JP4192351B2 - Image processing apparatus and image processing method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an image processing apparatus and an image processing method, and more particularly to an image processing apparatus and an image processing method capable of creating an image with reduced gradation by using a threshold value.
[0002]
[Prior art]
Digital image handling is the mainstream of current image processing. When displaying or outputting a digital image, it is often necessary to express the gradation of the image with a smaller gradation level due to restrictions due to the characteristics of the output device. From the beginning, various digital halftoning image processing techniques have been studied, such as a binarization processing method that reproduces gradation using only white and black dots as pseudo halftone processing.
[0003]
The systematic dither method, error diffusion method, and various methods of those systems still in use have been developed and improved since the 1960s. In recent years, with the development of calculation processing hardware, a method for directly searching for an optimal pixel arrangement has been developed, such as a cost minimization method.
[0004]
Each of these halftoning methods has advantages and disadvantages depending on the purpose of use, and various issues and countermeasures have been studied. For example, the organized dither method is easy to process and easy to use, but the reproduced image quality is not good. Although the error diffusion method requires more computation than the dither method, the image quality is excellent.
[0005]
Various optimization methods such as neural networks, genetic algorithms, and simulated annealing are used in direct optimization methods such as cost minimization methods. By adopting these methods, it is easy to incorporate visual models and output device models into the process, and the degree of freedom of processing is expanded. However, there is a problem that the amount of calculation becomes enormous because it repeatedly calculates and searches for an optimal state.
[0006]
These challenges change as technology advances. The problem of enormous amount of calculation in the case of using the method for performing optimal search directly may be solved by the advancement of hardware that defines the calculation processing speed. However, simpler calculation processing is desired from the viewpoint of widespread use of simple and high-quality output devices.
[0007]
Further, as a common problem, there is a problem of a trade-off between resolution and gradation. This may also be solved by increasing the output gradation level of the output device itself and improving the resolution characteristics. However, for example, an increase in the chance that characters are processed as an image may be considered, and it is desired to perform these processes as simply as possible.
[0008]
Conventionally, for example, a method of determining an image region requiring gradation and an image region requiring resolution and changing a processing method in each region according to the determination result, a method of using a plurality of processing methods in combination, etc. It has been studied as a means for improving image processing. However, in order to execute them, it is necessary to develop and add a new process such as area discrimination, which is not a simple method. From the standpoint of balance with hardware (output device), it would be desirable to achieve the object of maintaining good resolution and gradation by processing similar to the error diffusion method.
[0009]
FIG. 66 is a block diagram showing a configuration of an image processing apparatus that executes a conventional error diffusion method.
[0010]
Referring to the figure, the image processing apparatus includes an input unit 501 that inputs a pixel value of one pixel of a multi-valued image, a subtracter 503 that subtracts an error diffused from the input pixel value, and a subtracter 503. Are output as corrected pixel values, a threshold processing unit 507 that performs threshold processing on the output of the output unit 505 to form binary data, and a threshold processing unit 507 An output unit 509 that outputs the output as pixel data, a subtraction unit 511 that subtracts the output of the output unit 505 from the output of the threshold processing unit 507, and a pixel whose output is output from the subtraction unit 511 And an error memory 513 for diffusing the pixels around the pixel of interest.
[0011]
An image created by the error diffusion method has a unique texture. However, as the blue noise characteristic is pointed out, the texture is not visually noticeable. In the dither method, a method of setting a dither pattern so as to obtain the blue noise characteristic more easily has been studied. However, since the error diffusion method adaptively generates a dot pattern for the input image, the characteristics of the input image can be reflected more than the dither method.
[0012]
In this respect, the error diffusion method outperforms the dither method in terms of image quality, but there is also noise specific to the error diffusion method. This is because the texture changes in a gentle gradation change area, and there appears to be a boundary line even in an area without a boundary (texture shift), or white or black in a black or white gradation area. This is a phenomenon that makes it easier to arrange the dots in a line.
[0013]
In order to prevent these phenomena, various improved methods such as modulation of error diffusion weighting factors and threshold values have been developed. In addition, regarding the resolution, the inherent edge enhancement characteristic has been pointed out, but it cannot be said to be sufficient.
[0014]
Further, the error diffusion method functions so as to reproduce the pixel values of the input image on average based on the algorithm. That is, it functions to reproduce the local zeroth order component of the image. On the other hand, the error diffusion method has been improved to emphasize the first and higher order components.
[0015]
[Problems to be solved by the invention]
An object of the present invention is to provide an image processing apparatus and an image processing method capable of solving the above-described drawbacks of the image processing method and improving the image quality.
[0016]
[Means for Solving the Problems]
In order to achieve the above object, according to one aspect of the present invention, an image processing apparatus transmits a first image signal indicating a density level of each pixel with a predetermined number of gradations to a number of gradations smaller than the predetermined number of gradations. An image processing apparatus for converting to a second image signal, wherein the first image signal of each pixel is sequentially input, the density level of each pixel is compared with a predetermined threshold value, and converted to a second image signal And a feedback circuit for correcting the predetermined threshold value used in subsequent pixel conversion based on the signal level of the second image signal output from the converter and the threshold value.
[0017]
Preferably, the feedback circuit includes control means for controlling a feedback value in the feedback circuit.
[0018]
Preferably, the control means includes a feedback coefficient setting unit for setting a feedback coefficient.
[0019]
Preferably, the feedback coefficient setting unit can change the feedback coefficient.
[0020]
Preferably, the feedback coefficient setting unit sets a feedback coefficient that changes according to the density level of each pixel converted by the converter.
[0021]
Preferably, the feedback coefficient setting unit is configured to calculate a feedback coefficient based on a predetermined relationship between each density level of each pixel converted by the converter and the feedback coefficient, and a predetermined relationship used in the calculation unit. And means for changing.
[0022]
Preferably, the feedback circuit includes a correction value memory that distributes a feedback value from the feedback circuit to a plurality of peripheral pixels to be subsequently converted according to a weight set for each peripheral pixel.
[0023]
Preferably, the weight of the correction value memory can be changed.
Preferably, the image processing apparatus further includes a threshold value generation unit that generates a value that changes with the conversion of each pixel as a predetermined threshold value.
[0024]
Preferably, the threshold value generator changes the predetermined threshold value according to the position of the pixel to be converted.
[0025]
Preferably, the threshold value generator changes the predetermined threshold value according to the density level of the pixel to be converted.
[0026]
Preferably, the image processing apparatus further includes a multiplier that is provided in front of the converter and multiplies the density level of each pixel converted by the converter by a predetermined coefficient.
[0027]
Preferably, the predetermined coefficient can be changed.
Preferably, the image processing apparatus changes a gradation characteristic indicating a relationship between the level of the first image signal and the level of the second image signal after conversion, and changes at least one of a predetermined threshold value and a feedback coefficient. To make adjustments.
[0028]
Preferably, the image processing apparatus fixes the gradation characteristics with respect to the maximum level and the minimum level of the first image signal, and determines an intermediate level gradation characteristic between the maximum level and the minimum level as a predetermined threshold value and a feedback coefficient. Adjust by changing at least one of
[0029]
Preferably, the image processing apparatus superimposes a signal component unrelated to the first image signal to be converted on any signal in the image processing apparatus.
[0030]
Preferably, the superimposed signal component represents a periodic pattern.
Preferably, the periodic pattern is any one of a distributed dither pattern, a concentrated dither pattern, and a line pattern.
[0031]
Preferably, the signal component to be superimposed is random noise.
Preferably, the random noise is any one of white noise, blue noise, and pink noise.
[0032]
Preferably, the feedback circuit feeds back a difference between the inverted value of the signal level of the second image signal output from the converter and the threshold value.
[0033]
According to another aspect of the present invention, an image processing method converts a first image signal indicating a density level of each pixel with a predetermined number of gradations into a second image signal having a smaller number of gradations than the predetermined number of gradations. An image processing method for sequentially inputting a first image signal of each pixel, comparing a density level of each pixel with a predetermined threshold value, and converting it to a second image signal; And a feedback step of correcting a predetermined threshold value used in subsequent pixel conversion based on the signal level and threshold value of the output second image signal.
