JPH08195882A - Improvement method and device and for correcting tone by predistortion of picture data used in ink reduction device in binary printer - Google Patents

Abstract

PURPOSE: To provide improvement for reducing the required amount of calculation, reducing the required amount of ink and performing printing with a high resolution in a turn-to-halftone process by an error diffusion method or a dither method for converting gray scale pictures provided with continuous tones to halftone pictures suitable for a binary printer. CONSTITUTION: At the time of defining the resolution of printing as dpi, when the radius of a dot corresponding to one picture element formed by the printer is 1/dpi<= (r)<=√2/dpi, a de-vice for performing the turn-to-halftone process by the error diffusion method or the dither method selectively executes a halftone processing only to every other picture element in respective rows while processing picture data by each row. For instance, only the picture elements of odd-numbered columns are processed in odd-numbered rows and only the picture elements of even-numbered columns are processed in even-numbered rows. The printer prints only the processed every other picture element.

Description

Detailed Description of the Invention

[0001]

FIELD OF THE INVENTION This invention relates to digital printing devices and, more particularly, to techniques for reducing the amount of ink required for printing and processing time.

[0002]

BACKGROUND OF THE INVENTION Many computer-driven printing devices that produce hardcopy, such as laser printers, dot matrix printers, and ink jet printers, print in a binary format. , The output medium is divided into an array of picture elements or "pixels" and the device either prints a small colored dot at each pixel location or blanks out the relevant pixel location. In black and white printers all dots are printed in a single color, whereas in color printers the dot corresponding colors are selected from a small set of colors. In either case, the dots themselves have a uniform color and the resulting output consists of an array of colored and blank pixels.

On the other hand, an image generated by an imaging method based on photographic technology or computer technology is
The tone has continuity. When such an image is divided into pixels, each pixel will present a "grayscale" color whose tone value falls within some predetermined tone range. Therefore, in order to reproduce such a "continuous tone" image by electronic printing, the image format is converted to one that is generally binary and compatible with the characteristics of the printing device. I have to do it. This conversion process, which can take many forms, is commonly referred to as "halftoning,""halftoning,""halftoning," and the like. A halftoned image (hereinafter referred to as a halftone image) actually consists of only a spatial pattern of binary pixels (colored dots or blank dots), but it depends on the visual system of the human body. This pattern is integrated to produce the illusion of a continuous tone image.

During the printing process, the image to be printed is divided into a series of pixels, and
The image value at each pixel is quantized to produce a multi-bit digital word representing the tone value of that pixel. The image is thus converted into a stream of digital words and applied to the printing device. In order to convert the format of each word into a format suitable for reproduction on a digital device, a process of halftoning the digital word stream is performed during a process called preprocessing. Over the years, many halftoning techniques have been developed and improved. In their simplest form, the value of each digital word is compared to a threshold level and a binary output pixel value is generated depending on its relative value.

For example, a digital scanner that processes continuous tone images can produce a multi-bit word stream representing the detected light intensity.
The numerical values of these words typically range from 0 to 255, corresponding to 256-level grayscale or 8-bit words. When such a stream of digital words is reproduced on a binary printing device, the process of halftoning compares the output word of the scanner to either a single threshold or an array of thresholds to obtain the required 2 It is adapted to generate a stream of pixels for value output. In such a system, each 8-bit scanner word is effectively compressed into a 1-bit output word.

Of course, such compression results in a considerable loss of visual information, resulting in distortions in the reconstructed image that were not present in the original image. . Therefore, additional techniques have been developed to reduce the visual distortion caused by the halftoning process. One approach, known as "error diffusion", is "quantization error" (ie, an input value represented by a multi-bit word and an output represented by a single bit or two multi-bit words). The difference between the values) is proportionally "diffused" among adjacent pixels.

This diffusion processing is performed by adding a part of the quantization error to the input value of the next pixel in the processing line and the adjacent pixel in the next one line or a plurality of lines. The quantization error is added to the pixel values before they are processed so that they are "spread" over several pixels.

According to one embodiment of the error diffusion process, the input pixels represented by the input word are "raster".
They are processed in sequence (that is, each line is processed one row from left to right before the next lower line is processed).
In general, when the processing sequence goes from left to right, the error diffusion process produces a good image, but it is known as "worm" and "snow plowing" that degrades the image quality. A by-product noise pattern of "Worms" appear as curves or diagonal lines in a homogeneous gray image area,
Then, it produces a patterned appearance in the halftone image. As such, various techniques have been used to reduce or eliminate these types of noise patterns.

US Pat. No. 4,955,065, JP-A-4
-37256 and some of the known techniques of this kind disclosed in JP-A-4-51773,
From a conventional raster scan pattern (where each pixel within each line of the image is processed from left to right, and the lines are processed from top to bottom):
The pixel lines are still processed from top to bottom, but the arranged pixel lines are turned into a serpentine scan pattern which is processed in alternating directions.

These techniques correct the worm,
While this causes the worm to appear as horizontal lines in the homogeneous gray area of the image, and thus be less noticeable, the worm is still present as it would be with a conventional raster scan.

The other known methods for minimizing the appearance of the worm disclosed in Japanese Patent Laid-Open No. 3-151762 are also different from the conventional raster pattern in the processing sequence. According to one embodiment, the processing direction for each line of the image is selected based on the quantized "1" and "0" streams output from the random number generator or the noise pattern generator. Therefore, pixel processing for each line is performed from left to right (in response to "1") or from right to left (in response to "0"). According to another known embodiment, the pixels in the line are
Processing from left to right and from right to left according to a predetermined periodic pattern. For example, scanning twice from left to right and then twice from right to left. Although these conventional processing sequence techniques reduce the size of the worm, the worm is still present in the halftone image.

Another problem inherent in the error diffusion method is that a large number of calculations are required to process one image. Millions of calculations may be required to produce the final output for a typical high resolution image, as some calculations, including multiplications, must be done for each processed pixel. In addition, it is necessary to store at least one line of error values to be applied to one continuous line of image values. The combination of this extremely large number of calculations and the amount of storage required increases the time to process each image, and consequently reduces the processing power of the printer.

Therefore, in order to reduce the number of calculations required to process one image, an alternative technique to the error diffusion method was developed. One such alternative technique is called "regular dithering", where an array of pre-determined (and generally not the same) thresholds, or "dither arrays," is used as image pixels. It is conceptually arranged to cover the pixel array of the image with the same spacing.

If the dither array is smaller than the image array, the dither array is repeatedly tiled or "tiled" to form a repeating pattern.
Therefore, each pixel conceptually has two values associated with it. That is, the actual tone value of the pixel and the threshold of the cells of the dither array, which are compared. Alternatively, the value of the dither array may be added to the tone value of each pixel, and this added value may be compared with a fixed threshold value.
In either case, an output value is generated based on the comparison. Since the processing of each pixel is a simple comparison and is independent of the neighboring pixel values, the calculation time required to process one whole image is the calculation time required for error diffusion. Much less than.

However, the quality of the resulting processed image depends entirely on the threshold selected for the dither array. Generally, two types of dither arrays are used. In one type, the "lumped dot" type dither array, the thresholds are arranged in a small cluster of closely spaced values. These clusters mimic "dots" of various sizes, thereby producing a process that is visually similar to the image produced by traditional halftone screens that have been used for many years in the photolithography process. The generated image is generated.

On the other hand, in the "dispersed dot" type dither array, large and small threshold values are evenly distributed over the entire array. This technique improves printable grayscale as well as resolution. In an image reproduction device capable of clearly displaying a single separated pixel, a distributed dot dither array generally produces better quality images than a lumped dot dither array. The reason is that distributed dot arrays have better fidelity to high frequencies and can produce a better continuous gray area illusion than concentrated dot arrays with the same resolution and period. is there.

With both types of arrays, the number of discrete gray levels that can be represented (increases as the number of cells in the dither array increases) and the annoying low frequency geometry in the region of uniform gray. There is generally some kind of antinomy relation between the appearance of patterns, which also appears as the number of dither array cells increasing. Further, when the size of the dither array is increased, there is also a tendency that the resolution of the image decreases as the low frequency value dissipates. One way to reduce these patterns and maintain resolution is to make the size of the dither array very large and a homogeneous threshold pattern where consecutive thresholds result in a pattern that "turns on" pixels. Using a dither array that is homogeneous and random.
However, as the size of the array increases, assigning a threshold to the cells of the dither array to produce such a homogeneous pattern becomes difficult. Therefore, various prior art techniques have been developed to obtain random threshold sequences and assign them to the cells of the array.

One conventional way to produce a homogeneous regular dither array is called the "repetitive mosaic method". The goal of the iterative mosaic method is that each successive cell is numbered, ie, "on", the cells in the "on" state of the two-dimensional array are arranged as homogeneously as possible. Is to assign a threshold value to each of the cells of the dither array. Because of this, when the dither array is used as a threshold array in the halftoning process, the corresponding array of output binary dots will be distributed as homogeneously as possible for each gray level to be simulated.

The algorithm used in the iterative mosaic method for generating the threshold array is such that when a regular geometric shape or "tile" covers or is laid out over a two-dimensional dither array. Based on the fact that the center point of a shape and its vertices can act as a center point for laying a new tile in the array with the same geometric shape but sized one cell smaller. . Then, the center points and vertices of these new tiles can act as the center points for another laying by the tiles that are again reduced by one cell.

According to the iterative mosaic algorithm, the vertices of all tiles at each stage of the iterative mosaic are assigned a threshold number prior to the occurrence of the next laying sequence. The mechanism provided by this algorithm is for locating the point cloud that lies exactly at the center of the void that occurs between the tile vertices. Essentially, a new point is added at the largest "blank" or area where there are no "on" pixels. Robert Ulichney, 1990, Cambridge, on the iterative mosaic method and the resulting threshold array.
It is detailed on pages 128-171 of "Digital Halftoning" published by MIT Press, Massachusetts and London, England.
Although iterative mosaic methods can produce large dither arrays, these arrays, when used in conventional halftoning, have a strong periodic appearance of an unnatural appearance to the resulting image. That
There is a structural defect.

