JPH0336879A - Intermediate tone reproduction system - Google Patents

Abstract

PURPOSE:To obtain gradation expression with excellent reproduction capability of minute detailed part and sharpness by providing an energy gap equivalent to the sum of control levels in a threshold level matrix between a point in which the control level is not uniform of all recording dots and a point in which the control level is nearly uniform of all recording dots. CONSTITUTION:A graduation picture signal from a graduation picture signal source 10 is inputted to a multi-value dither circuit 12 via a filter 11, the circuit compares a gradation picture signal inputted for each dot with each element of (L-1) sets of threshold level matrices (TM) each comprising M elements to select the control level for each recording dot among L sets of levels 0-(L-1), an L level control level signal is outputted to a head drive circuit 13 for each recording dot to control the energizing energy of a thermal head 14. There is energy gap in the TM between the state of nearly unsaturation in which the control level to the recording energy is not uniform for all dots and the state of the saturation state in which the control level to the recording energy is uniform, and the later is used for edge emphasis.

Description

[Detailed Description of the Invention] [Object of the Invention] (Field of Industrial Application) This invention relates to a halftone reproduction method for recording halftone images in a thermal melt transfer printer, an optical printer, or other printers, and particularly relates to a halftone reproduction method for recording halftone images in a thermal melt transfer printer, an optical printer, or other printers. This paper relates to a halftone reproduction method using a multilevel dither method.

(Prior art) Thermal melt transfer recording, which uses heat-melting ink with wax or resin as a binder, has the advantage of being able to obtain high-contrast images with a simple recording process, so it is widely used in color graphic printers and ordinary printers. It is widely used in paper facsimile machines, Japanese word processors for personal e-niss, etc.

Incidentally, as a method of reproducing natural images including smooth gradations and subtle colors using a binary recording printer, systematic dithering has been widely used. However, since the systematic dither method has the characteristic that the number of reproduced gradations and the resolution are contradictory, it is difficult to satisfy both the resolution and the number of gradations at the same time. In order to improve this point, each threshold value in the threshold matrix of the systematic dither method is made multi-valued and the threshold matrix is multi-layered, thereby reducing the matrix size and increasing the number of reproduced tones. law is being considered.

Conventionally proposed multilevel dither methods can be roughly divided into the following two types. One is to set all of the matrix elements to an intermediate multi-value level when the density is moderate, excluding areas near the ends of the lowest density and the highest density, and is often used in multi-value displays. The other is low concentration,
The idea is to set only one of the matrix elements to an intermediate multivalue level when the density is moderate, excluding areas near both ends of high density, and this uses area gradation similar to binary recording. It is said to be suitable for non-ingot printers. All of these methods are based on the simple dot placement rule used in the binary dither method, and the multi-value dither method is a method that multi-layers the threshold matrix according to a monotonous control increase/decrease rule that is consistent over the entire variable range. Although there have been many reports on its application to thermal recording, displays, ink jet printers, laser printers, etc., there have been few reports on its application to thermal melt transfer printers. On the other hand, we use several distinctive pixel patterns to suppress "roughness" (transfer noise) and false contours (gradation jumps) that tend to occur when expressing gradations using thermal melt transfer.
A density pattern method has been announced that realizes continuous gradation by controlling the diffusion of heat (heat bleeding, interference) across multiple dots.

When a gradation image is reproduced by thermal transfer using the conventional binary dither method, a rough impression and false contours due to gradation jumps are likely to occur. In order to investigate these recording noises in more detail, we conducted an experiment in which a specific dot pattern was continuously and repeatedly recorded, the recording density was measured against the recording energy, and the change in image quality was observed at a resolution of 12 dots/l. Using a thermal head, recording speed 2
The test was carried out under the conditions of m5ec/line. Figure 25 shows 2
It is a micrograph showing the shape of transfer ink when only one dot is recorded in a pixel of ×2 dots. As shown in the figure, the isolated dot-shaped recording pixels arranged at appropriate intervals form a stable transfer state with little fluctuation in area, that is, little "roughness." However, when two dots are recorded in a 2×2 dot pixel and the transfer area ratio is close to 50%, adjacent dots interfere with each other, affecting the image quality.

FIG. 18 shows characteristics for three pixel shapes determined by the arrangement of two dots in 2×2. In this graph, the solid line in the graph indicates a region that has been determined to have very little "roughness" based on subjective evaluation, and the broken line indicates a region that has been determined to have a large amount of "roughness." In addition, as a result of experiments using various pixel shapes such as 3.times.3 dots and 4.times.4 dots as parameters, the following common facts were found.

■There is recording noise peculiar to area gradation (this causes "roughness") due to the pixel shape.

■There are some patterns that show good image quality within a specific density range.

(2) A pattern in which dots are arranged linearly in the sub-scanning direction (stripe pattern) has the least recording noise ("roughness") in a specific intermediate density area.