[0034]
Preferably, the feedback step includes a control step for controlling a feedback value in the feedback step.
[0035]
Preferably, the image processing method further includes a threshold generation step of generating a value that changes with the conversion of each pixel as a predetermined threshold.
[0036]
Preferably, the image processing method changes a gradation characteristic indicating a relationship between the level of the first image signal and the level of the converted second image signal, and changes at least one of the predetermined threshold value and the feedback coefficient. To adjust.
[0037]
Preferably, the image processing method fixes the gradation characteristics with respect to the maximum level and the minimum level of the first image signal, and determines an intermediate level gradation characteristic between the maximum level and the minimum level as a predetermined threshold and a feedback coefficient. Adjust by changing at least one of
[0038]
According to still another aspect of the present invention, the image processing apparatus converts the first image signal indicating the density level of each pixel with a predetermined number of gradations into a second image signal with a smaller number of gradations than the predetermined number of gradations. An image processing device for conversion, wherein the first image signal of each pixel is sequentially input and divided into sections corresponding to the number of gradation levels of the second image signal, and the section divided by the section circuit A normalization circuit that normalizes the input first image signal, and a comparator that sequentially inputs the first image signal normalized by the normalization circuit and compares the signal level of each pixel with a predetermined threshold value; Based on the comparison result output from the comparator and the predetermined threshold value, the feedback circuit corrects the predetermined threshold value used in the subsequent pixel conversion, and the comparison result output from the comparator is divided by the classification circuit. According to the classification , And a assignment circuit assigning the gray level of the second image signal to each pixel.
[0039]
Preferably, the feedback circuit includes control means for controlling a feedback value in the feedback circuit.
[0040]
Preferably, the feedback circuit includes a correction value memory that distributes a feedback value from the feedback circuit to a plurality of peripheral pixels to be subsequently converted according to a weight set for each peripheral pixel.
[0041]
Preferably, the image processing apparatus further includes a threshold value generation unit that generates a value that changes with the conversion of each pixel as a predetermined threshold value.
[0042]
DETAILED DESCRIPTION OF THE INVENTION
[First Embodiment]
FIG. 1 is a block diagram showing a configuration of an image creating apparatus according to the first embodiment of the present invention. The image binarization process executed by this apparatus is referred to as “threshold diffusion method”. In the drawing of this case, the part enclosed by a normal rectangle is a part that performs some arithmetic processing, and the part enclosed by a rounded rectangle is a part that does not perform arithmetic processing and only outputs a value. Is shown.
[0043]
Referring to the figure, an image creating apparatus includes an image (pixel value) input unit 101, a threshold processing unit 103, a binary image output unit 105, an inversion unit 113, and an initial threshold generation unit 107. , Subtractor 109, correction threshold value output unit 111, subtractor 115, coefficient multiplier 117, and correction value memory 119.
[0044]
One pixel value (0 to 1) of the multi-value image is input to the image input unit 101. For example, when handling a 256-level multi-valued image n (0 to 255), a value (n / 255) normalized to 0 to 1 is input to the image input unit 101. The threshold processing unit 103 compares the corrected threshold Th (x) output from the corrected threshold output unit 111 with the pixel value input to the image input unit 101. If pixel value ≧ correction threshold Th (x), the threshold processing unit 103 outputs “1”, and if pixel value <correction threshold Th (x), the threshold processing unit 103. Outputs “0”. As a result, the binary image output unit 105 outputs a binary image of “0” or “1”.
[0045]
The initial threshold value generator 107 outputs the initial threshold value Th (x) before correction. The initial threshold value Th (x) before correction may be a constant value or may be changed according to the pixel position so as to form a dither pattern.
[0046]
The subtractor 109 reads the correction value stored in the correction value memory 119 corresponding to the pixel to be processed (target pixel), and subtracts the correction value from the initial threshold value Th (x). The result is set as a correction threshold Th (x).
[0047]
The inversion unit 113 inverts the output of the threshold processing unit 103. That is, the inversion unit 113 outputs “1” if the output of the threshold processing unit 103 is “0”, and “0” if the output is “1”.
[0048]
The subtracting unit 115 subtracts the correction threshold value Th (x) from the output of the inverting unit 113 and outputs the result. The coefficient multiplier 117 multiplies the output of the subtractor 115 by a feedback coefficient β set between 0 and 1 and outputs the result. Note that β = 0 means that threshold value diffusion is not performed.
[0049]
The correction value memory 119 is a memory for distributing the output result of the coefficient multiplication unit 117 to the correction value of the threshold value for the peripheral pixels of the pixel to be processed. Referring to FIG. 2, when a pixel to be processed is indicated by a white circle, the output result of the coefficient multiplier 117 is a ratio (weighting coefficient) of 1 to 32 with the threshold correction value for the surrounding pixels. It is sorted and stored.
[0050]
As shown in FIG. 2, the number of pixels to which the output result of the coefficient multiplier 117 is to be distributed is increased compared to the normal error diffusion method. This is to eliminate the monotony of the value. In other words, the error diffusion method performs diffusion processing on input values (pixel values) that change in various ways, whereas the threshold diffusion method performs diffusion processing on a threshold value that is constant or nearly constant. It is because there are many cases to do. When the initial threshold value before correction is not a constant value but takes various values, the number of pixels to be distributed may be reduced.
[0051]
In comparison with the block diagram of the image processing apparatus using the error diffusion method (FIG. 66), in the apparatus using the threshold diffusion method, the inversion processing is performed by the inversion unit 113 when the output is fed back. . This is a process for changing the output result to the output viewed from the threshold side because the output result is the output viewed from the input value side.
[0052]
Further, in the feedback to the threshold value, the coefficient multiplication unit 117 multiplies the feedback coefficient β, but this only functions to reproduce the threshold value on average unless the feedback coefficient β is multiplied as described below. In the threshold value diffusion method, this is a process for causing the input value to be reproduced on average.
[0053]
The output image when the feedback coefficient β is changed will be described below.
9 to 11 are diagrams showing image samples binarized by the threshold value diffusion method with the feedback coefficient β set to 1, 0, 0.5, respectively. The initial threshold value was fixed at 0.5, the median value. In the case of the feedback coefficient β = 1 shown in FIG. 9, the dots of white and black are distributed so as to have a half density almost uniformly over the entire surface of the image, and a threshold value, that is, a density value of 50%. Has been reproduced. As the fine structure, only the edge portion of the input image is emphasized, and the whole is flat and only the contour is reproduced. That is, the local primary component of the input image is reproduced in the output image, but the local zero-order component of the input image is not reproduced, but the threshold value is reproduced instead. The image is composed of a local zeroth order component (low frequency component) and a local first order or higher order component (high frequency component).
[0054]
On the other hand, when the feedback coefficient β = 0 as shown in FIG. 10, this corresponds to the fact that threshold value diffusion is not performed. It becomes an image just made into.
[0055]
As shown in FIG. 11, when the feedback coefficient β is set to 0.5, which is an intermediate value between 1 and 0, an intermediate image between the image of FIG. 9 and the image of FIG. As can be seen from this sample image, when the feedback coefficient β is set to 0.5, the local zero-order component of the input image is reproduced instead of the threshold value. Of course, the local primary or higher order components of the input image are also reproduced.
[0056]
A mechanism in which an input value is reflected as a local zeroth-order component in the threshold value diffusion method will be described with reference to FIGS. 15 to 18 showing a relationship among an input value, an initial threshold value, and an output value. For reference, FIGS. 12 to 14 show the relationship among the input value, threshold value, and output value when the error diffusion method is used.
[0057]
In each figure, the origin is O, and the input value Im and the output value Out with respect to the initial threshold value Th in a three-dimensional space are displayed by hatched surfaces. Of course, since the original output can take only binary values of max (“1”) and min (“0”), the output value Out indicates an average reproduction value, that is, a local dot density.
[0058]
Further, the output value Out with respect to the input value Im when the initial threshold value Th is constant at the median value is indicated by a line on each three-dimensional space. FIG. 18 shows the relationship between the input value Im and the output value Out on a two-dimensional graph when the initial threshold value Th is constant at the median value.
