself, but is combined with the sequence between the threshold planes. As a result, it is possible to greatly reduce printing errors such as density unevenness and stripes. When the binary dither processing is performed by a binary printer, dots that are larger than the resolution pitch are connected between adjacent pixels from the beginning, so that they are more resistant to uneven density and streaks. Therefore, an effect is obtained only with the reference threshold value array.

Note that the switching between the arrangement of the threshold values by the periodic systematic dither and the arrangement of the locally aperiodic dither threshold values is about 1 / の of the number of pixels in the basic dither unit.
It turned out that about 10 was good. This is because if too many pixels are fixed at the threshold of the systematic dither, the degree of freedom of the vacant area becomes extremely low, and consequently unnatural textures etc. occur at specific gradations. This is because there is a possibility that it will be lost. Experimentally, good results have been obtained for a range composed of periodic systematic dither up to about 0 to 20% of the reference threshold range.

Next, the threshold matrix size will be described. In this embodiment, the optimal threshold matrix size changes depending on the number of levels of the multi-level quantization. This is because the greater the number of levels of the multi-level quantization, the smaller the number of tones responsible for reproduction in each assigned threshold plane. For example, in the case where about 256 gradations are reproduced by the pseudo gradation processing, in the case of reproducing with one pixel binary, since there is only one threshold plane, the pseudo gradation of 255 gradations composed of 255 different thresholds is used. In the case where reproduction is required and one pixel can be reproduced with four values, 255 / (4-1) = 85, and one threshold plane is responsible for pseudo gradation reproduction for 85 gradations composed of 85 different thresholds. Ok, 1
255 / (8-1) ≒ 3 when reproduction can be performed with eight pixel values
According to 6, one threshold plane may be in charge of pseudo-gray-scale reproduction for 36 gray-scale levels consisting of 36 different thresholds.

As described above, the number of gradations N that can be reproduced by one pixel
Increases, the number of pseudo gradations assigned to each threshold plane decreases, so that threshold optimization becomes easy, and as a result, the optimal threshold matrix size can be reduced.

Here, the actual threshold matrix size K when it is assumed that gradation reproduction of N gradations (N> 2) per pixel can be performed.
(Main scanning direction size) × L (sub-scanning direction size), but if the specified threshold matrix size is too small, the degree of freedom of arrangement for allocating thresholds for the required number of pseudo gradations is small. Is reduced, unnecessary texture is generated in the designated matrix similarly to the conventional systematic dither due to geometrical restrictions, and unnecessary local periodicity is generated. In addition, the repetition characteristics when the threshold matrices are stuck in a tile form become visually noticeable, and the image quality is significantly deteriorated.

From another point of view regarding the degree of freedom, when the threshold matrix size is increased to some extent, unlike the conventional threshold configuration of a dither matrix such as Bayer, a plurality of the same thresholds are used when pseudo-gradation is reproduced. I do. It can be said that the degree of the degree of freedom depends on the number of the same threshold values.

The relationship between the threshold matrix size K × L and the degree of freedom in the arrangement of the threshold values was examined for each situation of the number of tones N that can be reproduced by one pixel. It has been found that an image capable of eliminating at least unnecessary texture and unnecessary periodicity in tone reproduction can be obtained.

That is, there is a degree of freedom for assigning at least 16 or more identical threshold values for one gradation. This is because, as described above, as the number of assigned thresholds for one gradation increases, the threshold matrix size increases accordingly.
In other words, this is equivalent to increasing the degree of freedom.

From the 16 numbers, the necessary threshold matrix sizes K and L when one pixel can be reproduced with N gradations
, The following relational expression is automatically derived.

16 × 255 ≦ K × L × (N−1)

[0091]

(Equation 6)

If the above conditions are satisfied, K and L can use any positive integers. On the other hand, the upper limit of the threshold matrix size is a minimum condition that is set to a size that does not cause redundant processing in an actual hardware configuration.

Naturally, the larger the threshold matrix size, the greater the degree of freedom in the arrangement of the thresholds, so that there is a high possibility that a high quality image can be obtained. However, as a practical matter, in terms of human visual characteristics, if the threshold matrix size is a certain size or more, if the threshold arrangement pattern is ideally optimized, the difference can no longer be distinguished. Is gone. further,
If tone reproduction is performed in multiple levels of N tone per pixel, the above-mentioned AM-modulation processing is performed at a specified input tone or higher, so that it is even harder to distinguish.