■In addition to noise that depends on materials and mechanisms such as the smoothness of the recording paper and the pressing pressure of the thermal head, there is also noise that depends on the control of pulses applied to the thermal head and the processing of gradation recording. It is shown that. ■ suggests that by skillfully tracing a few specific pixel shapes, it is possible to reproduce the entire density range with good image quality, and this is due to the method of recording continuous gradation using the density pattern method. This is the theoretical basis. In fused thermal transfer recording,
In order to consider the threshold matrix configuration for the multivalued dither method to obtain high image quality, it is necessary to design a gradation control method based on these facts.

Although the stripe pattern that continues in the sub-scanning direction described in (2) has the above-mentioned features, it also shows the shape of the transfer ink when it is continuously recorded in one row within a 3 x 3 dot pixel.
As is clear from the micrograph in the figure, irregular cross-linking occurs between adjacent stripes in high-density regions, resulting in extremely poor image quality. Furthermore, while it is better for the texture to have no phase fluctuation or directionality, there is a problem in that the fundamental frequency of the stripe pattern varies greatly depending on the direction.

Therefore, the present inventors developed a striped
Instead of a pattern, I decided to focus on a grid pattern as a basic pattern. The reason for this is, first of all, that the pattern includes stripes, that is, it maintains the property of being less unstable during the ink transfer process to some extent. Second, there is little variation in the fundamental spatial frequency depending on the direction, that is, texture noise is visually inconspicuous. Thirdly, and this is the most important reason, in the case of a grid pattern, the transferred ink has sharp protrusions on all sides, so the "roughness" that occurs when pixels bridge each other is avoided. It's a small point. Figure 27 is 4x4
This is a micrograph showing the shape of the transfer ink when it is recorded in a grid pattern within the pixels of dots, and it can be seen that the "roughness" is significantly reduced.

FIGS. 19 and 20 show the results of measuring fluctuations in the area of each pixel as recording noise with respect to the average recording density of a concentrated dot pattern and a lattice dot pattern, respectively.

As a result of analyzing micrographs of areas where recording noise increases, as indicated by arrows in the figure, we found that the 21st Figure a.

The existence of four types of noise conceptually shown in b, c, and d was confirmed. In other words, these four types of noise are: (a) noise that occurs due to uncertain ink transfer when the amount of recording energy is small, (b) noise that occurs due to the fusion of dots, and (C) noise that occurs due to instability between pixels. This is the noise caused by crosslinking, especially when d is at a high concentration, where parts that should be white are turned black. Below, the noise of a is “jumped”
The noise in b is called "fusion of pixel points," the noise in c is called "bridge between pixels," and the noise in d is called "collapse."

Of these noises a, b, c, d, each -1
FIG. 22 shows the results of selecting regions where each type appears independently and separating the magnitude of area ratio fluctuation. In general, the reason that the concentrated dot type is said to have better gradation than the dispersed dot type among the dither methods is that the dot placement rule is that the dots are always adjacent to the stepped dot group or decrease. , in that it is possible to avoid fusion noise of image points. However, in the dot concentration type, among the four types of area ratio fluctuations, the influence of the "inter-pixel bridging" noise of C appears the most. Therefore, in order to reduce the "roughness" of thermal melt transfer recording, it is most important to suppress this "pixel-to-pixel bridging" noise.

Comparing Fig. 19 and Fig. 20, it is found that the grid-like dot pattern (second
Figure 0) shows that there is less recording noise. In other words, in the case of a 3x3 dot matrix, the area ratio fluctuation of about 0.068 that occurs with a concentrated pattern of 4 dots is reduced to about 0.034, which is about 50, by using a lattice dot pattern of 5 dots. %Decrease. It is experimentally clear that in order to suppress bridging noise between pixels in this way, it is sufficient to use a dot arrangement rule that forms a lattice-like dot pattern from a low-density region.

Applying complex rules to a multilevel dither with three-dimensional degrees of freedom has hardly been considered until now, but as mentioned above, grid-like dot patterns form stable pixels and are highly effective. It has become clear that the gradation characteristics are very high. Therefore, the inventors developed an organized multi-value matrix that forms a multi-value matrix with only the weak constraint that ``when the gradation changes gradually, it must pass through a grid-like multi-value density pattern.'' A halftone recording method using the dither method has already been proposed (Japanese Patent Application Laid-open No. 47547/1983).

However, in the above-mentioned halftone recording method using the multi-level dither method, there is a contradictory tendency that when emphasis is placed on smooth gradation expression, the reproducibility and sharpness of fine shape parts are sacrificed.

(Problems to be Solved by the Invention) As mentioned above, in the conventional halftone recording method, when trying to achieve smooth gradation expression, there was a problem in that the reproducibility and sharpness of fine shape parts were impaired. .