[0059]
The feedback coefficient β = 0 shown in FIG. 15 indicates that threshold diffusion is not performed. In the case of the feedback coefficient β = 1 shown in FIG. 16, the output Out depends only on the initial threshold Th. That is, if the initial threshold value Th is constant, the output value Out is also constant. Therefore, in the threshold value diffusion method, it is important to set the feedback coefficient β between 0 and 1.
[0060]
As the feedback coefficient β changes from 0 to 1, the function surface of the output value Out inclines, and when the feedback coefficient β = 0.5, the input value Im and the initial threshold value Th as shown in FIG. It has an inclination with respect to both axial directions. That is, in this case, the output value Out depends on both the input value Im and the initial threshold value Th. Therefore, by setting β to 0.5, the local structure of the input image is further reflected while reflecting the first-order or higher-order components of the input image as the fine structure by the threshold diffusion algorithm. The next component can also be reproduced.
[0061]
As described above, it is important to set the feedback coefficient β in order to make the input value Im reproducible on average. However, as can be seen from the figure, the feedback coefficient β is usually around 0.5. It means that. However, this varies with conditions. That is, the feedback coefficient β = 0.5 is appropriate only when the initial threshold value is constant or close to the median as shown in the example, and when the initial threshold value is different. The value of the optimum feedback coefficient β varies.
[0062]
For example, as shown in FIG. 19, the initial threshold value Th (x) may be modulated by other factors such as an input value. In this case, the feedback coefficient β varies depending on the degree of dependence on the input value. As inferred from FIGS. 15 to 17, as the degree of negative dependence on the input value Im increases, the output function surface rises and the optimum feedback coefficient β approaches zero. Further, as the degree of positive dependence on the input value increases, the output function surface goes to sleep and the optimum feedback coefficient β approaches 1. Accordingly, the initial threshold value may be anything, but it is necessary to change the setting of the feedback coefficient β accordingly.
[0063]
In this way, by setting the feedback coefficient β so as to take advantage of the characteristics of the output function surface, the threshold diffusion method preferentially reproduces the first-order or higher-order component of the input value, that is, the high-frequency component. It is also possible to reproduce local zero-order components, that is, low-frequency components.
[0064]
12 to 14 show the relationship between the three in the error diffusion method, and β = 0 in FIG. 12 shows a case where error diffusion is not performed. That is, in FIG. 12, binarization is performed using a simple threshold value. Β = 1 in FIG. 13 corresponds to performing a normal error diffusion method. The output value Out reflects the input value Im. In the case of the error diffusion method, even if β is between 0 and 1, the output function reflects the input value when the input value is close to the threshold value, and is simple when the input value is close to max and min. It only has binarization and does not have much meaning. That is, in the error diffusion method, even if β is between 0 and 1, it is merely insufficient error diffusion.
[0065]
The operation and effect of the threshold value diffusion method are shown below by comparison with the error diffusion method. As can be seen by comparing the image processing apparatus using the error diffusion method shown in FIG. 66 with the apparatus using the threshold diffusion method shown in FIG. 1, the error diffusion method uses the input pixel value and output. However, the threshold diffusion method is greatly different in that the difference between the threshold value and the output is fed back to the threshold value. In other words, the object for calculating the difference from the output and the object for feeding back the difference are input values (input pixel values) in the error diffusion method, whereas they are threshold values in the threshold diffusion method. Yes. In the threshold diffusion method, the same texture (blue noise characteristic) as that of the error diffusion method can be obtained by the feedback algorithm.
[0066]
FIG. 3 shows a halftone processing result using a systematic dither method using a 4 × 4 pixel fat-type pattern, FIG. 4 shows a halftone processing result using an error diffusion method, and FIG. 5 shows an initial threshold value. The halftone processing result by the threshold value diffusion method (feedback coefficient β = 0.5) when a constant value is used as Th (x) is shown.
[0067]
When the systematic dither method is employed, the gradation and resolution are the worst. When the error diffusion method is used, both gradation and resolution are improved as compared with the dither method. However, when the threshold value diffusion method is used, the error diffusion method exceeds the error diffusion method in particular. With regard to the gradation and texture, the threshold value diffusion method can obtain almost the same result as the error diffusion method. In particular, in the threshold value diffusion method, the occurrence of a texture shift is observed as in the error diffusion method. However, the defect that the dots are arranged in a line in a region close to black or white that is generated in the error diffusion method does not occur in the threshold diffusion method.
[0068]
FIG. 6 shows a halftone processing result by the threshold value diffusion method (feedback coefficient β = 0.5) using a fat-type dither pattern of 4 × 4 pixels as the initial threshold value Th (x). Shows a halftone processing result by an error diffusion method in which a fat-type pattern of 4 × 4 pixels is modified as a threshold value, and FIG. 8 shows a halftone processing result using an error diffusion method with edge enhancement. .
[0069]
By using the dither pattern as the initial threshold Th (x) in the threshold diffusion method, the texture shift can be improved. Similar improvements have been made in the error diffusion method. However, this improvement does not adversely affect other characteristics such as resolution in the threshold diffusion method, and the threshold diffusion method is still superior to the error diffusion method. By using the error diffusion method with edge enhancement, the resolution is still improved. However, even in this case, since it is restricted by the essential function of the error diffusion method of reproducing the average input value, sufficient reproducibility cannot be obtained particularly with a thin line with low contrast.
[0070]
The characteristics of the output image quality of the binarization processing by the threshold value diffusion method are as already described. The process for producing such image quality will be described in comparison with the error diffusion method.
[0071]
The error diffusion method uses the input value to the maximum, that is, it makes an image reflecting the input value by a method of adaptively determining the dot arrangement using the process of feeding back the error between output and input to the input. On the other hand, in the threshold diffusion method, the input value is not directly involved in the feedback. In the threshold diffusion method, the input value is only used for comparison to determine the output value. That is, the output on (“1”) or off (“0”) is fed back, but the input value itself does not enter the feedback process.
[0072]
However, in the threshold value diffusion method, as described above, by setting the feedback coefficient β to an appropriate value, the input value can be reflected in the output result, that is, the gradation of the input image can be reproduced. is there.
[0073]
Further, the error diffusion method essentially functions to reproduce the input value locally on the essence, but the threshold diffusion method does not. This also seems to be disadvantageous for reproduction of the input image, but conversely it is difficult to be constrained by the input value.
[0074]
A case will be described in which the function of the error diffusion method of average reproduction of input values is constrained. For example, assume a case where there is a gray line, ie, a low contrast thin line, against a white background. In this case, since it is gray, the dot density is determined according to the degree. Assuming 50% gray, half the dots of white and black must exist on average. Then, since it is a thin line, if half of the dots constituting the line are made white, the solid line may become a dotted line in an extreme case. In other words, we want to increase the number of black dots and bear white dots around them, but since the surroundings are originally white, it is impossible to distribute more white dots around them. That is, since the density of the thin gray line is reproduced on average, the characteristic as a line may be destroyed.
[0075]
In such a case, it is better to reproduce the line characteristic, that is, the low contrast edge characteristic, even if the gray level of the line is ignored to some extent. The function of the error diffusion method gives priority to the reproduction of the local zero-order component (low frequency component). However, depending on the nature of the image, it may be desirable to prioritize local first-order or higher-order components (high-frequency components). For example, in a portion where there are relatively small irregularities locally, It is desirable to give priority to the reproduction of the first or higher order component, that is, the unevenness, over the reproduction of the average level. Of course, the zero-order component may be reproduced in the gentle gradation portion.
[0076]
On the other hand, as described above, in the threshold value diffusion method, emphasis is placed on the reproduction of a local first-order or higher-order component. Therefore, even in the case where a thin gray line exists on the white background described above, which is limited by the function of the error diffusion method, emphasis is placed on the reproduction of a local first-order or higher-order component. In the threshold diffusion method, good results can be obtained because the edge characteristics of the line are reproduced.
[0077]
FIG. 23 is a diagram for explaining a specific calculation example by the threshold value diffusion method. Referring to the figure, it is shown how numerical values in the apparatus change by subtracting units 109 and 115, threshold processing unit 103, coefficient multiplying unit 117, and inverting unit 113.