Therefore, other prior art methods were developed to produce the dither array, one of these prior art methods is called the "blue noise mask". Generally, "blue noise" is the high frequency component of the white noise spectrum.
The power spectrum of blue noise is cut off at low frequencies and remains flat up to the limits of considerable high frequencies. In general, dots generated by blue noise masks
The patterns are non-periodic and radially symmetric, and these patterns produce non-periodic, non-correlated dither array configurations without graininess at low frequencies. The choice of blue noise mask came from a frequency study on the dot pattern produced by the error diffusion process described above. When these dot patterns are transformed into a power versus frequency power spectrum, what is observed here is that the low frequency cutoff point of the power spectrum is, as the name "blue noise" implies: It occurs in the frequency range of blue light.

Regarding the shape of the power spectrum, the dot noise that produces the power spectrum of blue noise
It can be used to construct a dither array by examining the pattern and generating a dither array from these dot patterns. However, for the high quality output produced by the resulting array, the transformation between the frequency and the spatial domain required to derive the array results in the complex algorithm required to produce the dither array. Is generated. For Blue Noise Mask, Proceedings SPIE, Image Processing
Algorithms And Techniques II, v. 1452, pps. 47-5
6, February 21st-March 1, 1991, `` Digital Halftoning Using A Bl '' by T. Mitsa and K. Parker.
ue NoiseMask (digital halftone processing using blue noise mask) ”.

Another prior art technique for producing large, uniform dither arrays is what is called the "blank and crowd method." Looking further at the goals of this latter method,
It is to start with an initial pattern and take the pixels from the tightest "crowding" and insert them into the largest "blanks" to produce a homogeneous distribution of 1s and 0s. See IS & T / SPIE Sy
mposium on Electronic Imaging Science and Technolo
Robe in gy, San Jose, California, February 3, 1993
`` The Void-and-Cluster Method Fo by rt Ulichney
r Dither Array Generation (Blank and congestion methods for dither array generation) ", and Society for Info
rmation Display International Symposium, San Fose,
Robert Ulichney, California, June 14-16, 1994
By `` Filter Design For Void-and-Cluster Dither
Arrays (filter design for blank and dense dither arrays), the contents of which are incorporated by reference herein.

In particular, according to the article mentioned above, the initial pattern used to generate the dither array is "Protot
It is called “ype Binary Pattern” and the terms “cluster” and “void” are “minority pixels” and “majority pixels”. Is defined by When less than half of the pixels in the original binary pattern are of a particular type (either 1 or 0), they are called minority pixels,
Another type of pixel is called a majority pixel. By using these definitions for minority pixels and majority pixels, the terms "dense" and "blank" are meant to mean an array of fractional pixels with a majority pixel background. That is, "blank" is a large space between a few pixels and "dense" is a tight gathering of a few pixels. In the process of processing a prototype binary pattern, a few pixels are always added to the center of the largest void and taken out of the most tightly packed center to homogenize the pattern. Replaced by pixels).

The blank and dense methods are: (A) generating a prototype binary pattern; (B) "homogenizing" the prototype binary pattern by extracting pixels from the dense;
From the three individual steps of assigning them to the blanks to generate a homogenized initial binary pattern, and (C) assigning a threshold number of dither arrays to the resulting homogenized initial binary pattern. It will be.

More specifically, the generation of the prototype binary pattern is accomplished by using any suitable technique which produces an input pattern in which less than half of the pixels are "1" and the rest are "0". Done. For example, white
The noise pattern will act as a prototype binary pattern.

The next step in the method finds "blanks" and "crowds" to allow pixels to move between the blanks and the crowds. White space and congestion are located by filters that examine adjacent pixels that touch each pixel being processed. In particular, the blank finding filter considers neighboring pixels around all majority pixels in the prototype binary pattern, and the dense finding filter considers neighboring pixels around all minority pixels. In the above mentioned paper, to find voids and congestion,

[0028]

[Equation 1]

A two-dimensional Gaussian filter represented by the following function is used. During this process, the values in the array are iteratively processed, moving a single pixel several times from dense to blank. This process is iteratively performed until the whitespace stops becoming smaller and the congestion becomes less coarse. The process is considered complete when the largest void is created by moving a "1" from the tightest packing in the pattern. The resulting homogenized pattern is
It is called an initial binary pattern.

According to the third step of the method, a dither array is constructed from the initial binary pattern by assigning a threshold for the dither array using the initial binary pattern. In performing this step, the initial binary pattern and the dither array are multiplied in parallel in three phases by a predetermined algorithm. The threshold value is assigned by "rank" (ra) of "1" in the initial binary pattern.
nk) ”value for the cells of the dither array. That is, in the first phase I, the minority pixel "1" is removed from the initial binary pattern one at a time, while in parallel, the minority pixel remaining in the initial binary pattern is The number, or rank, is inserted at the dither array position corresponding to the removed "1" position. In each scene, the few pixels removed are
The pixel in the center of the tightest crowd.

In the second phase (Phase II), the initial binary pattern is again used as the basis for activation, a small number of pixels (ie "1") are inserted one at a time into the initial binary pattern, On the other hand, in parallel with this, the rank is inserted into the dither array. In each scene, the minority pixel to be inserted is inserted at the "0" position in the center of the largest blank.

In the final phase III, the initial binary pattern generated by adding "1" in phase II is used as the starting point. In this Phase III,
Since "1" is added in Phase II, there are more "1" s than "0" s in the initial binary pattern.
The meaning of "minority pixel" is changed from "1" to "0". Here, the initial binary pattern will be filled with more "1" s, while the dither array will be filled with rank values. At each step, a "1" is inserted for the "0" position identified in the tightest cluster of few pixels. At the end of Phase III, the initial binary pattern is filled with all "1s" and the dither array is filled with a unique rank value for each cell, from 0 to the maximum number of pixels.

This blank-and-dense method produces a dither array, which is generally followed by a high quality, halftoned image that is noiseless, but with some distortion in the tone. Yes, there is room for further improvement in the resulting image. All digital halftoning techniques assume a perfectly square pixel, and that group of square pixels completely covers the printable portion of the recording medium. If this is true in practice, the printer output density will be directly proportional to the scan input density of the original image. So in this ideal case,
This occurs when the output density Y exactly matches the input density X, as indicated by the 45 ° line (dashed line) in FIG.
However, in practice, the response of real electronic printing devices is generally non-linear. This is because such devices produce round dots rather than squares, rather than perfectly adjacent but non-overlapping square dots, which are outside the boundaries of the ideal pixel. Because it protrudes. This protrusion of dots allows adjacent dots to be completely connected under various printing conditions without leaving uncolored space (
For example, the conditions necessary for fully colored areas to be completely printed without uncolored voids).
Thus, since the actual dots are larger than expected in the halftoning algorithm, the printer output density Y which is a function of the input grayscale density X
Plot of (expressed as normalized% dot area)
Normally follows a curve similar to f (x) in FIG. (In FIG. 12, a large value on the coordinate axis corresponds to a low density, and thus the origin corresponds to the maximum density.) Therefore, the actually printed image is more than expected. High density, that is, it will appear dark.
Here, the exact shape of the function f (x) is determined by the characteristics of the printer and the halftoning algorithm employed.

To compensate for this effect, the function f (x)
Is measured mathematically through individual concentration measurements and regressions,
Then, the inverse (transformation) function f-1 (x) is similarly determined. This latter function, when applied to the input source data, offsets the effects of printer functionality and corrects the data, so the processing by the halftoning algorithm makes the density characteristics of the printed image more consistent with that of the original image. It is possible to generate an image adapted to do so.

This technique, known as "predistortion", produces better printed images than unmodified halftone image data, but when corrected to an ideal linear response,
The most visually desirable printed copy is not always produced. The reason is due to the complex interaction between the human visual system and the patterns generated by the halftoning algorithm for different scales. That is, the subjective response often depends on the halftoning algorithm employed and the degree of predistortion. See, for example, Trontelj et al., "Optimal Halftoning Algorithm Dependent on Print Resolution," SID 92 Digest 749 (1992).

No. 08 / 269,601 filed Jul. 1, 1994, assigned to the same applicant as the present application, multi-bit source data representing various densities of the original or source image is filtered by a filter function. Provided is a printing device and method thereof as modified. The filtered data is then processed by the halftoning module and converted into a raster of single bit data suitable for transfer to a printing device. This raster data does not linearly correspond to the input source data, is brighter than the density pattern of the source image in the highlight region and the high-side intermediate tone region, and is the source image in the shadow region and the low-side intermediate tone region. The darkness pattern is darker than the darkness pattern.

Furthermore, the filter may be applied to the parameters involved in the halftoning algorithm rather than the source data. For example, in the dither method using threshold arrays, the filter function can be applied to those thresholds. In that case, when the rasterized pattern is generated by applying the filtered threshold value to the source data in the reverse manner, the rasterized pattern is weighted according to a predetermined nonlinear output function, and the filter function is set to the threshold value. Applied to. According to this method, the modified array can be repeatedly applied to the source data by modifying one dither array only once, and thus the processing load can be considerably saved. For example, one 6
When using a 4x64 dither array, it is necessary to apply a filter function to each cell and process 4,096 data values. On the other hand, if the filter function is applied to each source data value, assuming that the sampling density for the source image is 720 × 720 dots per inch, then each image area of size 8 inches × 10 inches. For, it is required to process 40 million data points.

In general, many digital halftoning techniques assume ideally formed pixels, such as square pixels that completely cover the printable area of the recording medium. This ensures that under various printing conditions adjacent dots completely overlap without leaving any uncolored space (e.g., completely colored areas leave uncolored voids between them). Without leaving
To be printed continuously). FIG.
The hard copy output of a printer with a resolution of 1 / dpi is shown. Here, dpi is the number of dots per inch.
As shown here, ideal pixels 1001-1004 are illustrated as "on". Note that there are no voids between adjacent pixels.