SUMMARY OF THE INVENTION It is an object of the present invention to provide a halftone reproduction method capable of expressing gradation with excellent reproducibility and sharpness of fine shape portions while using a multilevel dither method.

[Configuration of the Invention (Means for Solving the Problem) The present invention converts a gradation image signal into a predetermined control level using a threshold value matrix, and controls recording dots using this control level, thereby converting a gradation image signal into a predetermined control level. In the halftone reproduction method that expresses levels, there is a first point where the gradation level is located at a gradation level lower than the maximum gradation level and the control level is not uniform for all recorded dots, and a first point where the control level is not uniform for all recorded dots. and a substantially uniform second point, there is a gap such that the total value of the control levels within the threshold matrix at the first point and the second point is discontinuous. .

More specifically, a multilayer threshold matrix consisting of M elements is used to convert the gradation image signal into L-value control levels from O to L-1 (L is an integer of L≧3), By controlling the recording dots, N gradation levels [M<N
≦M (L-1)], the sum of the values within the matrix size of the control level for the gradation image signal is the maximum gradation level and the maximum gradation. The threshold matrix is characterized in that the threshold value matrix is determined so as to take a maximum value between the level and the gradation level one level before.

In the present invention, it is desirable to perform high frequency enhancement processing on the gradation image signal.

Further, the value of L of the threshold matrix is preferably in the range of 7≦L511, and furthermore, the size of the threshold matrix is 180
~250 μm opening is preferred.

(Function) When the threshold matrix is determined in this way, by combining the multilevel dither method with a high-frequency emphasis filter, it is possible to achieve not only smooth gradation expression, but also gradation expression with excellent reproducibility and sharpness of fine shape parts. This makes it possible to record high-quality images, especially images that include a mixture of gradation images such as photographic images and binary images such as character images.

Also, it becomes possible. When the value of L in the threshold matrix, that is, the number of gradation control levels, is set to 7 or more and 11 or less,
Smooth gradation expression without false contours can be achieved without unnecessarily increasing the number of gradations, that is, without unnecessarily increasing the circuit scale.

Furthermore, the size of the threshold matrix is set to 180 to 250 μm.
By using it, recording noise (area ratio fluctuation) is reduced, and an image with less roughness 2 can be obtained.

(Example) Hereinafter, the present invention will be described in detail with reference to the drawings. FIG. 1 is a block diagram showing the main structure of a thermal melt transfer printer to which the present invention is applied.

In FIG. 1, a gradation image signal source 10 consisting of a memory, etc.
The gradation image signal from the gradation image signal is inputted to a multilevel dithering circuit 12 via a high-frequency emphasis filter 11. Here, the gradation image signal is a signal representing the density to be reproduced for each pixel of a photographic image or character image. The multilevel dithering circuit 12 determines the control level of each recording dot by comparing the gradation image signal manually input for each dot with each element of the (L-1) layer threshold matrix consisting of M elements. is selected from among L values from O to (L-1) and outputted to the head drive circuit 13 as an L value control level signal for each recording dot, and is configured, for example, by a ROM table as described later. Ru.

The L-value control level signal output from the multi-level dithering circuit 12 is supplied to the head drive circuit 13, and is converted into a pulse width for controlling the energization energy (recording energy) applied to each heating resistor dot of the thermal head 14. It is converted and supplied to the thermal head 14. The thermal head 14 heats and melts the ink by selectively applying electricity to heating resistors arranged in a line while pressing the recording paper 15 against the platen roller 17 side via the ink ribbon 16. 15. The ink transferred and attached to the recording paper 15 forms a recorded image 18 7, and one other ink remains on the ink ribbon 16. The timing controller 20 supplies necessary timing signals to the gradation image signal source 10, the multilevel dither circuit 12, and the head drive circuit 13.

FIG. 2 is a block diagram showing details of the multilevel dithering circuit 12. The gradation image signal data (8 bits) from the gradation image signal source IO is stored in the buffers (RAM1, RAM2) 21a.
, 21b. Here, assuming that the thermal head 14 is a 25BO dot/line head in which 2560 heating elements are arranged in the main scanning direction (the width direction of the recording paper perpendicular to the traveling direction of the recording paper), the two buffers 21a, 2
The capacity of 1b is 8 bits x 2580 dots, each corresponding to one line of gradation image signal data. The reason why two buffers 21a and 21b are used is to support high-speed recording. That is, the first line of data is written to the buffer 21a, and the next line of data is written to the buffer 21b. After that, the buffers 21a, 2
Data of each line is written alternately to 1b. By doing this, while writing data to the negative buffer,
Data can be read from the other buffer. The buffers 21a and 21b enter a standby state when writing of one line of data is completed. When recording of one line is completed, a data read signal RAMIRD or RAM2RD is generated, and the data of the second line is read from the buffer 21g or 21b.