In this embodiment, the upper limit of the threshold matrix is derived as follows based on the psychological evaluation theory of visual characteristics. In other words, the minimum threshold matrix size is statistically defined as the patience limit, the maximum threshold matrix size is defined as the detection limit size, and this is applied to the threshold matrix size,

[0095]

(Equation 7)

Is defined. Therefore, from the threshold matrix sizes K and L through the entire range,

[0097]

(Equation 8)

Is derived. Here, when K = L square threshold matrix size,

[0099]

(Equation 9)

Is obtained as it is.

Therefore, according to the present embodiment, the threshold matrix size can be sufficiently reduced even for a large-scale matrix method such as a blue noise mask. For example, using a square threshold matrix, one pixel 4
The threshold matrix size L for tone reproduction with values is
In the range of “45 ≦ L ≦ 135”, the threshold matrix size L in the case of reproducing the gradation with eight values per pixel is “2”.
4 ≦ L ≦ 72 ”.

FIG. 14 shows an output state of an arbitrary gradation when, for example, the threshold matrix is set to 30 × 30 with eight values per pixel. Further, FIG. 15 shows an output state in which the 30 × 30 pattern of FIG. 14 is repeatedly tiled in the main / sub-scanning direction and enlarged twice. As can be seen from FIGS. 14 and 15, it is possible to obtain an output in which the unnecessary texture and unnecessary local periodicity in the pattern and the repetition characteristics when the pattern is laminated in a tile shape are hardly recognized. Can be. Therefore, by using the threshold matrix size under the above conditions, multi-value dither processing can be realized with a sufficiently simple configuration.

Also, regarding the matrix size, if this size is too small, periodicity or unnecessary textures can be seen, so an appropriate, non-redundant size is required. The optimum size changes depending on the basic characteristics of each dot and the compatibility with the paper. If the dot design of each basic gradation pixel does not show extremely large non-linearity, the input gradation image data of M gradations of one pixel is subjected to multi-value dither processing, and one pixel N (M>N> 2) When converting the image data into image data having a smaller number of gradations, the matrix size of the multilevel dither processing is set to K × L, and the number of output gradations after multilevel conversion is set to N gradations. When,

[0104]

(Equation 10)

In the case of a square matrix (L × L), it is set to be an integer in the range of

[0106]

[Equation 11]

If the values are set so as to be integers within the range, the periodicity and the generation of texture can be suppressed. This minimum limit indicates the visual patience limit,
The maximum limit indicates a limit that does not result in redundant size. The matrix size derived from this is smaller than the preferred size of 128 × 128 or 256 × 256 or more, which is generally referred to as binary stochastic dither, and cannot be called a large matrix size anymore. It can be realized by a hardware configuration.

Considering an integer value divisible by the fundamental period of the systematic dither, the above two matrix size conditions are satisfied. For example, as shown on the right side of FIG. It may be a matrix.

Although the example in which the minimum dither unit is a square matrix has been described above, the minimum dither unit may be a rectangular matrix as shown in FIG. In FIG. 16, input gradation image data of 8 bits and 256 gradations (0: white, 255: black) is subjected to pseudo halftone processing to obtain 2 bi-color images.
The case of converting to t, 4 gradations (0: white, 3: black) will be described as an example.

In this case, the maximum number x of different threshold values that can realize the maximum number of reproduction gradations not exceeding 256 gradations in the multi-value dither processing is 256 / {x * (4-1) +1}.
≧ 1 and therefore x ≦ 85. This is, in other words, 3
The number of gradations assigned to each threshold plane in the threshold array with planes is equivalent to 85 gradations. Note that it is not always necessary to use 85 gradations, and in order to make the description easy to understand, this time, each threshold plane assigns an output pattern for 80 gradations. Incidentally, at this time, the gradation of 80 * 3 + 1 = 241 gradations can be reproduced in all the threshold planes.

When the 80 gradations are constituted by a matrix of the minimum unit, the matrix becomes 10 × 8, and by turning on arbitrary pixels of this 10 × 8 threshold matrix one by one, the 80 × gradations are turned on.
The gradation of the gradation can be reproduced. Here, as shown in FIG. 16, when a 10 × 8 threshold array B1 is the minimum dither unit,
The total threshold array B2 of 40 × 40 has a size in which a total of 20 minimum threshold arrays of B1 are four in the main scanning direction and five in the sub-scanning direction.

If the matrix size is set to an integral multiple of the minimum dither unit, the seam of the repetitive processing by the systematic dither can be smoothly shifted, which is convenient. In FIG. 16, the minimum dither unit is constituted by 10 pixels in the main scanning direction and 8 pixels in the sub-scanning direction. However, this allocation method may be changed to the main / sub-scanning direction.