[0078]
Specifically, for the initial threshold value (threshold value th), the input value (input), and the coefficient β, the output of the subtraction unit 109 (correction threshold value Mod th) and the threshold value processing unit 103 , The output of the inversion unit 113 (inversion output), the output of the subtraction unit 115 (FB), and the output of the coefficient multiplication unit 117 (FB * β) are shown to change. Here, Shigeaki Kakutani, “High-definition technology of error diffusion method”, Journal of the Japan Photography Society, Vol. In accordance with 353 to 356, the threshold diffusion run-up section is estimated.
[0079]
As shown in the calculation example of FIG. 23, when one gradation is 1/255 (= 0.00392) with respect to the maximum output value, the threshold value is one-dimensionally diffused in one direction. The correction threshold (Mod th) is halved for each step. It is only the eighth step that the correction threshold reaches 1/255 with respect to the maximum value. This is the preferred run-up section ((255) ^1/2 / 2≈8) is almost equal. In the case of the error diffusion method, after 128 steps, the threshold value diffusion method is very advantageous compared to the error diffusion method. However, since the threshold value is actually two-dimensionally diffused, the run-up period is longer than that in the case of one-dimensional diffusion in one direction.
[0080]
Next, specific results of image processing will be described.
FIG. 24 is a diagram illustrating an image to be processed. This image has 256 gradations from 0 to 255. The black part in the image is a part showing a density of 251/255, and the central square part is a part showing a density of 4/255. In addition, a line having a density of 230/255 is drawn at a portion indicated by two arrows.
[0081]
With the image creation apparatus using the threshold value diffusion method of the first embodiment described above, the image binarization process shown in FIG. 24 was performed. The initial threshold value was 0.5 and the coefficient β was 0.5. In addition, the matrix shown in FIG. 2 was used for threshold diffusion.
[0082]
FIG. 25 is a diagram showing the image processing result. As shown in the figure, when the threshold value diffusion method is employed, dot generation delay and tailing are small.
[0083]
FIG. 26 is a diagram showing a result of processing the image shown in FIG. 24 by the error diffusion method (threshold is 0.5, and the same matrix as FIG. 2 is used for error diffusion). As can be seen from this figure, when the error diffusion method is employed, dot generation delay and tailing become severe. Thus, it can be seen that the threshold diffusion method is superior to the error diffusion method.
[0084]
For reference, FIGS. 27 and 28 show the results of threshold value diffusion processing when the initial threshold value is 0.5 and the input value is 0.2. In FIG. 28, the flow of each numerical value in the four steps (1) to (4) is shown.
[0085]
[Modification]
Instead of the configuration shown in FIG. 1, the configuration shown in FIG. 19 may be adopted for the threshold value diffusion method. In the configuration of FIG. 19, the initial threshold value Th (x) is modulated by the input value. In this case, it is necessary to adjust the feedback coefficient β.
[0086]
Further, instead of the configuration shown in FIG. 1, the configuration shown in FIG. 20 may be adopted. This includes a multiplication unit 302 that multiplies an input value by a coefficient α. The degree of the binarization effect can be controlled by the coefficient α, and the degree of edge enhancement can be increased, for example. Even in this case, since the appropriate value of the feedback coefficient β varies depending on the coefficient α, it is necessary to adjust the feedback coefficient β.
[0087]
In addition, the apparatus configuration shown in FIG. 19 and the apparatus configuration shown in FIG. 20 are added, and the modulation by the input value of the initial threshold Th (x) and the process of multiplying the input value by the coefficient α are performed. It may be performed simultaneously.
[0088]
As described above, the threshold value diffusion method functions so as to preferentially reproduce local first-order or higher components of the image input value. However, it is nevertheless possible to reproduce the local zero-order component by setting parameters. The algorithm uses the same feedback as in the error diffusion method, and the texture of the output image has the same blue noise characteristics as in the error diffusion method. Also, the amount of calculation is the same as the error diffusion method.
[0089]
By using the threshold diffusion method, it is possible to eliminate the disadvantage of the error diffusion method that dots in the background portion close to white or black are easily arranged in a line while maintaining the output image quality similar to that of the error diffusion method. Further, the threshold diffusion method has various advantages such as reproduction of low contrast edge components.
[0090]
With regard to halftone reproduction by binarization, it is considered that attention will be given to visual characteristics in the future. In that case, the balance between the overall characteristics of the observed image and the local characteristics is important. That is, it is necessary to reproduce the entire gradation while maintaining local correlation in the image. This is because the person who observes the image is thought to be psychologically balanced so that the characteristics of both are maximized. Therefore, a method is required that can easily reproduce local zeroth order components and first order or higher components of the input image. The threshold value diffusion method contributes to an image creation method that satisfies such requirements.
[0091]
[Second Embodiment]
The image processing apparatus according to the second embodiment uses a plurality of image creation apparatuses according to the first embodiment to reduce the gradation of a multi-value image composed of a plurality of color components.
[0092]
Here, the processing of a color image composed of four colors C (cyan), M (magenta), Y (yellow), and K (black) for a printer is targeted.
[0093]
FIG. 21 is a block diagram showing the configuration of the image processing apparatus according to this embodiment. Referring to the figure, the image processing apparatus inputs color image data, decomposes it into a multivalued image of C, M, Y and K, and outputs it to each of C, M, Y and K. The image forming apparatuses 100a to 100d corresponding to the image forming apparatuses 100a and 100d and an output unit 603 that outputs the binary images output from the image forming apparatuses 100a to 100d in a superimposed manner.
[0094]
Here, as one configuration of the image creating apparatuses 100a to 100d, the configuration of the first embodiment and its modification can be used.
[0095]
Different feedback coefficients β can be set in each of the image creating apparatuses 100a to 100d shown in FIG. That is, the feedback coefficient may be set for each color in consideration of the effect when the four colors are superimposed.
[0096]
Further, the initial threshold value Th (x) can also be set freely. That is, four types of appropriate patterns may be prepared and used as the initial threshold value. By such a device, it is possible to provide an advantageous effect on color reproduction when the four color image data are superimposed.
[0097]
That is, it is possible to perform processing so that noise due to interference does not occur when colors are superimposed. For example, in the process of using the dither screen as the initial threshold value Th (x), four types of screen patterns may be prepared by changing the pitch and angle of the screen, and corresponding to each color. In this way, processing specific to a color image can be performed in the present embodiment.
[0098]
Further, as shown in FIG. 21, it is desirable to process image data of four colors independently from each other for the purpose of simplifying the process, but it is not always necessary to do so.
[0099]
For example, the input may be considered as a set of vectors in a four-dimensional color space, and threshold diffusion may be performed in the four-dimensional space. In this case, the feedback coefficient is also set as a vector having components of each dimension. The initial threshold value is also set as a function whose output value is a vector (initial threshold value vector Th (x)).
[0100]
By adopting the threshold value diffusion method, the second embodiment also has the merits as described in the first embodiment, but this can be achieved by an appropriate method as in the second embodiment. By adapting to the color image, it is possible to further control the relationship of the color dot arrangement of each color on the output medium. Thereby, it can contribute to a more beautiful color reproduction. For example, in the present embodiment, this can be realized by setting a pattern that does not interfere with each other at the initial threshold value for each color.
[0101]
It is also possible to control the gradation of each color and adjust the color of the output image itself by independently adjusting the feedback coefficient (β) of each color. Thus, there is an advantage that a new degree of freedom in image quality adjustment can be achieved by applying the threshold value diffusion method to a color image.
[0102]
That is, as an effect when the threshold value diffusion method is adopted for the color image, the coefficient and the initial threshold value can be variably set for each color, and the adjustment operation for the dot arrangement at the time of output becomes possible. Also, tone reproduction at the time of output can be effected independently for each color, and the color can be adjusted.
[0103]
By appropriately performing these operations, it is possible to achieve better color reproduction such as the vividness of the color of the output color image itself and the optimization of the color.