This situation is shown in the timing diagram of FIG.

Such alternating writing and reading operations of gradation image signals for 2560 dots are repeated until recording of one page is completed (in the case of color recording, until recording of the first color is completed). According to this, the printing cycle can be set to 2 m5ec/line.

Data read from the buffer 21a or 21b by the RAMIRD signal or the R'A M 2 RD signal is supplied to the multi-value threshold matrix ROM 22. This threshold value matrix ROM 22 is a part related to the gist of the present invention, and performs conversion from a gradation image signal to a multivalued control level. The contents of this threshold matrix ROM 22 are such that when an input level and a recording dot position (x, y coordinates) are given as an address, control level data for the corresponding dot is output.

The dot position data (x, y coordinates) given as part of the address in the threshold matrix ROM 22 is obtained from the output (2 bits) of the line counter 23 and the output (2 bits) of the dot counter 24 in FIG. It will be done. That is, as is clear from the timing diagram of FIG.
It performs a counting operation from 0 to 3 in synchronization with the RD signal and outputs the X coordinate. Also, the line counter 24 is RA
Every time the MIRD or RAM2RD signal is counted 2560, it counts up and outputs the y coordinate from 0 to 3.

The output converted to a control level from the threshold matrix ROM 22 is serially supplied in 6 bits, and is sent to the buffer y.
25 a, 25 b (RAM B 1,
A M B 2 ). Buffer 25a, 2
The data in 5b represents the amount of energy supplied to each heating element of the thermal head 14. The buffers 25a and 25b each have a capacity of 6 bits x 2560 dots, and similarly to the buffers 21a and 21b, as shown in the timing diagram of FIG.
Write operations are performed alternately by the AMB2WR signal,
Read operations are performed alternately by the RAMB IRD and RAMB2RD signals.

FIG. 4 is a block diagram showing details of the head drive circuit 13 in FIG. 1, and FIG. 5 is a timing chart thereof. Here, it is assumed that the thermal head 14 is driven in two layers. Therefore, two systems of drive circuits are configured. The data converted into the amount of energization energy by the multi-level dithering circuit 12 is input to the shift register 30a. The output of the shift register 30a is the output of the shift register 30b.
will be forwarded to. The same clock signal is supplied to the shift registers 30a and 30b. shift register 3
The outputs of 0a and 30b are input in parallel to latch circuits 31a and 31b, respectively.

The same latch signal is supplied to the latch circuits 31a and 31b. Also, a latch circuit 31a.

31b has an enable signal ENI as shown in FIG.
, EN2 are supplied alternately. gate circuit 32a,
The output of 32b is a driver 33a.

The heat is supplied to the heating elements of each layer of the thermal head 14 via 33b. The head drive circuit 13 is, for example, disclosed in Japanese Patent Application Laid-Open No. 6-1-
227 (as described in Japanese Patent No. 174), the parallel signals may be generated manually. However, in that case, the output signals from the multilevel dithering circuit 12 must be made parallel.

In this embodiment, temperature data of the thermal head 14 is fed back to the head drive circuit 13. This is because the amount of ink transferred even with the same amount of energizing energy varies depending on the heat accumulated in the thermal head 14 itself and the salinity of the environment, so the energizing energy is appropriately controlled based on the detected temperature of the thermal head 14 itself. It's for a reason. By performing such control, as shown in Fig. 6,
Assuming that the energizing energy at normal m (Tn) is 100%, as the temperature increases, the energizing energy decreases, and as the temperature decreases, the energizing energy i increases, so that the current is always kept at a predetermined level regardless of the heat storage state of the thermal head. amount of ink is transferred. To detect the temperature of the thermal head 14, for example, as shown in FIG.
This can be done by supplying the signal to the head drive circuit 13 via the head drive circuit 13. Furthermore, the energizing energy of the thermal head 14 changes the pulse width of the enable signals ENI and EN2 supplied to the gate circuits 32a and 32b in FIG. 4 as shown in FIG.
The pulse widths shown in I and Bl are reduced to the pulse widths shown in A2.82, respectively, and the amplitude values of the output voltages of the drivers 33a and 33b are changed from the values shown in AI and A2 in FIG. 8 to A3. It can be reduced by reducing it to a value as shown in B3.

FIG. 9 shows a specific example of a multivalued threshold matrix used in this embodiment, in which a threshold matrix consisting of M-4×4 elements is provided in 12 layers indicated by P1 to PL2. in this case,
It is possible to realize an L-13 value dither method that reproduces N-84 gradation levels.

If a 64-level gradation image signal expressed as a numerical value of θ to 63 is input to this 12-layer threshold matrix pt-PL2 as shown in FIG. 10, this gradation image signal and threshold value matrix PL-PI3 A numerical value (1
~63) are compared and the magnitude relationship between them is changed.Threshold matrix number P i (i = 1.2...12)
is output as control level data.