Also, if the size of all dither matrices also meets the following conditions:

[0114]

(Equation 12)

Accordingly, a non-square matrix size such as 40 × 48 may be used.

In the examples shown in FIGS. 11 and 16, the dither matrix having the minimum number of pixels is forcibly provided with a diagonal component.
For example, as shown in FIG. 17, for a screen dither matrix having an appropriate number of tones, an appropriate angle, and an appropriate matrix size, threshold values in all the matrices can be similarly generated using only the pattern of the low tone portion. . Also in this case, the conditions of the two matrix sizes are naturally satisfied.

Further, the total number of tones to be reproduced does not always need to be 256 tones, and even when a screen dither matrix is used, the unique number of tones to be reproduced is determined. It just needs to be able to reproduce the key. As described above, in the low gradation portion, the reference threshold value array is calculated using a pattern conforming to the rule of the periodic systematic dither.

Next, another preferred example will be described. In the low gradation part, a reference threshold array of dither is set so that a pattern having a random output characteristic which is locally aperiodic is set. From the gradation portion to the high gradation portion, the reference threshold array of dither is set so that an anisotropic output pattern in which dots are consecutively arranged preferentially in the scanning direction where printing accuracy is relatively low.

This applies to the case where the dots are reproduced locally and non-periodically to reproduce the dots in the low gradation portion, rather than to reproduce the dots periodically, so that a clear output can be obtained in terms of visual characteristics. . Also in this case, the reference threshold value array is 0 to 20.
The range of% is a range showing characteristics different from the output characteristics in the intermediate gradation part to the high gradation part depending on the printing accuracy. That is, density unevenness and streaks are always easy to see regardless of printing accuracy in the intermediate tone portion to the high tone portion, but in the low tone portion the density unevenness and streaks are relatively difficult to see due to printing accuracy. There is a characteristic.

Here, even in a low gradation area where density unevenness and streaks due to fluctuations in printing accuracy are relatively inconspicuous, depending on the appearance of the output printed by the printer, the threshold value of the low gradation section is as shown below. May be switched.

That is, in the case of a printer characteristic in which print unevenness and streaks are not conspicuous in a low density portion,
Instead of the pattern shown in FIG. 11, an output pattern which is completely isotropic and stochastically generates a uniform number of output dots in each row / column in the main scanning direction / sub-scanning direction of the matrix is obtained. May be used as an initial pattern,
A pattern obtained by processing an arbitrary uniform gradation by an optimized error diffusion algorithm may be used as an initial pattern.

Then, using this initial pattern, a threshold value of a higher gradation portion is randomly created, or a convolution filter is used to form a non-printing pattern in which dots are preferentially connected in the scanning direction where printing accuracy is relatively low. A threshold value that is an isotropic output pattern is obtained. Similarly, it is desirable that the connection strength from the intermediate gradation part to the high gradation part be set optimally according to the printing accuracy. In addition, since the order of the thresholds on the low tone portion side is not determined from the beginning as in the case of the systematic dither pattern, the order is stochastically determined in advance in each low tone portion. A threshold value at which a uniform number of output dots can be obtained for each row / column may be obtained, or the threshold value calculation method for the high gradation side may be applied to the low gradation side for calculation. Since the matrix size is relatively small and the number of gradations on the low gradation side is known at most, it is easy to optimize manually.

Further, in the case of a printer characteristic in which density unevenness and streaks are actually relatively noticeable even in a low gradation area,
The dither reference threshold array is set so that an anisotropic output pattern in which dots are consecutively arranged preferentially in the scanning direction where printing precision is relatively low from the low gradation part. In this case, instead of the pattern shown in FIG. 11, a completely anisotropic output pattern that forcibly generates an output dot continuous in the main scanning direction of the matrix is obtained. Alternatively, an output pattern processed by error diffusion in which the coefficient of the error diffusion matrix is optimized so that the output of an arbitrary uniform gradation is continuous in the main scanning direction may be used as the initial pattern. In addition, as an initial pattern, a threshold array of systematic dither showing characteristics of coupling in the main scanning direction in a low gradation portion may be used. Then, using this initial pattern, a threshold value of a higher gradation portion is obtained in the same manner as the above method.

In this way, even in a case where it is preferable in terms of visual characteristics to locally reproduce dots in a non-periodic manner rather than to reproduce dots periodically in the low gradation portion, the optimum value is obtained in the entire gradation range. A reference dither threshold at which an output is obtained is determined. Further, a reference dither threshold value which becomes an anisotropic pattern can be obtained in the intermediate gradation portion to the high gradation portion in order to make print unevenness and streaks inconspicuous.