[0104]
Such an invention can be applied to an apparatus that decomposes a digitized full-color gradation image into color components that can be output and outputs binary values for each color component. In particular, a color image output device such as a color printer outputs a color material such as ink on paper according to a pixel position, but the output of the color material of each color component is restricted to an on / off binary value. Many. In such a case, depending on the dot arrangement and area of each color material, color reproduction such as hue and saturation and processing for expressing a halftone (binarization processing) are performed. Can be implemented and the effect comes to life.
[0105]
In other words, processing can be performed with relatively simple calculation and a memory that is not so large, the gradation of the original color image can be sufficiently reproduced, and the image quality such as color reproduction can be further improved. A binarization process can be provided.
[0106]
[Modification]
Note that the apparatus configuration shown in FIG. 21 may be modified to the configuration shown in FIG.
[0107]
Referring to FIG. 22, the image processing apparatus sequentially inputs C, M, Y, and K image data from input unit 701 that inputs C, M, Y, and K image data. Image creating apparatus 100 that creates a value image, output unit 703 that outputs output data from image creating apparatus 100 as C, M, Y, and K image data, and C, M, Y, and K image processing, respectively The setting unit 705 for setting the initial threshold Th (x) and the feedback coefficient β in accordance with the control unit 707 and the control unit 707 for controlling the entire apparatus.
[0108]
In this apparatus, sequentially input C, M, Y, and K image data are serially processed by one image creating apparatus 100 and output. Further, the threshold value and the feedback coefficient are changed according to each color component. By configuring the apparatus in this way, it is possible to provide an image processing apparatus that performs the same operation as the apparatus of FIG.
[0109]
[Third Embodiment]
FIG. 29 is a block diagram showing a configuration of an image creating apparatus according to the third embodiment of the present invention. Referring to the figure, as in the first embodiment, the image creating apparatus includes an input unit 101, a threshold processing unit 103, a binary image output unit 105, an inversion unit 113, and an initial threshold. A value generation unit 107, a subtraction unit 109, a correction threshold value output unit 111, a subtraction unit 115, a coefficient multiplication unit 117, and a correction value memory 119 are provided.
[0110]
In the present embodiment, the image creating apparatus further includes a coefficient β input unit 601 for inputting a coefficient β, an initial threshold th setting unit 603, and a coefficient β setting unit 605. That is, this image creating apparatus is characterized in that the user sets at least one of the initial threshold th and the coefficient β. Thereby, the γ characteristic (image reproduction characteristic) in image reproduction can be controlled.
[0111]
FIG. 30 is a diagram illustrating the γ characteristic when the initial threshold th and the coefficient β are changed. In each of the initial threshold value th and the coefficient β, the horizontal axis of the graph shown indicates the input value, and the vertical axis indicates the output value.
[0112]
According to the error diffusion method that is the prior art, in order to vary the γ characteristic, it is necessary to use a correction function calculation unit, a lookup table, or the like for correcting the separate characteristic. However, when the threshold value diffusion method is used, the γ characteristic can be controlled only by changing th and β.
[0113]
[Fourth Embodiment]
FIG. 31 is a block diagram showing a configuration of an image creating apparatus according to the fourth embodiment of the present invention. Referring to the figure, as in the first embodiment, the image creating apparatus includes an input unit 101, a threshold processing unit 103, a binary image output unit 105, an inversion unit 113, and an initial threshold value. A generation unit 107, a subtraction unit 109, a correction threshold value output unit 111, a subtraction unit 115, a coefficient multiplication unit 117, and a correction value memory 119 are provided.
[0114]
In this embodiment, the image creating apparatus further includes an initial threshold th setting unit 603 that sets an initial threshold th, a coefficient k output unit 609 that outputs a coefficient k for calculating a coefficient β, A β calculation unit 611 that calculates β for each pixel based on an input value of the input unit 101, an initial threshold th, and a coefficient k, and a k setting unit 607 that sets the coefficient k.
[0115]
FIG. 32 is a diagram for explaining a method of calculating β by the β calculating unit 611. As shown in the figure, when the input value from the input unit 101 is “0”, the initial threshold th is set to the value of β. If the input value is equal to th, k is set as the value of β. If the input value is “1”, 1-th is set as the value of β.
[0116]
FIG. 33 is a diagram for explaining an effect in the present embodiment. Referring to the figure, by changing β according to the input value as shown in (A), the output (dot density) when the input is 0 (minimum level) as shown in (B). It is possible to give the apparatus a γ characteristic in which the output is 1 when 0 is 0 and the input is 1 (maximum level).
[0117]
That is, in order to express white or solid, it is often necessary to express dot density 0 (= no dot is applied) and dot density 1 (= dot density 100%). In order to ensure continuous gradation, the γ characteristic needs to be continuous and monotonous (increase or decrease). Further, the γ characteristic obtained by the halftoning process is not necessarily linear (= straight line).
[0118]
In the present embodiment, it is possible to perform image processing that satisfies such a requirement. That is, when the input is 0, if β = th, the output increases monotonously when the input dot is 0 (= minimum value) and the output dot density is 0 and the input is at an intermediate level. Characteristics can be obtained. Similarly, when β = 1−th when the input is 1 (= maximum value), the output dot density is 1 when the input is 1. Further, when the input is other than 0 or 1, β and th can be freely determined continuously with respect to the input. In this embodiment, the parameter k is arbitrarily determined and the input and β are related. With such a configuration, the required characteristics can be easily obtained by simply changing th and k.
[0119]
FIG. 34 is a diagram illustrating changes in γ characteristics when th is changed in a state where k = 0.7. Thus, changing th can change the brightness of the entire output image.
[0120]
FIG. 35 is a diagram showing a change in γ characteristics when the value of k is changed in a state where th = 0.5. Thus, the contrast of the output image can be changed by changing k.
[0121]
[Application example]
FIG. 36A is a diagram showing the characteristics of a certain laser printer, and is a diagram showing the relationship between the density of dots to be printed and the lightness L (how it appears to human eyes). As is apparent from the figure, the dot density and the lightness L are not linear characteristics. If the characteristics shown in FIG. 37C are obtained in the course of image processing, the characteristics of linear dot density and lightness L shown in FIG. 36B should be obtained.
[0122]
38 and 39 are diagrams showing the relationship between the input image obtained by changing k and th and the output dot density. 40 and 41 have the γ characteristics by applying the image creating apparatus according to the fourth embodiment to the laser beam printer having the characteristics shown in FIG. 36A and adjusting k and th. It is a figure which shows the relationship between the input image at the time of letting it appear, and output brightness.
[0123]
As shown in FIG. 40D, when only the error diffusion method is employed and γ correction is not performed, a desirable γ characteristic as shown in FIG. 36B cannot be obtained. By adjusting k and th by the threshold value diffusion method as in the present embodiment, γ characteristics close to desirable γ characteristics ((1) to (3) in FIG. 40 and (1) to ▲ in FIG. 41) are obtained. 3 ▼) can be obtained.
[0124]
Further, the weighting coefficient shown in FIG. 2 may be changed based on the change of the input value.
[0125]
Further, in the present embodiment, the coefficient β and the weighting coefficient are changed based on the input value, but values within the threshold diffusion algorithm (output value, inverted output value, feedback value, correction threshold value, etc.) Based on the above, th, β, and the weighting coefficient may be changed.
[0126]
Furthermore, th, β, and the weight coefficient may be changed depending on the type of image. Here, the image type includes the color of the image (C, M, Y, K, etc.), color / monochrome, whether it is an image or text, a natural image, or a graphic.
[0127]
Conventionally, when digital halftoning technology using an error diffusion method or the like is employed, when processing CMYK image data having the same value (image data, which is likely to be graphics), the same dot pattern is generated, resulting in image quality. Was damaged. Therefore, if any one of th, β, and the weighting coefficient is made different from each other in CMYK, such a problem can be solved.
[0128]
Further, the color development characteristics are generally different for each color material of CMYK. Therefore, in order to obtain preferable characteristics, it has been conventionally necessary to use a correction function calculation or a lookup table for correcting the characteristics for each color. However, by using the threshold value diffusion method, it is possible to obtain preferable characteristics for each of CMYK only by setting th, β and the like appropriate for each color.