The 12 types of pulses obtained by multiplying the output matrix number Pi (control level data) by the unit pulse width are supplied to the head drive circuit 13 as energization pulses to the thermal head 14, so that the energization pulse width or It becomes possible to control recording energy according to the comparison result between the gradation image signal and the threshold value matrix. For example, if a 14-level gradation image signal is input to the position of the 1st row and 2nd column in the threshold matrix, the 8th layer P8
Since there is a threshold corresponding to the 14th level at which the magnitude relationship changes, if the unit pulse width is 80 μSeC, an energizing pulse with a pulse width of 8×80-640 μSeC is applied to the recording image point. FIG. 10 shows the pulse width (×80 μsec) output corresponding to each element position when a gradation image signal is manually applied to all of the threshold matrix.

The threshold value matrix shown in FIG.
Similar to No. 48, the increment of the elements of the threshold matrix is determined at each position (
In addition to the fact that there is not very strong regularity or uniformity between the matrix position) and each layer, and there are asymmetrical differences, the following points are major characteristics. In other words, this threshold matrix is set between the maximum gradation level (63rd level in this example) and the next gradation level (62nd level in this example).
The sum of the control levels within each matrix takes a maximum value, and a plurality of elements (3 to 4) are included that have the maximum gradation level (63 levels) throughout all the layers of the threshold matrix. As a result, as shown in FIG. 11, the total recording energy in the matrix jumps discontinuously between the maximum gradation level and the previous gradation level.

Design procedure for a threshold matrix with such characteristics (1)
~(4) is shown below. First, perform a recording experiment to search for and select a multi-valued matrix pattern with a lattice-like dot pattern that allows continuous gradation recording with high image quality (1), and make a list of the obtained multi-valued matrix patterns (2). . That is, the multivalued matrix pattern is arranged in the order of gradation levels 0 to N, and the control levels 0 to L-1 for each pixel are written.

FIG. 12 shows an example of L multi-value patterns obtained in this way, with matrix size M-3X3 and number of control levels L.
-3 (0 to 3), and the case where the number of gradation levels is N-13 is shown. O■■ indicates the control level of O to 3, respectively. As shown in the figure, between the gradation levels L1 and L2, the total control level in the matrix changes continuously by "1", but the gradation level immediately before the maximum gradation level L12 changes continuously by "1". Between the tone level Lll and the maximum tone level LL2, the total control level in the matrix is 8 m for the former and 12 for the latter, which is an increase of 4, that is, there is a discontinuous jump. ing. Here, in the present invention, the first
As shown in Figure 2, the maximum gradation level L12
corresponds to a minute shape portion such as an edge portion, and corresponds to a gradation level Lll one level below the solid density.

Next, the multivalued matrix pattern of FIG. 12 is converted into a multilayered threshold matrix (3). That is, a matrix of N layers corresponding to each control level is set, and each element of the matrix is filled with the number of gradation levels at which the density value of the gradation image signal passes through the corresponding control level. Create a threshold matrix. For example, if the number of gradation levels at which the density value of the gradation image signal passes through the control level ■ is entered in the elements of the matrix, the result will be as shown in FIG. 13(a), which becomes the first layer threshold matrix. . Similarly, if the number of gradation levels at which the density value of the gradation image signal passes through the control level ■ is entered in the matrix element, the second layer threshold matrix is obtained as shown in FIG. 13(b). .

The multi-value threshold matrix thus obtained is applied to the multi-value dithering circuit 12 in FIG. 1 as a systematic dither method (4
). In that case, as described above, it is desirable to combine the high-frequency emphasis filter 11 as a preprocessing means.

The most significant feature of the multi-value threshold matrix according to the present invention described above is that in FIG. 11, which shows the gradation levels with respect to (the total of) recording energy in the matrix, the control level with respect to recording energy A is not uniform for all dots. By having an energy gap between an unsaturated state (first point Xl: corresponds to solid density) and a saturated state (second point X2), which is uniform for all dots with respect to recording energy B, be. That is, the first point X1 is at a gradation level lower than the maximum gradation level of the multivalue, and the second point X2 is at the maximum gradation level, so that there is a discontinuous gap in recording energy between them. It may be said that it is set. Gradation expression is performed at all levels below the first point X1, and edge emphasis is performed only at the second point X2. Note that the recording energy B used for edge enhancement may have some variation.

In short, the sum of the control levels within each matrix takes a maximum value between the maximum gradation level and the gradation level immediately before it, and the total recording energy within the matrix takes a maximum value between the two gradation levels. Points that jump discontinuously do not contribute to expressing gradation as a whole pixel, but the combination of the high-frequency emphasis filter 11 and the multi-level dithering circuit 12 reduces image edges and fine parts. It has a great effect when expressing. That is, for example, a solid density portion of a normal gradation image will be at a level below the maximum gradation level, but a portion such as an edge portion with many high frequency components will reach the maximum gradation level at once due to high frequency enhancement.