As described above, the various types of the reference threshold array generated by this do not have a large correction effect on the printing accuracy by a multi-valued printer by itself, but the various threshold threshold planes described below are used. In combination with the above sequence, it is possible to greatly reduce printing errors such as density unevenness and streaks.

Although the reference threshold array has been described above,
Next, a method of expanding the reference threshold array obtained above in the direction of each multi-level plane will be described. As described above, the dot output characteristics of a printer that reproduces gradation by area modulation are greatly different due to the sequence of the multi-value dither processing. In the above two basic configurations (sequence between the reference threshold array and the threshold plane), print unevenness,
Speaking of compensation for the printing accuracy of stripes and the like, the effect of improving the image quality by changing the sequence between a plurality of multi-valued dither threshold planes is greater.

Unlike the case where the threshold sequence is limited due to the architecture of the printer, the sequence between the threshold planes can be relatively easily changed in the printer as in the present embodiment. However, basically, by changing the sequence between the threshold planes, it is possible to relatively easily suppress printing unevenness and streaks resulting from printing accuracy, but greatly affect resolution and gradation reproducibility. Therefore, it must be carefully designed.

Also, here, there are roughly two schemes for optimizing the threshold sequence. First, the first optimization will be described with reference to FIG.
FIG. 10 shows a general output density characteristic of an output device that reproduces an image by area modulation, as described above. The reproduction resolution of the low gradation part is lower than that of the intermediate gradation part to the high gradation part. This is because, in the low gradation part, dots of the same size as the basic one gradation are arranged in order on the white background portion of the paper in accordance with the reference threshold arrangement every time the gradation rises by one step. In other words, it means that when the dots after the second basic tone are appropriately interwoven and printed, it may be easier to obtain a smooth density change between adjacent tones in tone reproduction. become.

FIG. 18 shows this state. (A) of FIG.
Is a dot growth process in the same sequence as in FIG. 7A when the resolution is the highest. On the other hand, in FIG.
As shown in (b), the output of the same density can be theoretically obtained by interlacing dots of different sizes in addition to the configuration of only the minimum dots. Considering the output characteristics in FIG. 10, the change in density means that a smooth density change may be obtained between adjacent gradations in the growth process of FIG. 18B. . However, in this case, since a dot larger than the basic one tone dot is output to the low tone portion, it is premised that the uncomfortable pattern of that size is not visually recognized.

Here, as a preferable method for obtaining an optimum dot output pattern, instead of an output pattern in which only the basic one-tone dots showing growth as shown in FIG. As shown in FIG. 19B, only a portion up to the gradation indicating the periodic systematic dither output pattern, that is, several basic gradations before arranging the dots of the first basic gradation on the whole. The first is to grow the dots. If the dot has a basic tone that cannot be visually recognized, the tone looks much smoother than the output pattern shown in FIG.

FIG. 20 shows a threshold sequence in accordance with the output pattern shown in FIG. 19B. In this example, the reference threshold value indicating the output of the periodic systematic dither is up to 4, and dots up to the basic three gradations are preferentially output to the corresponding position. This process is effective not only in making the output pattern itself suitable for visual perception, but also in that an output pattern that is strong in printing accuracy and the like can be obtained because the distance between adjacent dots is large. . This effective range is more effective when its value is within a range of about 0 to 10% of the gradation range that can be taken by all the input image data.

In the above description, a case has been described in which dots up to the basic three gradations are preferentially output for a periodic output pattern. However, actually, this sequence is set according to the dot diameter of the basic gradation characteristic. Depends greatly on In other words, the basic dot size is determined by the net resolution of the printer.
In 00 dpi / 600 dpi, the basic minimum dot size and pitch are different. In addition, if the resolution increases, the characteristics of the dots actually measured with respect to the ideal dot diameter become nonlinear with a large shift due to the substantial difficulty of the design.

Accordingly, the sequence setting is arbitrarily optimized as shown in FIGS. 21 (a) and 21 (b) according to the above-mentioned rules by the different basic tone characteristics (particularly the dot diameter) due to these various factors. In addition, the setting of this sequence is similarly performed in accordance with the above-described rules according to the printing accuracy, and
Arbitrarily optimized as shown in (b).

On the other hand, another preferable example of the sequence, for example, in the case where dots are uniformly allocated in the first basic tone, the dot characteristics of the first basic tone are extremely small and have good characteristics. A case where an output that is more visually pleasing than a periodic systematic dither arrangement is obtained is described. A dither threshold array is provided for input images in this range so as to increase the spatial frequency by using the fact that density unevenness and vertical streaks are not conspicuous due to the small size of constituted pixels. FIG. 22 shows an example.