These ideally formed pixels are actually approximations. The actual pixel is generally represented by a more circular dot. As those skilled in the art will understand,
The size or radius of these dots depends on many factors, such as the printing mechanism, the ink, and the recording sheet.

FIG. 7 shows an example where the diameter of the printed dots corresponding to the pixels is less than or equal to 1 / dpi. In other words, the radius r is 1 / {√2 × dp
It is less than i}. As shown, pixels 1001a and
There is a void 1100 between the dots representing 02a. Therefore, even if two adjacent pixels are on, there is a portion with a void. In this example, unacceptable results occur, so dots with radii less than or equal to 1 / {√2 × dpi} are not used.

FIG. 8 shows another example in which the radius of a dot representing a pixel is expressed by the following equation.

1 / {√2 × dpi} ≦ r <1 / dpi As shown, there are no voids between dots representing adjacent pixels, such as pixels 1001c and 1002c. However, the dots representing the pixels 1001c and 1002c overlap in the area 1200.

FIGS. 9A and 9B show another example of dots representing pixels having a radius expressed by the following equation.

1 / dpi ≦ r ≦ √2 / dpi FIG. 9A shows every other pixel in every other row for simplification of description. As shown, every other pixel in the same row, eg, pixels 1001d and 1002d.
(Or, in the case of the same column, for example, pixels 1002d and 1003d) overlap. However, the dots representing the pixels 1001c and 1003c do not overlap. All pixels are shown in FIG. In the figure, the dots representing pixels 1001d and 1002d occupy the area of pixel 1004d, so the dots representing pixel 1004d provide redundant information. Therefore, the ink used to print pixel 1004d is unnecessary. Furthermore, additional computation and printing time are required to print all adjacent pixels. Generally, the value of 1 / dpi at 720 dpi is 0.0014 inches and r is 0.
0015 inches and √2 / dpi is 0.0020 inches.

Furthermore, regarding dots of this size,
A plot of printer output densities (represented by normalized percentage dot areas) is shown in FIG. 12 (higher values along the axis correspond to lower densities and the origin corresponds to the maximum density), as indicated by the dotted line f (x ) Is represented as a function of input grayscale density, which normally follows a curve similar to Therefore, the printed image appears as a darker image than expected. Further, as the curved shape becomes more rectangular as shown, the dynamic range of the image becomes narrower. Generally, the dynamic range is about half. Therefore, the inverse (conversion) function f-1 (x) shown by the chain line
The dynamic range of is also about half.

FIGS. 10A and 10B show examples of dots representing pixels having the following radii.

√2 / dpi ≦ r This example is similar to the example shown in FIGS. 9A and 9B, except that dots representing pixels 1001e and 1003e (called alternating diagonal pixels) intersect. is there.

The resolution of a conventional printer is generally 300.
It is about 360 to 360 dpi. The resolution of a printer having a higher resolution can reach 720 dpi. As will be understood, the resolution when printed at a resolution of 720 dpi is about four times that at 360 dpi. In other words, when printing at 720 dpi, 72 prints per square inch.
It contains 0x720 or 518,400 dots.
On the other hand, when printing at 360 dpi, 360 × 360, that is, 12 in a print area of 1 square inch.
It contains 9,600 dots. However, according to the current technology, when printing at 720 dpi using the conventional printing mechanism, ink and / or paper, the resulting dot size as shown in FIGS. 9 (A) and (B) Is expressed by the following equation.

1 / dpi≤r≤√2 / dpi Therefore, printing at 720 dpi with this dot size results in printing redundant information and requires additional computation and additional printing time. And

[0050]

It is an object of the present invention to provide an apparatus and method for improving the resolution of halftoned images produced by a binary printing device.

Another object of the invention is to provide a method and apparatus that overcomes the problems of conventional printers.

Still another object of the present invention is to provide a method and apparatus for printing with high resolution, which requires less ink than the conventional apparatus and method.

Still another object of the present invention is to provide a method and apparatus for printing with high resolution, which requires less calculation as compared with the conventional apparatus and method.

Yet another object of the present invention is to provide a method and apparatus for applying non-linear predistortion processing without loss of dynamic range.

Yet another object of the present invention is to provide a method that is relatively easy to implement, either on dedicated hardware or on existing printer drivers.

Yet another object of the present invention is to reduce the tone distortion produced by the halftoning process using a dither array produced by the blank and congestion technique.

Yet another object of the present invention is to create a regularly distributed dot dither array of arbitrary size with evenly distributed dots.

A further object of the present invention is to generate the processing time for a regular distributed dot dither array.

[0059]

SUMMARY OF THE INVENTION According to one aspect of the present invention, a series of electronically encoded sequences consisting of n columns containing m pixels, each pixel specifying a grayscale value. A method and apparatus is provided for converting a continuous tone image, represented as an imaged n × m pixel, into a binary raster suitable for electronic printing. In this case, a series of pixels is identified corresponding to the linear segment of the image. The identified pixel group is, for example, an odd-numbered pixel of m pixels corresponding to an odd-numbered column of n pixel columns and an m-th pixel of an even-numbered column of n pixel columns. It is composed of an even-numbered pixel of the pixels. The identified groups of pixels are processed according to an order corresponding to movement along the segment in a predetermined direction,
Converted to binary raster values. Pixel identification and processing is repeated until the image is completely processed.

According to another aspect of the invention, image processing means for converting a continuous tone image into a binary array suitable for printing operations is electronically encoded to represent grayscale values. And a storage means for storing a series of n × m pixels that have been stored. The reading means then reads every other pixel from the memory means. The halftoning means iteratively processes the series of pixels corresponding to the linear segments of the image read by the reading means in an order representing the movement in the predetermined direction along the segment, to obtain the remaining binary data. To generate.

According to yet another aspect of the invention, a printer for printing a continuous tone image as a pattern of monochrome dots on a print medium includes a digitizing means responsive to the continuous tone image. The digitizing means comprises a stream of m × n pixel values, each pixel value being electronically encoded such that each pixel value represents a gray scale of a portion of a continuous tone image. Memory means receives the stream of electronically encoded pixel values and divides it into a plurality of linear segments to store the electronically encoded pixel values. Each linear segment has a beginning and an end and is composed of pixel values that represent a contiguous portion of a continuous tone image. The reading means reads every other electronically encoded pixel from the memory means.
The halftoning means processes each electronically encoded pixel value read by the reading means in each of the plurality of linear segments in start-to-end order and end-to-start order. Produces a static output. Upon receipt of this resulting output, the printing mechanism produces a pattern of monochrome dots on the print medium.

According to yet another aspect of the present invention, a computer system has a memory for storing data and programs. A central processing unit controls and integrates the operation of the computer system in response to a program stored in the memory, and a digitizing means responds to the continuous tone image to cause each pixel to be continuous. Generate a stream of electronically encoded pixels that represent the grayscale values of each part of the toned image. The memory means receives the stream of electronically encoded pixel values, divides it into a plurality of linear segments and stores the electronically encoded pixel values. Each linear segment has a beginning and an end and is composed of pixel values that represent a continuous portion of a continuous tone image. The reading means is
Every other pixel is read from the memory means. The halftoning means processes each electronically encoded pixel value read by the reading means in each of the plurality of linear segments in start-to-end order and end-to-start order. Produces a static output.
Upon receipt of this resulting output, the printing mechanism produces a pattern of monochrome dots on the print medium.

Furthermore, according to another aspect of the present invention, there is provided a printing apparatus and method capable of obtaining a series of pixel data corresponding to a linear segment of a source image. Every other pixel data is selected, and the selected pixel data is filtered according to the filter function. When printed, the filtered pixel data is a halftone pattern that represents a density pattern that is brighter in the highlight and high midtone regions than the source image and darker in the shadow and low midtone regions than the source image. This halftone pattern is printed on the recording medium.

According to yet another aspect of the invention, a printer has a source of a series of encoded pixel data representing density levels associated with a source image. The selector is
Every other pixel data is selected. The halftoning engine processes the selected pixel data to produce a density pattern corresponding to the source image that is brighter than the source image in the highlight regions and darker in the midtone and shadow regions than the source region. To generate a raster pattern. Then, ink or toner is supplied to the recording medium according to the halftoned pattern.

The method and apparatus of the present invention can be easily incorporated into the driver software of the printing device, or can be embodied as hardware at the printer port or at the printer itself, at a relatively low cost. Is. According to the present invention, it is possible to minimize the generation of a worm type noise pattern in the error diffusion halftone processing and generate a high quality halftone image.

While the preferred embodiment of the invention has been described in detail by the detailed description and accompanying drawings, those skilled in the art will appreciate that many modifications and variations of the invention will occur.
It should be understood that the invention is included in the technical scope of the present invention defined by the appended claims.

[0067]

DETAILED DESCRIPTION OF THE INVENTION The present invention is directed to an operating system that resides on a personal computer such as an IBM®, PS / 2® or Apple® Macintosh® computer. Is preferably implemented as. Although a typical hardware environment is shown in FIG. 1, what is illustrated here is a typical hardware configuration of a computer 100 in accordance with the present invention.
The computer 100 is controlled by a central processing unit 102, which is, for example, a conventional conventional microprocessor. Many other units are all interconnected via the system bus 108 to accomplish particular tasks.
Although a particular computer may include only some of the units illustrated in FIG. 1, or may include additional components not shown, most computers include at least the illustrated units. I'm out.

More specifically, included in computer 100 shown in FIG. 1 is a random access memory (RAM) 10 for temporary storage of information.
6, Read only memory (R) for permanent storage of computer configuration and basic operational commands
OM) 104, and input / output (I / O) adapter 110 for connecting peripheral devices such as disk unit 113 and printer 114 to bus 108 via cables 115 and 112, respectively. User·
An interface adapter 116 is also provided to connect input means such as a keyboard 120 and other known interface means such as a mouse, speaker and microphone to the bus 108.
The visible output is a display device 12 such as a video monitor.
Display adapter 11 connecting 2 to bus 108
Given by 8. The workstation is controlled and cooperated by the operating system resident therein.