As a result, the multilevel dither method, which was originally devised with an emphasis on the expressive power of smooth gradations, can be given expressive power and sharpness for fine shape parts. This point will be explained in more detail below.

To record a gradation image, high frequency enhancement as a correction process is necessary for the following two reasons (A) and (B).

(A) Correction of blur caused by the heat storage effect and heat diffusion specific to thermal recording.

(B) Suppression of noise at image edges that tends to occur during dithering processing, that is, periodic texture and scattering of pixels.

The halftone reproduction method according to the present invention utilizes the principle of stabilizing the shape of the ink transferred to recording paper in a molten state by controlling the temperature distribution across a plurality of dots, and controlling fluctuations in the ink adhesion rate.

From this, the experimentally optimized matrix size is
This is thought to be approximately the same size as the area in which the heat emitted from a certain dot can affect its surroundings. Actually, 10
Since a matrix of ~250 μm 0 is used, it is estimated that the thermal blur is about 4 to 5 lines/angle. In this embodiment, a general high-frequency emphasis filter 1 using a Laplacian filter, for example, is used to deal with the blur caused by the accumulation effect in (A).
1 is used for correction.

Further, as described above, the method of the present invention in which several matrix elements whose recording energy is left at an unsaturated low level or zero are filled all at once at the maximum gradation level is a method of the present invention in which the high-frequency emphasis filter 11 is used to By overflowing the data, it is not only possible to express the inconspicuous fine edge shape of the grid-like texture, but it also has the effect of suppressing noise caused by scattering of pixels.

In order to confirm the effect of this high-frequency emphasis filter 11, the following experiment was conducted. In the experiment, an image synthesized from a part of the Institute of Image Electronics Engineers facsimile test chart No. 1 ((C) 1972) was used. A filter having a size of 4×4, which is the same as the size of the threshold matrix shown in FIG. 14, was used to prevent moire (aliasing noise) from occurring due to dithering. The filter shown in FIG.
--1 is a low-pass filter, and K>0 is a Laplacian filter that becomes a high-frequency emphasis filter. For image quality evaluation, Fig. 21(a) shows the grid-like texture density pattern method, Fig. 21(b) shows the multilevel dither method without filter processing, and Fig. 21(e) and (d) show the high-frequency emphasis filter 11. 3 shows samples recorded by the multilevel dither method according to this embodiment, which were preprocessed in the following manner. As a result of a subjective evaluation comparing these, the image quality of the gradation image part was the best in FIG. 21(e) with a moderate amount of high-frequency enhancement. When high frequency emphasis is applied too much, Figure 21(d)
The noise contained in the human image is emphasized, and the gradation becomes worse, resulting in a so-called "hard image quality."On the other hand, the image quality of text improves as the amount of high-frequency emphasis increases.
6.5 point characters (kanji) are reproduced almost as good as the original. Therefore, the larger the amount of high frequency enhancement, the smaller the dithering noise.

From the above recording results, the multi-value dither method according to this embodiment has better image quality than the multi-value density pattern method, which has better gradation than the binary dither method but lacks sharpness. It was confirmed that this multilevel dither method is particularly effective in using a high-frequency emphasis filter to correct blur caused by heat diffusion. It was also confirmed that the optimal amount of high-frequency enhancement differs between gradation images and character images.

The basic design guidelines for the multilevel dither method in this embodiment described above are as follows.

■Use a multilayer threshold matrix that concentrates heat in a grid pattern.

■Determine several high-quality multi-value patterns that should be passed during increase/decrease as constraint conditions.

■Elements other than those that make up the grid dot pattern are
It is not necessarily necessary to use it for gradation expression.

■When expressing the edges (fine parts) of an image,
Use all elements of the threshold matrix.

In this case, it is necessary to determine the pattern as a constraint condition, the size of the threshold value matrix, and the number of multi-value level control, and also to achieve both of the following. These will be described below.

Regarding pattern determination, based on the empirical facts obtained by the inventors through experiments in which many density patterns were recorded, in the case of a 4×4 matrix size, the three types of lattice-like dot patterns shown in FIG. 15 were set as constraint conditions. I found it useful to use it as These three types of patterns form pixel shapes with little variation in area within each density range of low density, medium density, and high density, and exhibit stable transfer characteristics. The cross-shaped pattern used at low density as shown in Figure 15(a) is such that in the highlight area where only the central dot is desired to be transferred, the surrounding four dots act as a thermal bias to form a stable thermal state. do. In the area where the density increases and transitions to the next step pattern, the lattice pattern shown in FIG. 15(b), the transfer ink is characterized by good connectivity in terms of texture and energy distribution. ing. In addition, the pattern shown in FIG. 15(C) used at high density and the pattern shown in FIG. 15(d) which is spatially shifted from the pattern shown in FIG. 15(d) have the noise characteristics shown in FIG. It mimics the heat distribution in the specific low-noise region (indicated by the black arrow) that exists between after the collapse is completed and before the collapse begins, and has the effect of suppressing noise at high concentrations.