As a result, the number of types of dot patterns appearing with the converted multi-valued image data when the input image data is in the low gradation portion is substantially smaller, and the range of the input image data in the range from the intermediate gradation portion to the high gradation portion is reduced. In the case of the data of (1), the types of dot patterns appearing by the converted multi-valued image data are substantially larger than those of the low gradation portion.

As a result, pixels in a low tone portion, which is a very important element in tone reproduction of a printer, are made inconspicuous, tone reproducibility is improved, and portions where density unevenness and streaks are conspicuous are determined by the type of dot. Can be dispersed to make density unevenness and streaks inconspicuous. Also, unlike the case where thresholds are randomly arranged, since there is a correlation between the threshold planes, the threshold of each plane can be automatically obtained from the basic dither matrix, and simplification of hardware can be expected.

Further, the setting of the sequence based on the different basic tone characteristics (particularly the dot diameter) due to various factors is arbitrarily optimized as shown in FIGS. is there. In addition, the setting of this sequence is performed in accordance with the above-described rule in accordance with the printing accuracy in the same manner as in FIG.
As shown in (a) and (b) of FIG. Note that the example relating to the threshold sequence is more effective when combined with the above-described optimization of the reference threshold array.

FIG. 24 (a) shows the on / off characteristics of the threshold in the intermediate gradation part when focusing only on the normal locally non-periodic and uniformly dispersed reference threshold array. Show. FIG. 24 (b) shows the approximate intermediate gradation by the reference threshold array generated so as to form a pattern showing anisotropic output characteristics in which dots are preferentially connected in the scanning direction with relatively low printing accuracy. 6 shows on / off characteristics of a threshold value in a section.

On the other hand, each pattern in FIG. 25 is a diagram schematically showing an output pattern when actually printed on paper by various multi-value dither processing.
The pixel corresponding to the portion indicated by is shifted rightward due to, for example, the misdirection of the ink head.
Note that (a) to (c) in FIG. 25 respectively correspond to the sequences (a) to (c) in FIG. FIG. 25 (c)
Is based on an isotropic rule as shown in FIG.

On the other hand, FIG. 25 (d) is a diagram showing the state of the output pattern according to this embodiment, and the reference threshold value array corresponds to the anisotropic one shown in FIG. 24 (b). In the case of (d) in FIG. 25, the reference threshold arrays are preferentially connected in the horizontal direction. In other words, the threshold value between the adjacent pixels in the horizontal direction is relatively easy to take a value close to the threshold value, and the dot is preferentially grown in the horizontal direction.

As a result, the correction effect can be expected to some extent in (c) of FIG. 25, but the output pattern as shown in (d) of FIG. 25 can be obtained. With this configuration, even when the printing position accuracy is low, an effect of making density unevenness and streaks less noticeable is exhibited.

Next, a case where gamma correction is incorporated into the multi-valued dither threshold array will be described. In general, the number of theoretical gradations that can be reproduced by the pseudo halftone processing is determined by the total number of different threshold values in the unit matrix. A screen-based dither can reproduce a unique gradation depending on a combination of a pattern size, an angle, and the like, and a stochastic dither and error diffusion can normally reproduce full 256 gradations.

However, this theoretical gradation number is a case where only the pseudo halftone processing section is assumed. Actually, tone loss may occur in all the image processing units before the pseudo halftone processing, so that the image data finally input to the pseudo halftone processing unit has only a limited number of gradations. There are output patterns that are not used at all in the section.

In the normal flow of image processing, color conversion → BG
/ UCR → gamma correction → pseudo halftone, and in each image processing unit, a rounding error due to digital arithmetic processing or a gradation loss due to color gamut compression or the like occurs. Here, since the tone loss in the color conversion unit and the UCR unit cannot be basically restored, the tone reproducibility in the gamma correction unit and the pseudo halftone unit will be described.

The gamma correction process is a process for correcting the basic gradation characteristics of the engine to, for example, target characteristics such as linear luminance and linear density, and has a relationship shown in FIG. That is, by converting the input image data using the target gamma correction curve from the measured basic tone characteristics of the engine to the target characteristics, the process of adjusting the finally output tone characteristics to the target characteristics It is. Although the target curve is a straight line in the figure,
It can be replaced with an arbitrary curve.