A printing device, typically included in a computer system such as that shown in FIG. 1, is electrically connected to the computer system and produces a permanent image on a selected medium. To be controlled by it. To print documents that are displayed on the monitor or stored in memory, it is necessary to perform some operations. First, because print media is usually of fixed size, the printable information must be divided into pieces that are large enough to fit the selected media. Called. In addition to this, the information is reformatted from the format in which it is displayed or stored to a format suitable for controlling the printing device to actually perform the printing process on the medium. There is a need to. The reformatting process in this latter step may include a preceding process step in which the graphic representation is converted to the form used by the printing device by performing the halftoning process described above.

In general, computerization, although the pagination and reformatting processes necessary to convert printable information into a printable form, can also be performed by hardware configured for a particular configuration.
It is executed by a software program running in the system. The pagination is performed either by an application program that produces the initial output, or by the operating system, which is a collection of unit programs that perform basic file processing functions. The reformatting process, including the process of halftoning, is specific to the printing device and is typically contained in a software program called a "driver," which may be part of the operating system.
It is necessary to have a special association with a particular printing device. The driver program receives the information about the text or image from the computer system and performs the processing as described above to generate a signal that directly controls the printing device.

For example, FIG. 2 is a schematic illustration of a typical computer system using application programs, operating systems and printer drivers. Here, the computer system is represented schematically by the dotted box 200, the application program is represented by box 202, and the operating system is represented by box 206. The interaction between application program 202 and operating system 206 is schematically illustrated by arrow 204. This dual programming system is used in many types of computer systems, from large computers to personal computers.

However, the method for performing the printing process is computer dependent, and in this regard, a prior art personal computer system is shown in FIG. In order to provide the print processing function, the application program 20
2 is made to interact with printer driver software 210 (as indicated schematically by arrow 208). Printer driver software 210
Generally performs a halftoning process, but reformatted information, including embedded commands and converted graphic information, is shown schematically by arrow 214. It is possible to take another action to generate a stream of. The transformed information stream is then applied to a printer port 212 which contains circuitry to transform the incoming information stream into electrical signals. An "imaging engine", typically included in printer 218, is a hardware device or device for capturing the incoming stream of information and converting it into the electrical signals necessary to drive the actual print processing elements. RO
It is an M program type computer. The resulting "hard copy" output is available on the selected media. The apparatus for performing the error diffusion processing of the present invention is the printer 21.
8 may be incorporated in special hardware installed in the printer port 212 of itself.

As already mentioned, printing is performed at a resolution of 720 dpi and the radius is 1 / dpi≤r≤√2 / dpi.
A printer that has a dot that is i will print redundant information. Referring to FIG. 9B again, the pixel 100
The dots representing 1d and 1002d overlap in the area of pixel 1004d. Therefore, it is not necessary to perform any processing or printing on the pixel 1004d. According to the invention, only every other pixel in each row is processed and printed. For example, as shown in FIG. 17, only pixels in odd-numbered columns are processed in odd-numbered lines, and only pixels in even-numbered columns are processed in even-numbered lines (of course, even-numbered columns in odd-numbered lines are processed). Pixels and pixels in even-numbered lines and odd-numbered columns may be processed and printed). By printing in this way, 3
60 × 720 dpi or 259 per square inch,
A resolution of 200 pixels can be obtained. This resolution is twice the resolution of 360 dpi. As already mentioned, these pixels are halftoned for printing on a binary printer. For example, this type of processing includes error diffusion and dithering methods, which are described in detail below.

1. Error diffusion process The error diffusion process itself is well known, for example Robert Uli.
"Digital Halftoning" by chney (Robert Uricine) (MIT Press, Cambridge, Massachusetts and London, UK 1990, 239).
Pp. 319). During the error diffusion process, the pixels that make up the original image are processed in a line-by-line manner, with each line processing the pixels in a single direction (left to right, or right to left). A general line processing pattern is shown in FIG. 3, in this case each pixel line 300, 302, 304, 306.
And 308 are processed from left to right, then the next line is processed from left to right, and subsequent lines are sequentially processed from left to right until the entire image is processed from top to bottom. It is processed.

FIG. 4A shows a mode in which an error generated during the processing of each pixel is diffused to the adjacent pixels when the processing progresses from the left to the right along the line. That is, each pixel is processed by comparing its pixel value with a predetermined threshold, in which case
The pixel "value" is the original grayscale value plus the error adjustment value resulting from the previous processing of the other pixels. If this pixel value exceeds the threshold,
"1", that is, dots are output. Conversely, if the pixel value is less than the threshold value, "0" is output. That is, no dots are output. The error value is calculated by subtracting the actually output dot value from the next input value. This error value is then "diffused" or dispersed into several adjacent but unprocessed pixels.

This "diffusion" process is shown in FIG. 4A, where the pixels to be processed are shown as box 400. In the illustrated arrangement, the error resulting from the processing process goes to the pixel immediately to the right of processed pixel 400 (indicated by arrow 402), and arrows 404, 40.
It is spread to three adjacent pixels in the next line as shown at 6 and 408. The error value is multiplied by a proportional constant before it is added to adjacent pixels. The values of these constants q1, q2, q3 and q4 are q1 + q2 +
It is configured such that q3 + q4 = 1. For example,
The following weights can be used: q1 = 5/16, q
2 = 1/16, q3 = 7/16, and q4 = 3/16
Is. After the processing of pixel 400 is complete, the pixel to the right of pixel 400 is processed by adding a proportional error value to that pixel value and processing it in the same manner as pixel 400. After each pixel in a line is processed in this way, the next line in the image is processed in the same way.

The diffusion process is shown in FIG. 4B, where the pixel processing is from right to left, and the pixel is shown in box 410. In the configuration shown, the error resulting from the processing process is to the pixel immediately to the left of processed pixel 410 (indicated by arrow 412) and in the next three adjacent pixels of lines 414, 416 and 418. Spread to. Again, the error value is multiplied by a proportionality constant before it is added to adjacent pixels. These constants q
The values of 1, q2, q3 and q4 are q1 + q2 + q3 + q4 =
It is configured to be 1. For example, the following weights can be used. q1 = 5/16, q2 = 1/1
6, q3 = 7/16, and q4 = 3/16. After processing pixel 410, the pixel to the left of pixel 410 is processed by adding a proportional error value to that pixel value and processing it in the same manner as pixel 410. After each pixel in a line is processed in this way, the next line in the image is processed in the same way.

In general, when the processing sequence is from left to right, the error diffusion process produces a good image, but reduces the quality of the image such as "worm" and "snow plowing". A well-known byproduct noise pattern, also called, also occurs. "Warms" appear as curves or diagonal lines in a homogeneous gray image area and produce a patterned appearance in the halftone image. As such, various techniques have been used to reduce or eliminate these types of noise patterns.

US Pat. No. 4,955,065, JP-A-4
-37256 and some of the known techniques of this kind disclosed in JP-A-4-51773,
From a conventional raster scan pattern (where each pixel within each line of the image is processed from left to right, and the lines are processed from top to bottom):
The pixel lines are still processed from top to bottom, but the arranged pixel lines are turned into a serpentine scan pattern which is processed in alternating directions.

These techniques correct the worm,
While this causes the worm to appear as horizontal lines in the homogeneous gray area of the image, and thus be less noticeable, the worm is still present as it would be with a conventional raster scan.

The other known method for minimizing the appearance of the worm disclosed in Japanese Patent Laid-Open No. 3-151762 is also different in the processing sequence from the conventional raster pattern. According to one embodiment, the processing direction for each line of the image is selected based on the quantized "1" and "0" streams output from the random number generator or the noise pattern generator. Therefore, pixel processing for each line is performed from left to right (in response to "1") or from right to left (in response to "0"). According to another known embodiment, the pixels in the line are
Processing from left to right and from right to left according to a predetermined periodic pattern. For example, scanning twice from left to right and then twice from right to left. Although these conventional processing sequence techniques reduce the size of the worm, the worm is still present in the halftone image.

US patent application Ser. No. 08 / 269,708, dated July 1, 1994, assigned to the same applicant as the present application, describes the line-to-line processing direction as a warm noise pattern in a halftone image. A method and apparatus is provided that minimizes According to this technique, the processing direction of each line in the image is adaptively changed depending on the content of the image. More specifically, the processing direction of each line in the image is determined depending on the grayscale value and / or the quantization error associated with one or more pixels of previously processed lines in the image. To be

[First Embodiment] FIG. 11 shows a first embodiment of the present invention. As already mentioned, only every other pixel in each row, or alternating pixels, is processed by error diffusion. As shown in this figure, only odd column pixels are processed in the odd numbered rows and only even column pixels are processed in the even numbered rows. These pixels are called selected pixels. More specifically, in this error diffusion technique, each selected pixel is processed by comparing its value with a predetermined threshold. Here, the “value” of a pixel is a value obtained by adding an error adjustment value resulting from previous processing of another pixel to the original grayscale value.
When the pixel value exceeds the threshold value, "1", that is, a dot is output. Instead, if the pixel value is smaller than the threshold value, "0" is output, that is, no dot is output. The error value is then determined by subtracting the value of the dot actually output from the input value. This error is then "diffused", that is, distributed to surrounding pixels other than the unprocessed pixels.

As already discussed, the direction of error diffusion can be determined by various methods. FIG.
(A) and (B) show examples of error diffusion from right to left and from left to right, respectively.