[0129]
Furthermore, preferable γ characteristics vary depending on the type of image. For example, a high contrast is required in a text screen, but if the threshold value diffusion method is used, a high contrast can be obtained by setting β small. In addition, in a monochrome natural image, a high-key screen is generally preferable, so th can be reduced. In general, one image forming means must correspond to other types of images, and it is preferable to give different γ characteristics to each image data. Such a request can also be realized by giving different th, β, and weighting coefficients depending on the type of image.
[0130]
Further, although the correction value memory 119 is placed after the feedback coefficient multiplication unit 117, it may be located at a position indicated by (*) in FIG.
[0131]
Further, the weight shown in FIG. 2 may be arbitrarily changed by the user.
[Fifth Embodiment]
Further, the setting by the initial threshold th setting unit 603 and the k setting unit 607 shown in FIG. 31 may be automatically performed based on the output of the sensor 651 as shown in FIG. That is, an environmental value is detected by the sensor 651, and th and k (or a weighting factor and β) are changed according to the environmental value. Here, the environmental value is a value determined by temperature, power supply voltage, humidity, paper type, image forming method, remaining amount of color material, type of color material, secular change, photoconductor, light emitting element, and the like.
[0132]
The γ characteristic of the image forming means generally varies depending on the environmental value. In order to compensate for the change in the γ characteristic due to the change in the environmental value, the th, k, β and weighting factors are changed. Further, instead of detecting the environmental value, the output from the output unit 105 may be directly detected.
[0133]
[Sixth Embodiment]
FIG. 43 is a block diagram showing a configuration of an image creating apparatus according to the sixth embodiment of the present invention. Referring to the figure, the image creating apparatus is similar to the first embodiment in that input unit 101, threshold value processing unit 103, binary image output unit 105, inversion unit 113, and initial threshold generation are performed. Unit 107, subtraction unit 109, correction threshold value output unit 111, subtraction unit 115, coefficient multiplication unit 117, and correction value memory 119.
[0134]
In this embodiment, the image creating apparatus further includes an n setting unit 701 that sets a maximum value (coefficient n) of a noise level, a coefficient n output unit 703 that outputs a coefficient n, an input value, a coefficient n, and an initial value. A noise generation unit 705 that generates noise based on the threshold th and an addition unit 707 that adds the initial threshold th and noise are provided.
[0135]
Similarly to the fourth embodiment, the image creation apparatus includes a coefficient k output unit 709, a β calculation unit 701 that calculates β for each pixel based on the input value, the initial threshold th, and the coefficient k. It has.
[0136]
FIG. 44 is a block diagram showing a configuration of the noise generator 705 of FIG.
Referring to the figure, a noise generator 705 generates a noise level (N) based on an input value, a coefficient n, and an initial threshold th, and generates a periodic noise pattern. A noise pattern generator 705b to be generated, and a multiplier 705c that multiplies the output of the noise level generator 705b and the output of the noise pattern generator 705b.
[0137]
FIG. 45A is a diagram illustrating processing of the noise generation unit 705, and FIG. 45B is a diagram illustrating processing of the β calculation unit 701.
[0138]
As shown in FIG. 45A, the noise level is 0 if the input value is 0, n if the input value is th, and 0 if the input value is 1.
[0139]
Since the process of the β calculation unit 701 shown in FIG. 45B is the same as that of the fourth embodiment, description thereof will not be repeated here.
[0140]
46 to 51 are diagrams showing specific examples of patterns output by the noise pattern generator 705b. 46 to 49 are concentrated dither patterns, FIG. 50 is a distributed dither pattern, and FIG. 51 is a line pattern.
[0141]
Such a pattern may be stored in the memory in advance, or the pattern may be calculated from information on the pixel position. By adding such a regular pattern to the threshold value, the image quality of the output image can be improved.
[0142]
That is, the threshold diffusion method has a problem that crystallization and texture shift are larger than the error diffusion method (dots are more regularly arranged in a grid pattern than the error diffusion method, and the texture shift is conspicuous) As shown in FIG. 34 and FIG. 35, there is a problem that the graph showing the γ characteristic has irregularities. In the present embodiment, these problems can be solved by adding noise.
[0143]
In the case of improving the texture by actively showing the pattern, it is preferable to use a concentrated dither pattern or a line type pattern. On the other hand, if you do not want to show the pattern, it is better to use a distributed pattern.
[0144]
[Effect of the sixth embodiment]
52 and 53 are diagrams showing changes in the γ characteristics when the coefficient k and the initial threshold value th are changed, and are diagrams corresponding to FIGS. 34 and 35. FIG. Here, the coefficient n is set to 0.2.
[0145]
As is apparent from the figure, when noise is added to the threshold according to the present embodiment, it is possible to prevent the γ characteristic curve from being uneven.
[0146]
FIG. 54 is a diagram showing a specific example of texture shift. In the figure, one square indicates the result of binarizing the current image having a constant density by the threshold value diffusion method. As shown in the figure, a pattern (texture shift) appears due to a change in the dot pattern within one square.
[0147]
55, 56 and 57 are diagrams showing the results of image processing with the patterns of FIGS. 46, 47 and 51 added in the present embodiment, respectively. As shown in the figure, it can be seen that the present embodiment can prevent the occurrence of texture shift.
[0148]
In the above-described embodiment, a periodic pattern is added as noise, but random noise may be added. Here, the random noise indicates, for example, white noise, blue noise, or pink noise. White noise is noise that uniformly includes all frequency components, blue noise is noise that includes more high frequency components, and pink noise is noise that includes more low frequency components.
[0149]
FIG. 58 shows an example in which white noise is added to the original image corresponding to FIG. 54 to improve the texture shift.
[0150]
The most common noise is white noise and is easy to generate. When the resolution is important, generally better results can be obtained by using blue noise. On the contrary, when a soft texture is desired, it is better to use pink noise.
[0151]
Although noise is added to the threshold value, noise may be added to any signal in the image creating apparatus such as a feedback coefficient, a weighting coefficient, an initial threshold value, or an input value. Furthermore, it is good also as noise by changing whether the range which diffuses a threshold value is widened or narrowed.
[0152]
In the present embodiment, the noise intensity is changed according to the input value. However, values within the threshold diffusion algorithm (for example, output value, inverted output value, feedback value, initial threshold value, feedback coefficient, etc.) Based on the above, the intensity and pattern of noise to be added may be changed.
[0153]
Further, the intensity and pattern of noise to be added may be changed depending on colors such as CMYK. That is, when printing a plurality of colors in an overlapping manner, if the same pattern exists in different colors, they interfere with each other and cause a reduction in image quality. Therefore, such a problem can be solved by adding a different signal pattern for each color, changing the intensity, changing the angle of the pattern, or changing the period.
[0154]
Furthermore, the intensity of the signal pattern to be added and the type of pattern may be changed depending on the type of image. The types of images include color / monochrome, image / text, natural image / graphic, and the like.
[0155]
Furthermore, the intensity and pattern of the signal pattern to be applied may be changed according to environmental values that are values based on temperature, humidity, power supply voltage, remaining amount of color material, type of color material, aging, photoconductor, light emitting element, etc. .
[0156]
This is because the optimum value of the signal (noise) to be added differs depending on the type of image and the environmental value.
[0157]
[Sample output]
59 to 63 are diagrams showing image print results using the threshold value diffusion method and the error diffusion method.
[0158]
59 to 61 and FIG. 63 show the output results by the threshold diffusion method, and FIG. 62 shows the output results by the error diffusion method.
[0159]
59, the coefficients used in the threshold value diffusion method are k = 0.9, th = 0.7, n = 0.2, and in FIG. 60, k = 0.9, th = 0. 75, n = 0.2, k = 0.9, th = 0.8, n = 0.2 in FIG. 61, and k = 0.9, th = 0.7, in FIG. n = 0.2. The laser printer used for printing these images is a printer having the characteristics shown in FIG.
[0160]
As can be seen from these output results, when the error diffusion method is used (FIG. 62), the image is crushed black, but when the threshold diffusion method is used, a good image can be obtained. effective.