In general, the basic characteristics of an area modulation output device that expresses dots by circles are characteristics above the straight line in the graph g in FIG. 26 where the gamma is higher than the target characteristics.
The actual processing of the γ correction is often performed by a digital 1 LUT calculation for each color. In FIG. 26, a digital 1 LUT
Is specifically converted as shown in FIG. In the low gradation part, the data is repeatedly converted into the same data by digital rounding error, and in the high gradation part, the data is converted into discrete values.

In other words, in the reproduction of the low gradation part which is important for gradation reproduction, the halftone patterns to be output tend to be exactly the same even if the input images are different, and the halftone patterns which are not used in the high gradation part are used. Exists,
As a whole, the number of reproduced gradations decreases, resulting in very inefficient image processing. In this phenomenon, the more the basic characteristics of the engine are far from the target characteristics, the lower the digital conversion accuracy is, and the number of reproduced gradations is greatly reduced. However, the characteristics of the ink jet printer are relatively close to ideal.

Therefore, here, the gamma correction is incorporated into the halftone processing, and a matrix for theoretically suppressing the gradation loss is generated. As for the binary dither processing, a method of generating a dither threshold incorporating gamma correction is well known. However, in the case of the multi-value dither processing, since the basic gradation characteristics between the planes are not linear, various threshold planes may be used. However, the conventional method is difficult to apply when all the sequences are adapted.

For example, when one pixel has eight values and the matrix size is 32 × 32, it is simultaneously turned on for an arbitrary gradation.
The number of dots where / OFF is switched is 32 × 32 × (8−
1) / 255 ≒ 28, and in the normal pseudo halftone processing, this number is equally allocated to each gradation.

The basic principle of this processing is that the gamma conversion characteristic is incorporated in the threshold between all threshold planes for determining ON / OFF of pixels in the halftone processing, and the number of pixels to be turned on in each threshold processing of all threshold planes is determined. Control according to gamma characteristics. That is, the gamma correction unit of the digital conversion that causes the gradation loss is passed, and the process is realized only by adjusting the total number of ON dots in the pseudo halftone unit, thereby restoring the actual number of gradations. In this case, when the number of ONs is multi-valued, there are a maximum of 7 ONs per pixel depending on the pixel level direction, that is, up to which of the seven threshold levels the threshold is turned ON. ing.

A method of calculating a multi-valued dither threshold array incorporating gamma correction is shown in the flowchart of FIG. First,
In step S11, a substantial gray scale characteristic of the engine is obtained by multi-value dither processing using a multi-value dither matrix in which the number of dots is uniformly allocated in advance. Next, step S
At 12, the determination of the γ target is made, and step S13 is performed.
Sets the gradation to 0.

Subsequently, in step S14, in accordance with the determined target characteristics, similar to the normal gamma conversion,
For each input tone value, a gamma correction tone value to be converted according to the target characteristic is calculated. At this time, if the gamma correction gradation value is calculated as a real number instead of an integer, the accuracy can be further improved.

Subsequently, in step S15, from the calculated output gamma correction gradation value, a value closest to this value is set to a curve of the output characteristic of a multi-valued dither matrix in which the number of changing dots is equally allocated to each gradation. It is obtained from above and converted to the number of ON dots at this time. The output curve is obtained by arbitrary interpolation using actually measured points. At this time, if the output gamma correction gradation value is calculated by a real number, a resolution up to one dot unit can be obtained.

Subsequently, in step S16, ON pixels are extracted from the threshold priority table, and in step S17, multi-value dithering is performed in ascending order of priority among all threshold planes according to the number of dots to be turned on. Are assigned, and in step S18, the gradation is incremented by one. By repeating the processing of steps S14 to S18 for all gradations, all thresholds can be filled in all threshold planes.

At this time, in the reference threshold array of the multi-valued dither, all priorities have already been calculated. Therefore, as shown in the example of FIG. It is developed in advance and the final priority order among all the threshold planes is obtained.

When a periodic pattern is output in the low gradation portion, the priority order of the reference threshold value array is a priority order according to the rules of the systematic dither. Has already been determined in the reference threshold value array in the calculation process for determining.

In the example of FIG. 29, the reference threshold is set to 4 and the threshold plane is set to 3 for the sake of simplicity. Note that the priority of the reference threshold is actually 1 to 1
It is between six. Here, looking at the sequence between the threshold planes in FIG. 29, first, the part of the reference threshold 1 and the first threshold plane is targeted, and the parts corresponding to these parts are assigned priority. Next, the portion of the reference threshold 1 and the second threshold plane is targeted, and the portion corresponding to this portion is assigned the next priority. By performing this in the entire order of the sequence from 1 to 12, a priority of 1 to 48 is assigned between all the threshold planes.