This "diffusion" process is shown in FIG. In this case, the pixel being processed is the pixel 1501. As already mentioned, only every other pixel is processed. Therefore, the error diffusion processing is not performed on the unselected pixels. In the array shown in the figure, the error caused by the processing is as follows: the peripheral pixel 1502 on the right side of the processed pixel 1501 (skips the pixel that is directly adjacent to it), and the two peripheral pixels 1503 and 1506 (pixel 1501 on the next line). Pixels immediately below is skipped), and are further distributed to the three peripheral pixels 1504, 1505, 1507 in the next line.
The error value is multiplied by the proportional weighting factor before being added to the surrounding pixels. These coefficients, q1, q2, q3, q
The values of 4, q5, and q6 are q1 + q2 + q3 + q4
It is distributed so that + q5 + q6 = 1. For example, the following weighting factors are used: q1 = 7/16, q2
= 8/32, q3 = 1/32, q4 = 5/32, q5 =
8/32 and q6 = 3/32. These values are a preferred example, and those skilled in the art should be able to find other useful values. After pixel 1501 has been processed, pixel 1502 to the right of pixel 1501 is added with the error value prorated to the pixel value and then processed in the same manner as pixel 1501. After each pixel of one line has been processed in this way, the next line in the image is processed in the same way. Of course, the error is
It may be diffused to another line.

FIG. 11 shows the diffusion process when the processing of the pixel 1601 is performed from right to left. In the figure, the error due to processing is the processed pixel 160.
The pixel 1602 on the left side of 1 and the two peripheral pixels 1603 and 1606 on the next line and the three peripheral pixels 1604, 1605 and 1607 on the next line are dispersed. Before being added to the surrounding pixels, the error value is also multiplied by the proportional weighting factor. These coefficients q1, q2, q3, q
The values of 4, q5, and q6 are q1 + q2 + q3 + q4 +
It is distributed so that q5 + q6 = 1. For example, the following weighting factors can be used: q1 = 7 /
16, q2 = 8/32, q3 = 1/32, q4 = 5/3
2, q5 = 8/32 and q6 = 3/32. These values are a preferred example, and those skilled in the art should be able to find other useful values. After pixel 1601 has been processed, the pixel 1602 to the right of pixel 1601 is added to the pixel value with the proportional error value, and pixel 16
1602 is processed in the same manner as in the case of 01.

A schematic configuration of the error diffusion circuit in this embodiment is shown in FIG. The one in FIG. 5 is a preferred example, and those skilled in the art should understand that various modifications and changes can be made thereto.

The error diffusion circuit conventionally receives a stream 500 of grayscale image pixels represented by digital words from an imaging device (not shown). The stream of image pixels is continuous in the conventional input buffer 50.
Entered in 2. Buffer 502 is basically a buffer of sufficient capacity to store the input image pixel word for the entire pixel line.

On the other hand, the buffer 502 outputs the stored values to both the adder circuit 506 and the numerical value / logical operation circuit (hereinafter, referred to as numerical value / logical operation circuit) 514 in line order, and the buffer control circuit 510 (arrow 512). (Schematically indicated by). In response to the buffer control 510, as described above, the odd-numbered column pixels in the odd-numbered lines and the even-numbered column pixels in the even-numbered lines are output. The adder circuit 506 is used to subtract the "error" value caused by the processing of the preceding pixel, as described in detail below. Numerical / logical operation circuit 514 is used to control the processing direction according to the techniques described above. That is, under the control of the buffer control circuit 510 (this control is schematically indicated by arrow 512), the buffer 502 sequentially outputs the stored pixel data in a "left to right" order (last in, first out). Can be done in the order of "right to left" (first in first out)
It can also be output sequentially.

The input data stored in the input buffer 502 is sent from the input buffer output 504 to the threshold circuit 528 via the adder circuit 506 and the signal line 508.
The output of the threshold circuit 528 is quantized (“0” and “1”).
A binary image, which is a pixel value (each pixel “value” is the original input image value and an adder circuit 50).
(Composed of the "error" adjustment value added by 6) with a predetermined fixed threshold value, and outputs "1" if the pixel value is greater than the threshold value, and if the pixel value is less than or equal to the threshold value. It is generated by outputting "0". As an example, the threshold circuit 528 can use a fixed threshold (for example, when the grayscale value ranges from 0 to 1) of 0.5.

The quantized binary signal generated by the threshold circuit 528 is sent through the output line 532 to the second adder circuit 53.
Sent to zero. The second addition circuit 530 uses the signal line 508.
The above unquantized input signal is subtracted from the quantized signal output from the threshold circuit 528 on the output line 532 to calculate the quantization error value. The quantization error value is sent to the filter circuit 522. The filter circuit 522 calculates the diffused error value by multiplying the error value by a proportional constant,
One of these error values is added to the next pixel to be processed by the adder circuit 506 and the remaining error value is added during processing of the next pixel line.
18 is stored.

The error buffer 518 further outputs the selected stored value to both the adder circuit 506 and the numerical / logical operation circuit 514 (via the output line 524) (as indicated schematically by arrow 516). Buffer control circuit 510
Controlled by. Under the control of the buffer control circuit 510, the error buffer 518 outputs the error value to the input buffer 50.
In order to match the input data shifted out from 2, the stored error values can be sequentially output in the order of “left to right” (last in, first out), and the stored error values can be output from “right to right”. It is also possible to sequentially output in the order of "left" (first in first out).

During pixel line processing, the buffer control circuit 510 sequentially sends the input pixel values from the buffer 502 to the adder circuit 506 and also the error buffer 518.
Both the input buffer 502 and the error buffer 518 are controlled so as to sequentially send the error value from 1 to the adding circuit 506. The adder circuit 506 then sequentially sends the error diffused pixel values to a threshold circuit 528 which produces a quantized output. The output of the threshold circuit 528 is also sent to the output buffer 534, which also outputs the buffer control circuit 5.
Controlled by 10. The output buffer 534 outputs on output line 536 a continuous stream of binary pixels to be sent to the printing device.

2. Predistortion Processing [Second Embodiment] First, reference is made to FIGS. 13A and 13B showing the operation of the second embodiment of the present invention. In the case of this figure, instead of processing every pixel as is done by conventional techniques, only every other pixel is processed. Pixels to be used for processing are selected by the method already described. Further, instead of processing the source data according to the conversion function f-1 (x) which produces the linear response shown in FIG. 12, a different form of the function g (x ) Is used. When this function is applied to the source data, the source data is modified, and the final print image is subjected to the halftoning process so that
3 (B) so that a raster is generated that shows the variation of the normalized value (100 percent dot coverage) denoted as h (x) = f (g (x)) (because the printer This is because the function f (x) operates on the data pre-distorted by g (x), not on the source data x). Specifically, the highlight and high-side midtone areas of the original image (that is, the relatively light density areas) are expressed in lighter densities than the corresponding densities of the original image, while the shadow and low-side midtone areas are represented. (That is, a relatively dark density area) is represented by a darker density.

Here, the percent dot coverage is defined by the following equation.

[0096]

[Equation 2]

Here, DT = density of screened area Ds = density of plain area n = constant concerning light scattering characteristic of paper, and Dγ and Ds are values measured on paper.

The appropriate transformation function used in this embodiment is most conveniently based on the printer function f (x), which, when called, represents the input source data as it actually appears on the printer. Related to the concentration One preferred conversion function g (x) is:

G (x) = N−f (N−x) where x is the density of the source data points expressed as a numerical value. N is a gray scale density (for example, N = 255 in an 8-bit gray scale) that numerically shows the highest value represented by the source data, and corresponds to the zero point of the print density. Further, f is a printer function.

The function g (x) can also be parameterized in order to obtain a family of transformation curves that allows the user to easily control the output (ie the appearance of the final image according to h (x)). Such control is based on the actual density pattern,
Given the complex relationship between the pattern's visual effects, it may be necessary to achieve a visually optimal output image across various printing conditions, halftoning algorithms and image formats. Preferably the function g (x)
Is parameterized using the weighting factor w 1 and the available function families are:

G (x) = 2w [N−f (N−x) + x] + x However, 0 ≦ w ≦ 1. The results obtained are shown in Figure 1.
A curve group as shown in 3 (B) is obtained. The curve c is w
The output h (x) = f (g (x)) in the case of = 0.5 is shown. The curve a corresponds to the case of w = 1, in which case the output h (x) is equal to the function g (x). Conversely, curve e corresponds to the case of w = 0, in which case the output corresponds to uncorrected f (x). Curves b and d represent intermediate cases.

FIG. 15 shows a typical hardware configuration of this embodiment. In the figure, the multi-bit image data is
It arrives as a stream of N-bit values (eg, from a digital scanner that discretely samples the density of the original image over a fixed range of grayscale values) and is stored in a data buffer 20. N-bit data is
Transferred to a filter 22, which modifies each N-bit density value according to the conversion function described above. The user operation adjustment function 24, which may be in the form of a slide switch or a digitally programmable input device, modulates the action of the filter by multiplying the transform filter function by a selected value of the weighting factor w 1.

The modified N-bit image data is transferred to the halftone engine 26, and its conversion unit converts the input data into a set pattern of 1-bit elements according to an appropriate halftone processing algorithm. This embodiment is suitably used in connection with various algorithms, such as globally fixed level threshold algorithm, constrained average threshold algorithm, dynamic threshold algorithm, orthogonal tone scale generation. There are arithmetic methods, so-called electronic selection algorithms, various dithering algorithms (including regular dithering techniques such as diffusion dot type dithering) and error diffusion techniques.

The filter function of the above-mentioned form operates on the printer function, but cannot be used for the printer function which is symmetric with respect to the filter function. This is because the required S-shaped asymmetric output function h (x) is not generated even if the filter function is applied. For example, g (x) shown in the above equation cannot be used in conjunction with a lumped-dot dither halftoning algorithm that produces a symmetric printer function with respect to h (x). However, other mathematical constructs can be utilized to generate the required output function.

The raster output of the halftone engine 26 is sent to a conventional electronic printer 28 which deposits ink or toner on a recording medium in accordance with the raster pattern.

[Third Embodiment] In a third embodiment of the present invention, a halftoning algorithm is filtered instead of image data, and FIG. 15 shows the hardware configuration thereof. In this case, the halftone filter 30 uses the halftone engine 2 so that the final halftone output image is biased according to the function g (x).
Fix 6.