[0161]
[Seventh embodiment]
In the first to sixth embodiments, the input multi-valued image is binarized. In the seventh embodiment, multi-value processing is performed by the threshold value diffusion method. is there. Here, the multi-value conversion processing refers to processing for converting an input multi-value image of, for example, 256 gradations into a multi-value image of, for example, 4 gradations having a smaller number of gradations.
[0162]
First, the basic concept of multilevel processing will be described. FIG. 64 is a graph for explaining conversion from a 256-gradation input image to a 4-gradation output image. In FIG. 64, the input image is divided into three sections, and binarization processing using the threshold value diffusion method is performed for each section. That is, threshold diffusion that outputs two gradations of gradation 0 and gradation 1 is performed in section a, and threshold diffusion that outputs two gradations of gradation 1 and gradation 2 is performed in section b. In section c, threshold value diffusion for outputting two gradations of gradation 2 and gradation 3 is performed. By the binarization process using the threshold value diffusion method for each section, an input image with 256 gradations is converted into an output image with 4 gradations from gradation 0 to gradation 3. The boundary value used for classification is determined according to the output gradation. For example, if the output gradation is 4 gradations of 0, 0.4, 0.8, and 1, the input image is 0 to 102 (= 255 * 0.4), 103 to 204 (255 * 0.8). ), 205 to 255.
[0163]
FIG. 65 is a block diagram showing the configuration of the image creation apparatus in the present embodiment. In the figure, a portion surrounded by a broken line A is a portion that performs binarization processing by the threshold value diffusion method, and other portions perform processing such as input image segmentation and output gradation assignment. Since the binarization processing part can use the same configuration as in the first to sixth embodiments, description thereof will be omitted. Needless to say, the threshold value Th (x) and the feedback coefficient β can be changed manually or automatically as in the above-described embodiment. Hereinafter, the configuration other than the binarization processing part will be described.
[0164]
First, a multi-value image (0 to 1) is input to the input unit B1. For example, in the case of an image with 256 gradations, a value normalized to 0/255 to 255/255 is input. The input image is divided into three sections according to the input gradation by the section B2. That is, when the input image is 0 to 102/255, it is classified into category a, when the input image is 103/255 to 204/255, it is classified into category b, and when the input image is 205/255 to 255/255, it is classified into category c. It is divided. On the other hand, the normalization processing unit B3 outputs a value Rin obtained by normalizing the input image from 0 to 1 in each section according to the classification result of the classification unit B2.
[0165]
The threshold processing unit B4 outputs “1” if Rin ≧ threshold Th (x) with respect to the normalized value Rin, and “0” if Rin <threshold Th (x). The normal binarization process of outputting “is performed. In the assigning unit B5, output gradations are assigned based on the classification result in the sorting unit B2 and the binarization processing result in the threshold processing unit B4. That is, when the input image is category a, if the binarization processing result is 0, gradation 0 is assigned, and if the binarization processing result is 1, gradation 1 is assigned. When the input image is category b, gradation 1 is assigned if the binarization processing result is 0, and gradation 2 is assigned if the binarization processing result is 1. When the input image is a section c, if the binarization processing result is 0, gradation 2 is assigned, and if the binarization processing result is 1, gradation 3 is assigned. Thus, binarization processing is performed for each section. It is possible to convert a 256-gradation input image into a 4-gradation output image.
[0166]
In the above description, only conversion from an input image of 256 gradations to an output image of 4 gradations is shown, but conversion from an arbitrary input gradation to an arbitrary output gradation is also possible with the same method. is there.
[0167]
The embodiment disclosed this time should be considered as illustrative in all points and not restrictive. The scope of the present invention is defined by the terms of the claims, rather than the description above, and is intended to include any modifications within the scope and meaning equivalent to the terms of the claims.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of an image creating apparatus according to one embodiment of the present invention.
FIG. 2 is a diagram for explaining the operation of a correction value memory 119;
FIG. 3 is a diagram showing a halftone processing result using a systematic dither method.
FIG. 4 is a diagram illustrating a halftone processing result using an error diffusion method.
FIG. 5 is a diagram showing a halftone processing result by a threshold diffusion method.
FIG. 6 is a diagram showing a halftone processing result by a threshold diffusion method.
FIG. 7 is a diagram illustrating a halftone processing result by an error diffusion method.
FIG. 8 is a diagram illustrating a halftone processing result using an error diffusion method with edge enhancement.
FIG. 9 is a diagram showing a threshold diffusion processing result when a feedback coefficient β = 1.
FIG. 10 is a diagram illustrating a threshold value diffusion process result when a feedback coefficient β = 0.
FIG. 11 is a diagram showing the result of threshold value diffusion processing when the feedback coefficient β is 0.5.
FIG. 12 is a diagram showing a relationship among an input value Im, a threshold value Th, and an output value Out when β = 0 in the error diffusion method.
FIG. 13 is a diagram illustrating a relationship among an input value Im, a threshold value Th, and an output value Out when β = 1 in the error diffusion method.
FIG. 14 is a diagram illustrating a relationship between an input value and an output value when a threshold value is a median value.
FIG. 15 is a diagram illustrating a relationship among an input value Im, a threshold value Th, and an output Out when β = 0 in the threshold value diffusion method.
FIG. 16 is a diagram showing a relationship among an input value Im, a threshold value Th, and an output Out when β = 1 in the threshold value diffusion method.
FIG. 17 is a diagram illustrating a relationship among an input value Im, a threshold value Th, and an output Out when β = 0.5 in the threshold value diffusion method.
FIG. 18 is a diagram illustrating a relationship between an input value and an output value when a threshold value is a median value.
FIG. 19 is a block diagram of an image creating apparatus using a threshold value diffusion method for modulating an initial threshold value Th (x) by an input value.
20 is a diagram showing a modification of the apparatus of FIG.
FIG. 21 is a block diagram illustrating a configuration of an image processing apparatus according to a second embodiment.
22 is a block diagram showing a modification of the apparatus of FIG. 21. FIG.
FIG. 23 is a diagram for explaining a specific calculation example by the threshold value diffusion method;
FIG. 24 is a diagram illustrating an image to be processed.
FIG. 25 is a diagram showing a processing result using a threshold value diffusion method.
FIG. 26 is a diagram illustrating a processing result by an error diffusion method.
FIG. 27 is a diagram illustrating a result of threshold value diffusion processing.
FIG. 28 is a diagram showing the result of threshold value diffusion processing for each step.
FIG. 29 is a block diagram illustrating a configuration of an image creating apparatus according to a third embodiment.
FIG. 30 is a diagram illustrating a γ characteristic when an initial threshold th and a coefficient β are changed.
FIG. 31 is a block diagram illustrating a configuration of an image creating apparatus according to a fourth embodiment.
FIG. 32 is a diagram for explaining a method of calculating β by a β calculating unit 611;
FIG. 33 is a diagram for explaining an effect in the fourth embodiment;
FIG. 34 is a diagram showing changes in γ characteristics when th is changed in a state where k = 0.7.
FIG. 35 is a diagram showing changes in γ characteristics when the value of k is changed in a state where th = 0.5.
FIG. 36 is a diagram illustrating a relationship between dot density and brightness of a laser printer.
FIG. 37 is a diagram illustrating a preferable γ characteristic for the printer of FIG. 36.
FIG. 38 is a diagram showing changes in γ characteristics when th is changed.
FIG. 39 is a diagram illustrating changes in γ characteristics when k is changed.
FIG. 40 is a diagram showing changes in γ characteristics by the threshold value diffusion method.
FIG. 41 is a diagram showing changes in γ characteristics by the threshold value diffusion method.
FIG. 42 is a block diagram of an image forming apparatus according to a fifth embodiment.
FIG. 43 is a block diagram illustrating a configuration of an image creating apparatus according to a sixth embodiment.
44 is a block diagram showing a configuration of a noise generation unit 705. FIG.
FIG. 45 is a diagram illustrating processing performed by a noise generation unit and a β calculation unit.
FIG. 46 is a diagram illustrating a first example of a noise pattern.
FIG. 47 is a diagram illustrating a second example of a noise pattern.
FIG. 48 is a diagram illustrating a third example of a noise pattern.