Determination of matrix size The noise of the lattice dot pattern was measured using the matrix size and the resolution of the thermal head 14 as parameters. As shown in FIG. 16, the noise decreases at intermediate density or higher, but the level It shows a characteristic that is inversely proportional to the pixel area. That is, the thermal head 14
In a pixel area of 188 .mu.m', which corresponds to a 3.times.3 matrix size at a resolution of 16 dots/square, the area ratio fluctuation is about 5%. On the other hand, the resolution is 16 dots/
250 μm, which corresponds to a 4 x 4 matrix size in m+s and a 3 x 3 matrix size with a resolution of 12 dots/step.
Regarding the pixel area of the mouth, the area ratio fluctuation was 3%, about half the level. According to the results of subjective evaluation based on observation of actually recorded samples, it was found that the former had more "roughness" than the latter, and the concentration change due to heat accumulation was also significant.

Therefore, the minimum matrix size that shows good gradation is 4 x 4.12 dots/mm at a resolution of 16 dots/mm and 3 x 3 at a resolution of 16 dots/mm (this is equivalent to 2508m0100 lines/inch as a pixel area for halftones). )I think that the. However, in the experiment, wax-based ink with a thickness of 2.6 μm and single-sided art paper with a smoothness of 300 seconds were used. Synthetic paper has better ink adhesion than plain paper and has less transfer noise.
There is also a possibility that the matrix size can be further reduced.

Determination of the number of multi-value level controls In general, in order to reproduce smooth gradations by recording using the normal binary dither method, it is insufficient if the number of elements constituting the threshold matrix is 32 or less. In addition, in order to reproduce smooth gradations using the 4x4 size multilevel dither method, at least 6
~Multi-value conversion with 7 or more values (threshold matrix with 5 to 6 layers or more)
is necessary.

Therefore, in thermal melt transfer recording using a thermal head with a resolution of 16 dots/dot, 5 to 6
It is desirable to use a threshold matrix of more than one layer to control the number of gradations of 64 levels or more.

This point will be explained in more detail.

As one of the most severe tests to examine the gradation recording performance of a printer, we output a smoothly changing halftone image using image data that does not contain any noise, such as computer graphics, and create a false contour (also known as a false contour). There is a method that uses as an evaluation criterion the absence of step-like density noise called .

FIG. 17 shows the results of evaluating the multilevel dither method of this embodiment using this method. The vertical axis represents the percentage (%) of those who cannot see the pseudo contour among the five subjects, and the horizontal axis represents the number of multilevel control levels. From this result, in order to reproduce smooth gradations without visible false contours with thermal melt transfer recording, the number of multilevel control levels of 7 or more is required.
In other words, it can be seen that a threshold matrix of six or more layers is required.

Further, FIG. 24 shows samples in which photographic images of pots were recorded with various multi-value control levels, and the false contours are almost completely invisible at L-7 and above.

In addition, in thermal melt transfer recording, the number of gradation levels is 32.
It is clear from experiments that more than gradations are required.

Therefore, the number of gradation levels exceeding 64 gradations is wasted in terms of image quality and only unnecessarily increases the circuit scale.
The 12-value multi-level control described in No. 02294, which has been experimentally confirmed to have the ability to express 111 gradation levels, is unnecessary, and the number of multi-value level controls may be 11 or less. .

According to this embodiment described above, in thermal melt transfer recording, it is possible to provide a halftone reproduction method using a multi-value dither method that can obtain recording reproduction of good image quality for both gradation images and character images. This multi-value dither method constructs a multi-value threshold matrix having the following characteristics.

■Grid pattern (180 to 250 in both main and sub-scan directions)
A multi-value threshold matrix of 5 to 6 layers or more is used to control the implantation energy so as to form a heat concentration with a micrometer pitch.

■The maximum density should be reached at one level before the maximum gradation control level, and the maximum density should be reached before the energy gap created in relation to ■ is not filled (i.e., unsaturated). Set the injection power.

■At the maximum gradation control level, the energy gap is filled in, and between the maximum gradation control level and the previous gradation control level, the pixel internal conflict energy (total recording energy control level) is ).

By combining this threshold matrix and a high-frequency emphasis filter, it is easy to correct blur caused by heat diffusion and suppress pixel scattering at edges, making it possible to reproduce images with sharpness.