This embodiment can be most advantageously applied to array-based halftoning techniques where a predetermined (or gradually changing) sequence of numbers is applied to the input source data. Therefore, the present embodiment is particularly useful in cooperation with the dither method. In the present example, each array value is treated as an individual intensity level and a filter function is applied to it to obtain a modified value. Since the array values represent (or are converted to) thresholds rather than intensities, the inverse function g-1 (x) will be applied to the array values rather than g (x). Halftone engine 26
Applies this filtered dither array to the pixel data selected as described above. As described above, the selected pixel data consists of every other pixel data representing the source image, which is selected by the pixel selection circuit 21 shown in FIG.

An application example of this technique is shown in FIGS. 14A and 14B. Since the printed output density of the printer follows a typical f (x) pattern, the dither array threshold compared to the source data is generally the opposite intrinsic bias f
Adjusted to reflect -1 (x). By doing so, when comparing the input source data with the dither array threshold, the overall output density will be approximately the same as the input density. F required to obtain the final printed output image
In order to perform dithering according to the (g (x)) pattern (that is, application of the function g (x) to the input source data described above), the values in the dither array are inverse functions g-1 (x) = N-
Biased according to f-1 (N-x), FIG.
The curve c of (B) is generated. This curve c is the inverse of the required output curve, and as a result of the thresholding algorithm on the array values thus reverse biased, the final image matches the required output curve.

Here again, it is desirable to adjust the conversion function g-1 (x) so that the user can select the degree of density correction in order to adapt to a specific condition. This is done in the adjustment function 24, which weighs w.
Is used to obtain the following set of available conversion functions.

G−1 (x) = x−2w [x−N + f−1 (N−x)], where 0 ≦ w ≦ 1. A representative curve that follows this function is shown in FIG. Here, the curve c is w =
The final output f (g (x)) in the case of 0.5 is shown. Curve a
Corresponds to the case of w = 0 which does not reflect the dither correction,
Curve e represents the same final output case as function g (x), and curves b and d represent the intermediate case.

As described above, the values in the modified dither array are directly compared to the source data values as a halftoning threshold that determines whether the corresponding raster point is colored or uncolored. Alternatively, the modified dither array is conceptually placed on top of the source data value and the value of the corresponding dither array cell augments the density value of each original pixel, after which the augmented source data value is It is compared with a fixed threshold.

3. Generation of Dither Array [Fourth Embodiment] As described above, in the present embodiment, only every other pixel is processed. As shown in FIG. 17, only the pixels marked with Δ are processed. The other pixels are not processed or printed and are set to a non-printing value such as zero. If a conventional dither array generation process is performed, pixels that should not be processed, i.e., pixels that are set to zero, may have a printed value, and thus printing may occur at that pixel. is there. This can impair the advantages afforded by the present invention. Therefore, this embodiment modifies the dither array generation process by obtaining a dither array whose elements correspond only to the pixels to be processed. As described in detail below, the dither array and the intermediate array used to drive the dither array include only enough elements to accommodate every other pixel processed. In other words, for an image that is dithered by a p × q dither array and represented by n × m pixels, every other pixel, ie {n /
Only 2} × m pixels are processed. Thus, the dither array and all intermediate arrays contain lxk elements, each element corresponding only with respect to {n / 2} xm elements.
The method and apparatus for producing such a dither array is detailed below.

Generation of the dither array is done in the computer system shown in FIG. 1, but in general
It is done "offline" prior to the start of the halftoning process. In this manner, the printer driver software 210 only compares the image data with the dither array thresholds during image processing. FIG.
The typical activation pattern shown in (A) ("Prototype 2"
Value pattern)) can be used according to the prior art method of creating blank and dense dither arrays.
The pattern as shown in FIG. 19A is obtained by, for example, extracting the output of the white noise generator, comparing the white noise value with a certain fixed threshold value, and when the noise value exceeds the threshold value. "1" is inserted, and if the noise value does not meet the threshold value, "0" is inserted, and the noise value is quantized to generate a binary "1" and "0" pattern. You can This FIG.
The pattern shown in (A) is actually 4
Individual identical patterns 300, 302, 304 and 306
It consists of The prototype binary pattern used as the starting point in the prior art dither array generation method actually consists of one of those patterns, but how is the pattern used to tile a two-dimensional image? Four patterns are shown here to show how they can be used.

If the prototype binary pattern shown in FIG. 19A is used in the prior art blank and crowd method, then the pixels from the tightest crowd are taken and placed in the largest blank. By doing so, the pattern is directly homogenized. However, it has been found that the quality of the output image obtained using the generated dither array depends critically on the activation pattern used to start the process. A pattern, such as the pattern 300 shown in FIG. 19A, is not initially homogeneous due to the coarse quantization.

In accordance with the principles of the present invention, a more homogenous pattern is obtained before the further homogenization process is performed by taking the pixels from the tightest packing and placing them in the largest white space. Inevitably generated. The generation of this more homogeneous activation pattern begins with an array filled with equivalent gray scale values. Each of the values represents a constant gray tone with a value less than half the overall scale. For example, in an 8 bit per pixel system (with a maximum of 255 for each pixel), each element of the array is initially filled with the same number, where the number is greater than 127. It is a small value.

This is then followed by an error diffusion applied to this constant gray input (referred to as the "prototype gray pattern") to produce a more homogenous first, as shown in FIG. A binary pattern 308 is generated. As a result of this increased homogeneity, a more homogenous initial binary pattern is obtained even after the blank and congestion processing is complete, which in turn results in a better output. Note that in FIG. 19B, the error diffusion pattern is still composed of four separate patterns 308, 310, 320 and 322.

As already mentioned, the error diffusion process used to process the original gray pattern is applied to every other pixel, for example as shown in FIG. Therefore, a description of this error diffusion will be omitted here.

The binary pattern resulting from the application of the error diffusion process to the original gray pattern (hereinafter referred to as the "starting binary pattern") is the starting pattern for the homogenization step in the prior art blank and congestion processing. Can be used directly as However, it has been found that additional modifications to the activation binary pattern prior to performing the whitespace and congestion processing result in a more improved dither array. In accordance with the principles of the present invention, an additional modification to the activation binary pattern is made to account for "oversized" dots produced by printing devices such as ink jet printers and laser printers.

In particular, in the conventional blank and dense dither array generation techniques, the output pixels are square, so it is assumed that the pixels are adjacent to each other without overlap. The shape of this pixel is shown in FIG. In this figure, the pixels assumed by the prior art blank and congestion method are shown as square pixels 1003. However,
As shown in FIG. 9B, and as already mentioned, many printing devices actually produce circular dots, such as dot 1002d. Since the diameter of this dot is generally larger than the square pixel area, the dots corresponding to adjacent pixels overlap in order to remove the solid white spot in the "solid black" area. As already mentioned, the dots 1001d and 1002d are sufficiently overlapped that it is unnecessary to process and print the pixel 1004d located between these dots. Therefore, since the dot 1001d is larger than the pixel represented by this dot,
Overlapping areas are not considered in the blank and dense methods. According to the invention, in order to consider this overlapping area, three distinct overlapping areas (areas shown shaded) that occur when the oversized area shown in FIG. 18 overlaps with other pixels are considered. . Further, in the blank and dense method, the prototype binary pattern, the prototype gray pattern, and the activation binary pattern must contain as many elements as there are pixels to be printed. For example, if there are nxm pixels, every other pixel, or (n / 2) xm pixels, will be printed.
Therefore, each of the above patterns includes only elements corresponding to (n / 2) × m pixels. In practice, this is accomplished by replacing the array of these patterns by, for example, 45 degrees, as shown in FIG.

More specifically, what is shown in FIG. 18 is three "on" state pixels 600, 602.
And 604 and associated printer dots 60
6, 608 and 610, these dots 606,
608 and 610 are three pixel regions 600, 60
It is used to represent 2 and 604. Due to the overlap due to oversized printer dots, three types of areas can be used, either alone or in combination, to precisely adjust adjacent areas for dot overlap. These areas are illustrated as shaded areas in FIG. For example, the first type of area 612 is shown as an “α” region. The second type of area 614 is shown as a "β" region, and the third type of area 616 is what is called a "γ" region. The sizes of the areas α, β and γ are
Each is fixed on a specific type of printer,
It also depends on the printer dot radius and the printer resolution. The radius of a dot on any particular printer can be determined by examining the dots produced by the printer with a microscope. Once the dot radius is determined, the normalized radius (γ0) can be calculated by using the following formula.

[0121]

(Equation 3)

Here, "dpi" is the resolution of "dot / inch" of the printer. As already mentioned, the radius of the pixel targeted by the present invention is generally 1 / dpi ≦ r ≦
It is within the range of √2 / dpi. Once γ0 is calculated, the three regions α, β and γ are determined according to the following formula.

[0123]

[Equation 4]

20A and 20B, the prototype binary pattern (white noise pattern used in the prior art blank and crowd method) or the starting binary pattern (after error diffusion has been applied) is shown. Prototype gray pattern)
For either of the above, a correction method for taking the excessive dot into consideration is shown. Further, these figures show that the patterns were replaced by 45 degrees. FIG. 20 (A) shows a part of the pattern, which is a part of the pattern of FIG. 19 (A) or FIG. 19 (B) generated assuming, for example, a rectangular pixel. The pattern includes a rectangular array of "1" s and "0" s. For example, the array portion shown in FIG. 20A mainly consists of "0" s, and three "1" s are elements 700 of the array. , 702 and 704.

The array portion shown in FIG. 20B has been corrected for oversized dots. For example, the array element 708 in FIG.
It is adjusted by adding the value of "β" to the original value ("0"). The adjustment is from the overlap area created by the dots located in element 710. Therefore, this value has changed from the original "0" value in the corresponding element 706 in FIG. Similarly, the values in the array elements surrounding the three “1” s are as illustrated in FIG.
Adjusted according to the overlap area created by oversized dots. These adjustments impart "grayscale" values to surrounding elements to further refine the dot pattern. Because of this added gray scale value, this pattern is henceforth referred to as the "activation gray pattern".