By setting a threshold value indicating an output corresponding to the number of turning on of a pixel of a corresponding gradation in accordance with the priority order, a sequence between any complicated threshold planes is set. A threshold is uniquely determined in all threshold planes.

As described above, in the present embodiment, by optimally combining the sequence between the reference threshold array and the threshold plane, each dither threshold plane can be set in accordance with the printing precision and the actual dot output characteristics according to the set of dither threshold planes. It is possible to perform multi-value dither processing for optimal output characteristics between tones, and by incorporating gamma correction processing into the multi-value dither threshold itself, it is possible to obtain images with higher tone reproducibility. It became.

In the above-described embodiment, basically, a configuration of a single color, for example, in the case of black, has been described. However, it is easy to extend this to a color image as it is. However, it is necessary to pay attention to the relationship between the output patterns between the colors. In the case of a color image, multi-value dither processing is usually performed for each color. Here, the configuration of the multi-value dither processing having the reference threshold array of the above-described embodiment is applied to at least two or more colors.

Some colors do not need to use the configuration of the reference threshold value array of the above embodiment. For example, Yel
For a color in which dot particles are extremely inconspicuous visually such as low, a simple conventional dither threshold array may be applied without using the configuration of the reference threshold array of the above-described embodiment, or a black like In the case of a color whose edge is to be emphasized more, a multi-valued error diffusion process can be applied to perform a process for further enhancing the edge effect.

On the other hand, in the case of a color to which the processing of the above-described embodiment is applied, for example, if exactly the same threshold pattern is applied to each color, a dot-on-dot output pattern is obtained. If the position is displaced, there is a problem that the reference threshold array is different for each color because there is a problem that the reference threshold array is different for each color. In this case, the reference threshold array may be individually created for each color. However, it is easier to realize the reference threshold array once created by an operation such as inversion, rotation, or shift.

In general, in the case of a binary printer, a severe design is required with respect to color moiré due to the correlation between the patterns of each color. However, in the case of a multi-valued printer, the design of each color pattern itself is required. Since color moiré due to the combination is unlikely to occur, by a relatively simple threshold change operation,
This is because a high-definition image can be expected to be obtained.
In addition, since the threshold array is originally a highly dispersive array, the above threshold operation is sufficient. However, when an image is reproduced with a low-gradation part using a periodic regular systematic dither, the pattern corresponding to the gradation part has a very strong periodicity. Considering this, an appropriate threshold value design is required.

As a preferred example, the assignment of the threshold value in the minimum dither unit is rearranged by an operation such as inversion, rotation, or shift as shown in FIG. 30, or a completely different pattern is newly created. The thresholds in all the matrices may be regenerated based on the newly created reference threshold pattern. At this time, as shown in FIG. 30, in the low gradation part between the colors, it is preferable that the dots are arranged so as not to overlap at the same position and to be juxtaposed as much as possible. This is because a portion where dots overlap each other in spite of being a low gradation portion is substantially a secondary color, and the existence of dots is more visually noticeable. Further, since the gradation portion has a periodicity, the periodicity also becomes conspicuous.

On the other hand, regarding the sequence between the color-related threshold planes, the sequence can be optimally set according to the substantial printing accuracy of each color component.
This is because each multi-value dither process for each color is processed independently, and this configuration can be easily realized.

Further, it is known that, even if the statistical printing accuracy is the same, the influence of density unevenness and vertical streaks on the visual sense generally differs greatly for each color. For example, when the printing accuracy is the same, it is considered that the noise is more noticeable in the order of Y → C → M → K. Therefore, in the multi-valued dither processing for each color, a more optimal output image can be obtained by performing pseudo gradation processing by appropriately changing the sequence between the threshold planes for each color.

In the above, the four colors of CMYK have been described. However, this is not limited to the four colors, and the three colors of CMYK are not limited.
Colors or other combinations of colors can be easily implemented. Further, in the present embodiment, multi-value dither processing has been generally described. However, except for a part limited to multi-value processing such as setting of a sequence between threshold planes, most of the processing can be easily performed in binary dither processing. Applicable.

In this embodiment, the multi-value dither processing has been described. However, the present invention is not limited to this, and those skilled in the art can easily apply the present invention to the density pattern method and the like. Also,
In this embodiment, a case has been described in which a printer is used as the two-dimensional plane image output means. However, the present invention is not limited to this, and a display such as a CRT display or a liquid crystal display may be used as the two-dimensional plane image output means. Good.

[0169]

According to the present invention, pseudo gradation processing can be realized with a simple configuration.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an overall hardware configuration according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a configuration of an image processing unit according to the embodiment;

FIG. 3 is a block diagram showing a configuration of a printer engine according to the embodiment.