In addition, as a result of recording using a boat-sized ink ribbon and a facsimile thermal head, the "roughness" is less than half that of conventional recording, the fineness is enough to reproduce 6.5 point characters, and 64 It was confirmed that smoothness exceeding gradation was obtained and good gradation images were reproduced.

[Effects of the Invention] As described above, according to the present invention, it is possible to express smooth gradations and reproduce intermediate tones with excellent detail reproducibility and sharpness. Moreover, smooth gradation expression without false contours can be achieved with a relatively small-scale circuit configuration without unnecessarily increasing the number of gradations.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of a thermal transfer printer to which the halftone reproduction method of the present invention is applied, and the second diagram is a block diagram showing details of the multilevel dithering circuit in the same printer.
Fig. 3 is a timing diagram showing the operation of Fig. 2, Fig. 4 is a block diagram showing details of the head drive circuit in the same printer, Fig. 5 is a timing diagram showing the operation of Fig. 4, and Fig. 6 is a thermal Figure 7 is a block diagram showing the temperature detection means of the recording head; Figure 8 is a waveform diagram showing an example of the temperature compensation method for the recording head; FIG. 9 is a diagram showing an example of the threshold value matrix in this embodiment, and FIG. 10 is a diagram showing how the threshold value matrix in FIG. 9 is used to convert a gradation image signal into pulse width control data for the energizing pulse to the thermal head. Figure, 11th
The figure shows the non-linearity of the recording energy control amount between each gradation level in the same embodiment, Figures 12 and 13 are diagrams for explaining the design procedure of the multilevel dither method in the present invention, and Figure 14. 1 is a diagram of the Laplacian filter used as the high-frequency emphasis filter in FIG. 1, FIG. 15 is a diagram showing three types of lattice-like dot patterns that are the constraint conditions in this example, and FIG. 16 is a diagram showing the lattice-like dot patterns of FIG. 15. Diagram showing the area ratio fluctuation characteristics with respect to the average recording density of the pattern, 1st
Figure 7 is a diagram showing the relationship between the number of multilevel control levels and the percentage of people who can see pseudo gradation in the same example, Figure 18 is a diagram showing changes in recording density characteristics depending on the arrangement of recording dots, and Figure 19 is FIG. 20 is a diagram showing the area ratio fluctuation characteristics with respect to the average recording density of a concentrated dot pattern. FIG. 20 is a diagram showing the area ratio fluctuation characteristics with respect to the average recording density of a lattice dot pattern. FIG. 21 is a diagram showing the area ratio fluctuation characteristics with respect to the average recording density of a lattice dot pattern. 22 is a diagram conceptually showing the types of recording noise shown in FIG. 21, and FIG. 23 is a diagram showing the measurement results for each recording noise shown in FIG.
Figures 24 and 24 are photographs showing various recording samples for detailed explanation of the present invention, Figures 25, 26, and 2.
FIG. 7 is a microscopic photograph showing the particle structure of transfer ink for explaining the problems of the prior art. 10... Gradation image signal source, 11... High frequency emphasis filter, 12... Multi-value dithering circuit, 13... Head drive circuit, 14... Thermal head, 15... Recording paper , 16... Ink ribbon, 17... Platen roller, 18... Recorded image, one... Ink, 20...
timing controller.

Claims (5)

[Claims] (1) In a halftone reproduction method that expresses a gradation level by converting a gradation image signal into a predetermined control level using a threshold value matrix and controlling recording dots using this control level, the gradation level is between a first point located at a gradation level lower than the maximum gradation level and where the control level is not uniform for all the recording dots, and a second point where the control level is substantially uniform for all the recording dots; A halftone reproduction method characterized by having a gap in which the total value of control levels within the threshold matrix at the first point and the second point is discontinuous. (2) Using a multilayer threshold matrix consisting of M elements, convert the gradation image signal to a control level of L value from 0 to L-1 (L is an integer of L≧3), and use this control level to In a halftone reproduction method that expresses N gradation levels [M<N≦M(L-1)] by controlling recording dots, the threshold matrix is configured to control the control level for the gradation image signal. Halftone reproduction characterized in that the sum of values within a matrix size is determined to take a maximum value between a maximum gradation level and a gradation level one level before the maximum gradation level. method. (3) Using a multilayer threshold matrix consisting of M elements, convert the gradation image signal into a control level with an L value from 0 to L-1 (L is an integer of L≧3), and use this control level to In a halftone reproduction method that expresses N gradation levels [M<N≦M(L-1)] by controlling recording dots, the value of L in the threshold matrix is 7≦L≦11. A halftone reproduction method characterized by: (4) The halftone reproduction method according to claim 1, 2 or 3, wherein the gradation image signal is subjected to high frequency enhancement processing. (5) The size of the threshold matrix is 180 to 250.
Claim 1, 2 or 3 characterized in that μm^□
The halftone reproduction method described in .