According to the principle of the present invention (FIG. 19 (B))
The activation binary pattern (as shown in FIG.
Both activation gray patterns (as shown in 0 (B)) are then followed by an initial binary pattern and an initial gray pattern by moving pixels from the congestion to the voids according to the processing of the voids and congestion. -Generate a pattern. It has been found that the quality of the dither array produced by the blank and compaction method of the present invention is further enhanced due to the additional adjustment as illustrated in FIGS. 19 (B) and 20 (B). Has been done.

FIG. 21 is an exemplary flow chart of the blank and congestion method of the present invention incorporating the two improvements described above. Both of the refinements have been applied to the method illustrated in FIG. 21, but any of these refinement steps may produce:
It should be understood that it has nothing to do with the other. In particular, the method of the present invention begins at step 800 and proceeds to step 801, where the elements of the gray pattern are arranged to correspond to every other pixel (selected pixel). As shown in FIG. 18, the gray pattern is replaced by about 45 degrees so that each element of the gray pattern array corresponds to each selected pixel.
In step 802, a prototype gray pattern is generated, eg, using a constant value for each element of the array. Following this, in step 804, an error diffusion process is applied to the prototype gray pattern to produce a starting binary pattern. The activation binary pattern array contains as many elements as there are selected pixels, step 80.
6 described above and in FIGS. 20 (A) and 2
The activation binary pattern is adjusted for oversized dot patterns to produce an activation gray pattern, as shown at 0 (B). Following this, in step 808, the initial binary pattern and the initial gray pattern are moved by moving pixels from the dense to the blank according to the prior art blank and dense method.
Generated from the activation binary pattern and activation gray pattern, respectively.

Finally, in step 810, as shown in FIG.
A threshold is assigned to the dither array by using an initial binary pattern and an initial gray pattern using the method shown in FIGS. The routine ends at step 812.

22, 23 and 24 disclose a method according to the invention for co-locating to assign thresholds to a dither array using an initial gray pattern and an initial binary pattern. There is. Although the method according to this disclosure is similar to that used in the prior art blank and congestion method, this method is the same in the blank and congestion method as the initial binary pattern used in the prior art method. Instead of the initial gray
It has been modified to use patterns. Early gray
The pattern is tuned for oversized dots, thus providing a more accurate position for the tightest congestion and largest voids. In particular, the illustrated method consists of three processing phases. In the process of Phase I (shown in FIG. 22), the number of elements (“the number of 1s” −1)
Dither array elements from 0 to 0 are entered. Phase
In II (shown in FIG. 23), dither array elements from "number of 1s" to "lk / 2-1" are added,
And in Phase III (shown in FIG. 24),
Elements from "lk / 2" to "lk" are entered.

More specifically, the exemplary routine begins at step 900 and proceeds to step 902 where the gray pattern array is loaded with an initial gray pattern. This routine uses two arrays, called a "gray pattern" array and a "binary pattern" array. These arrays have as many elements as dither arrays (optionally, they form l columns and k rows). In step 902,
The gray pattern array is loaded with the values taken from the initial gray pattern array generated as described above.

In the next step 904, the binary pattern array is loaded with the values of the initial binary pattern array generated as described above. In the following step 906, the variable "1 count" is set equal to the number of 1's in the binary pattern, and in step 908 the variable "rank" is set equal to the value of "1 count"-"1". It

In step 910, it is checked whether the variable "rank" is less than zero. This check determines if Phase I of the threshold assignment routine has ended. If the "rank" value is not less than 0, the routine proceeds to step 912 to determine the location of the tightest congestion in the gray pattern array. As mentioned above, this determination is made using the two-dimensional Gaussian filter (described above). This filter takes into account adjusted grayscale values in the gray pattern to locate the tightest congestion.

In the next step 914, the new 2
The binary value "1" is removed from the binary pattern at the position determined in step 912 to form the value pattern. In the following step 916, the value of the variable "rank" is entered in the corresponding position of the dither array to form a new threshold.

In step 914, the fact that the removal of "1" from the binary pattern removed "1" and inserted "0" at the corresponding position was calculated,
Since the readjustment of α, β and γ is required, readjustment of the gray scale value is performed. As a result, in step 918, the gray pattern is recomputed from the new binary pattern generated by removing the "1" in step 914.

In step 920, the value of the variable "rank" is decremented and the routine returns to step 910 to further check if Phase I is complete. If so, the routine proceeds via connectors 922 and 924 to Phase II, shown in FIG. On the other hand, in step 910, if the value of the variable “rank” is not smaller than 0, step 9
12-920 is repeated until Phase I is complete,
The routine continues this aspect.

FIG. 22 shows the steps carried out in phase II of the threshold allocation method according to the invention. More specifically, the routine proceeds from connector 924 to step 926 where the previously determined initial gray
The gray pattern array is initialized by loading the pattern values. Similarly, step 928
At, the binary pattern array is initialized by loading the previously determined initial binary pattern values. In step 930, the value of the variable "rank" is set equal to the value of the variable "1 number".

In the next step 932, a check is made to determine if the value of the variable "rank" is less than the number of columns times the number of rows divided by 2 (lk / 2). Done. Since the value of the variable "rank" increases each time the loop of the routine is passed, the value of the variable "rank" is increased.
Finally exceeds the value (lk / 2), and step 9
As indicated at 32, the routine proceeds to Phase III of the threshold assignment method.

However, when it is determined in step 932 that the remaining "blanks" need to be processed, the routine proceeds to step 934 to re-use the two-dimensional Gaussian filter described above.
The position of the largest blank in the gray pattern array is determined.

In the next step 936, the binary value "1" is inserted in the corresponding position of the binary pattern, thereby replacing the original "0" and forming a new binary pattern. Then, in step 938, the value of the variable "rank" is placed in the corresponding position of the dither array.

Following this, the gray pattern array is recomputed from the new binary pattern for adjustment to the replacement of "0" s and "1" s in step 936. Then, in step 942, the value of the variable “rank” is incremented by 1. Following this, the routine returns to step 932, where the value of the variable "Rank" is checked again to determine if Phase II has ended. If Phase II has not ended, step 934.
-942 is repeated. The value of the variable "rank" is (lk /
When Phase II ends due to crossing 2), the routine proceeds via connectors 944 and 946 to Phase III of the threshold assignment routine as detailed in FIG.

Referring now to FIG. 24, the routine proceeds from connector 946 to step 948 where the definition of "minority pixel" is changed from "1" to "0" (as described above, Phase II. The binary value "1" is now a "minor" pixel, but is now a "major" pixel because the array is filled with a binary value of "1" during the process. Following this, the routine proceeds to step 950 where it is checked whether the value of the variable "rank" is less than the number of columns times the number of rows (lk). Phase II
At I, the value of the variable "rank" increases, so that it eventually equals lk, at which point the routine ends. At this point, each lxk element of the dither array corresponds to each of the selected (n / 2) xm pixels.

However, in step 950,
If it is determined that the value of the variable "rank" is less than lk, the routine proceeds to step 952, where the gray
The location of the tightest packing in the pattern array is determined. In the following step 954, a "1" is inserted into the binary pattern at the position determined in step 952 to form a new binary pattern. Then, in step 956, the value of the variable "rank" is input to the corresponding position in the dither array.

At step 958, the gray pattern array is recomputed from the new binary pattern and at step 960 the value of the variable "rank" is increased. The routine returns to step 950 and a determination is made whether Phase III has ended. If yes, step 9
The routine ends at 62, otherwise steps 952 through 960 are repeated until all elements in the dither array have been assigned a threshold value.

Although the above description has focused on a single color plane, those skilled in the art should understand that these techniques are applicable to each color plane. In other words, these techniques are applicable to color printing.

Furthermore, since all matters contained in the accompanying drawings and the above description are illustrative and not restrictive, it is out of the scope of the present invention when implementing the aforementioned technique. It is possible to make various changes and modifications without incident.

[Brief description of drawings]

FIG. 1 is a schematic block diagram of a computer system in which halftone processing using an error diffusion technique according to the present invention is operable.

FIG. 2 is a schematic configuration diagram of a computer system in the prior art showing a relationship between an application program, an operating system, and a printer driver on which halftone processing or preprocessing is performed.

FIG. 3 is an explanatory diagram showing a processing sequence for processing pixels in a prototype binary pattern by using an error diffusion technique.

FIG. 4A shows a method of selecting a proportional coefficient used for error diffusion processing when pixels are processed from left to right, and FIG. 4B shows pixels processed from right to left. Explanatory drawing which shows the selection method of the proportional coefficient used for error diffusion processing in a case.

FIG. 5 is a schematic diagram of the configuration of an error diffusion halftoning device.

FIG. 6 is a diagram showing a group of adjacent ideal pixels.

FIG. 7 is a diagram showing pixels with a radius r ≦ 1 / {√2 dpi}.

FIG. 8 is a diagram showing a pixel having a radius of 1 / {√2 dpi} ≦ r ≦ 1 / dpi.

FIG. 9A shows a radius of 1 / dpi ≦ r ≦ √2 / dpi.
i show every other pixel on every other line that is i,
(B) is a diagram showing adjacent pixels each having a radius of 1 / dpi ≦ r ≦ √2 / dpi.

FIG. 10 (A) shows that the radius is √2 / dpi <r1.
Shows every other pixel on every other line, (B) is
FIG. 3 is a diagram showing adjacent pixels with a radius of √2 / dspir ≦ 1.

FIG. 11 is a diagram showing an error diffusion technique applied to every other pixel.

FIG. 12 is a characteristic diagram showing how the output of an electronic printer tends to deviate from the source continuous tone image in terms of density level and how this deviation is corrected in the prior art.

FIG. 13 is a characteristic diagram showing a parameterized family of conversion functions derived in accordance with the present invention and the resulting printer output.