FIG. 49 is a diagram illustrating a fourth example of a noise pattern.
FIG. 50 is a diagram illustrating a fifth example of a noise pattern.
FIG. 51 is a diagram illustrating a sixth example of a noise pattern.
FIG. 52 is a first diagram showing effects according to the sixth embodiment.
FIG. 53 is a second diagram showing the effect of the sixth embodiment.
FIG. 54 is a diagram illustrating an example of texture shift.
FIG. 55 is a diagram showing a state in which the pattern of FIG. 46 is added.
FIG. 56 is a diagram showing a state in which the pattern of FIG. 47 is added.
FIG. 57 is a diagram showing a state in which the pattern of FIG. 51 is added.
FIG. 58 is a diagram illustrating an example in which white noise is added to improve texture shift.
FIG. 59 is a first diagram showing an output result by a threshold value diffusion method.
FIG. 60 is a second diagram showing an output result obtained by the threshold value diffusion method.
FIG. 61 is a third diagram showing an output result obtained by the threshold value diffusion method.
FIG. 62 is a diagram showing an output result by an error diffusion method.
FIG. 63 is a fourth diagram showing an output result by the threshold value diffusion method.
FIG. 64 is a diagram illustrating processing according to the seventh embodiment;
FIG. 65 is a block diagram illustrating a configuration of an image creating apparatus according to a seventh embodiment.
FIG. 66 is a diagram illustrating a configuration of an image processing apparatus using an error diffusion method.
[Explanation of symbols]
101 Image Input Unit, 103 Threshold Processing Unit, 105 Binary Image Output Unit, 107 Threshold Generation Unit, 109 Subtraction Unit, 111 Correction Threshold Output Unit, 113 Inversion Unit, 115 Subtraction Unit, 117 Feedback Coefficient Multiplication Part, 119 correction value memory.

Claims (26)

An image processing apparatus for converting a first image signal indicating a density level of each pixel with a predetermined number of gradations into a second image signal having a smaller number of gradations than the predetermined number of gradations,
A converter that sequentially inputs a first image signal of each pixel, compares the density level of each pixel with a corresponding threshold value, and converts it to a second image signal;
A feedback circuit that generates a threshold value corresponding to a subsequent pixel based on the signal level of the second image signal output from the converter and the threshold value used for the conversion of the second image signal ;
The feedback circuit includes:
Means for calculating a difference between an inverted output of the signal level of the second image signal and a threshold value used for conversion of the second image signal;
Means for multiplying the difference by a feedback coefficient to calculate a feedback value;
And a means for generating a threshold value corresponding to a subsequent pixel by correcting a reference threshold value by a value obtained by multiplying the feedback value by a predetermined weight constant . It means for calculating the feedback value includes a feedback coefficient setter for setting the feedback coefficient, the image processing apparatus according to claim 1. The feedback coefficient setter is changeable the feedback coefficient, the image processing apparatus according to claim 2. The feedback coefficient setter changes the feedback coefficient in accordance with the density level of each pixel to be converted by the converter, the image processing apparatus according to claim 3. The feedback coefficient setting unit includes:
Means for calculating a feedback coefficient according to a predetermined relational expression based on the magnitude relationship between the density level of each pixel converted by the converter and a predetermined initial threshold value ;
And means for changing the predetermined relationship, the image processing apparatus according to claim 4. 2. The correction circuit according to claim 1, wherein the feedback circuit includes a correction value memory that distributes the calculated feedback value to a plurality of peripheral pixels to be subsequently converted according to a weight set for each peripheral pixel. Image processing apparatus. The image processing apparatus according to claim 6 , wherein the weight is changeable. The threshold generator for the variable ized threshold of the reference for each transformation of each pixel further comprising, an image processing apparatus according to claim 1. The image processing apparatus according to claim 8 , wherein the threshold generation unit changes the reference threshold according to a periodic pattern corresponding to a position of a pixel to be converted.   The image processing apparatus according to claim 1, further comprising a multiplier provided in front of the converter and multiplying a density level of each pixel converted by the converter by a predetermined coefficient. The image processing apparatus according to claim 10 , wherein the predetermined coefficient is changeable. A gradation characteristic indicating a relationship between the level of the first image signal and the level of the converted second image signal is adjusted by changing at least one of the reference threshold value and the feedback coefficient. The image processing apparatus according to claim 1. A gradation characteristic with respect to the maximum level and the minimum level of the first image signal is fixed, and an intermediate level gradation characteristic between the maximum level and the minimum level is set to at least one of the reference threshold value and the feedback coefficient. The image processing apparatus according to claim 12 , wherein adjustment is performed by changing. The image processing apparatus according to claim 1, wherein a signal component independent of the first image signal to be converted is superimposed on the reference threshold value . The image processing apparatus according to claim 14 , wherein the superimposed signal component represents a periodic pattern. The image processing apparatus according to claim 15 , wherein the periodic pattern is any one of a distributed dither pattern, a concentrated dither pattern, and a line pattern. The image processing apparatus according to claim 14 , wherein the superimposed signal component is random noise. The image processing apparatus according to claim 17 , wherein the random noise is any one of white noise, blue noise, and pink noise.   The image processing apparatus according to claim 1, wherein the feedback circuit feeds back a difference between an inverted value of a signal level of a second image signal output from the converter and the threshold value. An image processing method for converting a first image signal indicating a density level of each pixel with a predetermined number of gradations into a second image signal having a smaller number of gradations than the predetermined number of gradations,
A first image signal of each pixel are sequentially inputted, as compared to the corresponding threshold density level of each pixel, a conversion step of converting the second image signal,
A feedback step of generating a threshold value corresponding to a subsequent pixel based on a signal level of the second image signal output from the converter and a threshold value used for conversion of the second image signal ;
The feedback step includes
Calculating a difference between an inverted output of the signal level of the second image signal and a threshold value used for conversion of the second image signal;
Multiplying the difference by a feedback coefficient to calculate a feedback value;
And correcting a reference threshold value by a value obtained by multiplying the feedback value by a predetermined weight constant to generate a threshold value corresponding to a subsequent pixel . 21. The image processing method according to claim 20 , further comprising a threshold generation step of changing the reference threshold for each pixel conversion. Adjusting a gradation characteristic indicating a relationship between the level of the first image signal and the level of the second image signal after conversion by changing at least one of the reference threshold value and the feedback coefficient; The image processing method according to claim 20 . A gradation characteristic with respect to the maximum level and the minimum level of the first image signal is fixed, and an intermediate level gradation characteristic between the maximum level and the minimum level is set to at least one of the reference threshold value and the feedback coefficient. The image processing method according to claim 22 , wherein adjustment is performed by changing. An image processing apparatus for converting a first image signal indicating a density level of each pixel with a predetermined number of gradations into a second image signal having a smaller number of gradations than the predetermined number of gradations,
A classification circuit that sequentially inputs a first image signal of each pixel and divides the first image signal into sections corresponding to the number of gradation levels of the second image signal;
A normalization circuit that normalizes the input first image signal within the section divided by the section circuit;
A comparator that sequentially inputs the first image signal normalized by the normalization circuit and compares the signal level of each pixel with a corresponding threshold value;
A feedback circuit on the basis of the threshold used in the comparison result and the comparison output from the comparator, and generates a threshold value corresponding to a subsequent pixel,
An allocation circuit that assigns the gradation level of the second image signal to each pixel according to the comparison result output from the comparator and the classification divided by the classification circuit;
The feedback circuit includes:
Means for calculating a difference between an inverted output of the comparison result output from the comparator and a threshold value used for the comparison;
Means for multiplying the difference by a feedback coefficient to calculate a feedback value;
And a means for generating a threshold value corresponding to a subsequent pixel by correcting a reference threshold value by a value obtained by multiplying the feedback value by a predetermined weight constant . The feedback circuit, the feedback value calculated, the plurality of peripheral pixels that are subsequently converted to a correction value memory for dispersing in accordance with the weight set for each peripheral pixel, according to claim 24 Image processing apparatus. 25. The image processing apparatus according to claim 24 , further comprising a threshold generation unit that changes the reference threshold for each pixel conversion.

Images