FIG. 4 is a diagram showing a pixel size of each gradation in the embodiment.

FIG. 5 is a block diagram showing a configuration of a pseudo gradation processing unit according to the embodiment;

FIG. 6 is a diagram showing a dither reference threshold value array according to the embodiment;

FIG. 7 is a view showing an example of a sequence in a threshold plane according to the embodiment;

FIG. 8 is a diagram showing each output example by the multi-value dither processing of FIG. 7;

FIG. 9 is a diagram showing an example of pixel growth in the sequence of FIG.

FIG. 10 is a graph showing basic tone dot characteristics of multi-value dither processing.

FIG. 11 is a diagram showing a reference threshold array according to the embodiment;

FIG. 12 is a flowchart showing a threshold generation process according to the embodiment;

FIG. 13 is a view showing dot reproduction at each basic dither threshold in the embodiment.

FIG. 14 is a diagram showing an output state of an arbitrary gray scale when the threshold matrix is set to 30 × 30 with eight values per pixel in the embodiment.

FIG. 15 is a diagram showing an output state in which FIG. 14 is enlarged twice in the main / sub-scanning direction in a tile shape.

FIG. 16 is a diagram showing another reference threshold array according to the embodiment;

FIG. 17 is a diagram showing another reference threshold array according to the embodiment;

FIG. 18 is a diagram showing an example of pixel growth in each sequence according to the embodiment.

FIG. 19 is a diagram showing another example of pixel growth of each sequence in the embodiment.

FIG. 20 is a view showing an example of a sequence in the embodiment.

FIG. 21 is a diagram showing a modification of the sequence shown in FIG. 20;

FIG. 22 is an exemplary view showing another example of the sequence in the embodiment.

FIG. 23 is a view showing another example of the sequence in the embodiment.

FIG. 24 is a view showing dot reproduction at a basic dither threshold value in the embodiment.

FIG. 25 is a diagram showing an output pattern in each sequence including print unevenness.

FIG. 26 is a diagram for explaining gamma characteristics and correction thereof.

FIG. 27 is a diagram showing gamma conversion by normal table conversion.

FIG. 28 is a flowchart showing processing for determining a dither threshold between all threshold planes incorporating gamma correction in the embodiment.

FIG. 29 is a view for explaining an operation for obtaining a priority order among all threshold planes in the embodiment.

FIG. 30 is a diagram showing an arrangement relationship of dots of a low gradation portion between respective colors in the embodiment.

FIG. 31 is a diagram showing a line recording head and a printing example thereof.

FIG. 32 is a diagram showing an algorithm of binary dither processing.

FIG. 33 is a view showing an example of print output by the binary dither processing of FIG. 32;

FIG. 34 is a diagram showing an algorithm of multi-value dither processing.

FIG. 35 is a diagram showing an example of print output by the multi-value dither processing of FIG. 34;

FIG. 36 is a diagram showing a sequence of a multi-value dither process.

[Explanation of symbols]

 24 pseudo tone processing section 32-35 ink jet head 51 main counter 52 sub-counter 53 encoding section 54 LUT section

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 2C262 AA02 AA03 AA06 AA16 AB07 AB20 AC17 BB02 BB03 BB06 BB23 BB24 BC11 5B057 AA11 CA01 CA07 CA08 CB01 CB07 CB08 CE13 CE14 CE16 DB02 DB06 DB09 5C077 NN19 NN19 PP08 NN19 NN RR09 RR14 TT02 TT05 TT06

Claims (3)

[Claims] 1. A multi-level dithering process for input gradation image data of one pixel M gradation using a reference threshold array by dither processing means to output image data of one pixel N (M>N> 2) gradation. In the image processing apparatus which outputs the image after the conversion by the two-dimensional plane image output means, the dither processing means sets a matrix size of a reference threshold array to be used to K × L (K and L are integers) and (K × L
L) ^1/2 is given by: An image processing apparatus characterized by being set in the range of: 2. An input gradation image data of M gradation of one pixel is subjected to multi-value dither processing by a dither processing means using a reference threshold value array to output image data of N gradations of one pixel (M>N> 2). In the image processing apparatus for outputting an image by a two-dimensional plane image output unit after the conversion, the dither processing unit sets a matrix size of a reference threshold array to be used to L × L (L is an integer), and ] An image processing apparatus characterized by being set in the range of: 3. The image according to claim 1, wherein the dither processing means sets the matrix size of the reference threshold array to be used to be an integral multiple of the basic period of the organized dither composed of the minimum number of dots. Processing equipment.