FIG. 14 is a characteristic diagram illustrating a parameterized family of conversion functions derived for application to a dither array.

FIG. 15 is a block diagram showing a preferred hardware configuration of a second aspect of the present invention.

FIG. 16 is a block diagram showing a preferred hardware configuration of a third aspect of the present invention.

FIG. 17 is a schematic diagram showing every other pixel to be processed in accordance with aspects of the present invention.

FIG. 18 is an explanatory diagram showing three adjustment areas used to correct the initial binary pattern for oversized dots.

FIG. 19 (A) shows a prototype binary pattern generated using white noise, and FIG. 19 (B) is an improvement generated by applying blank and congestion techniques to the pattern shown in (A). Explanatory drawing which shows the produced prototype binary pattern.

FIG. 20 (A) shows a portion of an initial binary pattern generated by prior art blank and congestion techniques,
FIG. 3B is an illustration showing the same portion of an initial binary pattern generated by the blank and congestion technique improved according to the principles of the present invention.

FIG. 21 is a process flow diagram of an improved blank and congestion technique.

FIG. 22 is a Phase I flow diagram of a process for adding a threshold to a dither array in accordance with the principles of the present invention.

FIG. 23 is a phase II flow diagram of a process for adding a threshold to a dither array in accordance with the principles of the present invention.

FIG. 24 is a Phase III flow diagram of a process for adding a threshold to a dither array in accordance with the principles of the present invention.

[Explanation of symbols]

21 Pixel Selection Circuit 24 User Operation Adjustment Function Unit 26 Halftone Engine 28 Printer 30 Halftoning Filter 300, 302, 304, 304 Prototype Binary Pattern 308, 310, 320, 322 Initial Binary Pattern 502, 600, 604, 608 , 1001, 100
2, 1003, 1004, 1005 Oversized dots 612, 614, 616 Overlapping area 1501, 1502, 1503, 1504, 1505,
1506, 1507, 1601, 1602, 1603,
1604, 1605, 1606, 1607 Selected pixel

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification number Internal reference number FI Technical display location G06F 3/12 L G06T 5/00 H04N 1/41 B B41J 3/12 G G06F 15/68 320 A

Claims (28)

[Claims] 1. A method of printing, comprising: a. Obtaining a series of pixel data corresponding to linear segments of the source image; b. Selecting every other pixel data acquired in step a, and c. Filtering the pixel data selected in step b according to a filter function, d. The pixel data filtered by the process c is
The process of processing into a halftone pattern that, when printed, has a density pattern that is brighter than the source image in the highlight and high midtone areas and darker than the source image in the shadow and low midtone areas. , E. Printing a series of dots on a recording medium according to the halftone pattern. 2. The method according to claim 1, wherein the printing step is carried out by an electronic printer having a printer conversion function f, wherein the density of pixel data of the source image is x, and the highest gray scale represented by the source image. The method wherein the filter function is g (x) = Nf (Nx), where N is the density. 3. The method according to claim 1, wherein the printing step is performed by an electronic printer having a printer conversion function f, wherein the density of pixel data of the source image is x, and the highest gray scale represented by the source image. The density is N and 0 ≦ w ≦ 1
Then, the filter function is g (x) = 2w {N
-F (N-x) -x}, further comprising the step of selecting an appropriate value for w. 4. The method of claim 1, wherein the processing step is accomplished by an error diffusion method. 5. The method of claim 1, wherein the processing step is accomplished by dithering. 6. The method of claim 5, wherein the processing step is accomplished by a distributed dot dither method. 7. The method of claim 5, wherein the processing step comprises: a. Associating individual pixel data of the source image with binary raster points; b. Applying a dither array to the pixel data of the source image; c. Comparing the pixel data to which the dither array has been applied with a pixel threshold, and d. Selecting the binary number of the corresponding raster point based on whether the pixel data to which the dither array is applied exceeds a threshold value. 8. The method of claim 5, wherein the processing step comprises: a. Associating individual pixel data of the source image with binary raster points; b. Sequentially comparing the pixel data of the source image with the numbers in the dither array, c. Selecting the binary number of the raster point based on whether the pixel data exceeds the corresponding numerical value. 9. The method according to claim 1, wherein, when dpi is a resolution of a pixel expressed by the number of dots per inch, the radius r of each dot is 1 / dpi ≦ r ≦ √2 /
a method of being within a range of dpi. 10. A method of printing, comprising: a. Obtaining a series of pixel data corresponding to linear segments of the source image; b. Selecting every other pixel data acquired in step a, and c. When the pixel data selected in step b is printed, a density pattern that is brighter than the source image in the highlight region and the high-side intermediate tone region and darker than the source image in the shadow region and the low-side intermediate tone region Processing into a halftone pattern showing d. Printing a series of dots on a recording medium according to the halftone pattern. 11. The method of claim 10, wherein the processing step comprises: a. Providing a dither array having a plurality of elements, each element associated with a numerical value; b. Associating individual pixel data of the source image with binary raster points; c. Modifying the value of the element according to a bias function, d. Applying the dither array to the pixel data of the source image by sequentially comparing the density values of the pixel data of the source image with the values of the dither array; e. 2 of the corresponding raster points based on whether the pixel data of the source image exceeds the numerical value of the corresponding element.
A step of selecting a number of values, and a method comprising :. 12. The method of claim 10, wherein the processing step comprises: a. Associating individual pixel data of the source image with binary raster points; b. Providing a dither array, each element having a plurality of elements associated with a numerical value; c. Modifying the value of the element according to a bias function, d. Applying a dither array to the pixel data of the source image by adding the numerical value of the element to each pixel data of the source image; e. Comparing the pixel data to which the dither array has been applied with a pixel threshold, and f. Selecting the binary number of the corresponding raster point based on whether the pixel data to which the dither array is applied exceeds a fixed threshold value. 13. The method according to claim 11, wherein the printing step is performed by an electronic printer having a printer conversion function f, wherein the density of pixel data of the source image is x, and the highest gray scale represented by the source image. The method wherein the bias function is g-1 (x) = Nf-1 (Nx), where N is the concentration. 14. The method of claim 12, wherein the printing step is performed by an electronic printer having a printer conversion function f, wherein the density of pixel data of the source image is x, and the highest gray scale represented by the source image. The method wherein the bias function is g-1 (x) = Nf-1 (Nx), where N is the concentration. 15. The method of claim 11, wherein the printing step is performed by an electronic printer having a printer conversion function f, wherein the pixel data density of the source image is x, and the highest gray scale represented by the source image. The density is N and 0 ≦ w ≦ 1
, The bias function is g-1 (x) = x-2
w {x−N + f (N−x)}, further comprising the step of selecting an appropriate value of w. 16. The method of claim 12, wherein the printing step is performed by an electronic printer having a printer conversion function f, wherein the pixel data density of the source image is x, and the highest gray scale represented by the source image. The density is N and 0 ≦ w ≦ 1
, The bias function is g-1 (x) = x-2
w {x−N + f (N−x)}, further comprising the step of selecting an appropriate value of w. 17. The method of claim 10, wherein dp
When i is the resolution of a pixel expressed by the number of dots per inch, the radius r of each dot is 1 / dpi ≦ r ≦ √
A method of being within a range of 2 / dpi. 18. A printing apparatus comprising: a. A source of sequentially coded pixel data representing density levels associated with the source image; b. A selector for selecting every other pixel data, c. A filter for modifying the pixel data selected according to the filter function, d. When printed, it is brighter than the source image in the highlight and high midtone areas, and
A halftone engine that processes the modified data into a halftone pattern that exhibits a darker density pattern in the shadow and lower midtone regions than the source image; e. And a unit for supplying dots of ink or toner to a recording medium according to the halftone pattern. 19. The apparatus according to claim 18, wherein the printing device has a printer conversion function f, the density of pixel data of the source image is x, and the maximum grayscale density represented by the source image is N. , The filter function is g (x) = Nf (Nx). 20. The method according to claim 18, wherein the printing device has a printer conversion function f, the density of pixel data of the source image is x, the highest grayscale density represented by the source image is N, and 0 ≤w≤1
Then, the filter function is g (x) = 2w {N
-F (N-x) -x}, further comprising means for selecting an appropriate value for w. 21. The apparatus of claim 18, wherein the halftone engine operates according to an error diffusion method. 22. The apparatus of claim 18, wherein the halftone engine operates according to the dither method. 23. The apparatus of claim 22, wherein the halftone engine operates according to a distributed dot dither method. 24. The device of claim 18, wherein dp
When i is the resolution of a pixel expressed by the number of dots per inch, the radius r of each dot is 1 / dpi ≦ r ≦ √
A device characterized by being in the range of 2 / dpi. 25. In a printing device, a. A source of sequentially coded pixel data representing density levels associated with the source image; b. A selector for selecting every other pixel data, c. A halftone engine that processes the selected pixel data into a raster pattern that corresponds to the source image and is brighter than the source image in the highlight region and darker than the source image in the intermediate tone region and the shadow region. And d. And a unit for supplying dots of ink or toner to a recording medium according to the halftone pattern. 26. The apparatus according to claim 25, wherein the printing apparatus has a printer conversion function f, the density of pixel data of the source image is x, and the maximum grayscale density represented by the source image is N. , The halftone engine functions g-1 (x) = N-f-1 (N
A device characterized in that it is biased according to -x). 27. The apparatus according to claim 25, wherein the printing apparatus has a printer conversion function f, the density of pixel data of the source image is x, the maximum grayscale density represented by the source image is N, and 0 ≤w≤1
Then, the halftone engine uses the function g-1.
An apparatus characterized in that it is biased according to (x) = x-2w {x-N + f (N-x)} and further comprises means for selecting an appropriate value for w. 28. The device of claim 25, wherein dp
When i is the resolution of a pixel expressed by the number of dots per inch, the radius r of each dot is 1 / dpi ≦ r ≦ √
A device characterized by being in the range of 2 / dpi.