JPH09247446A - Signal processor and method therefor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce the occurrence of a pseudo contour and to obtain an excel lent image quality by preventing continuous occurrence of a single quantization representative value in any input signal value. SOLUTION: In the case of using plural quantization values for an input signal and generating and outputting a value finner than a quantization interval based on an incidence ratio, the input signal is classified into plural sets, e.g. four ranges depending on the value of the input signal, a quantization representative value D not a member of the classified set I (D belongs to 1), for example, a quantization representative value 170 is used for a signal within a range of 64-128 and a quantization representative value 85 is used for a signal within a range of 128-192 for quantization.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a signal processing apparatus and method, and more particularly to a signal processing apparatus and method for pseudo-grading image information.

[0002]

2. Description of the Related Art Conventionally, in digital signal processing for converting image information into pseudo gradation, sampling and quantization have been essential steps. In general quantization technology,
When the number of representative values that can be roughly represented is limited, signals having values other than the representative value are represented by the closest representative value.

The selection of the representative value can be classified into a so-called linear quantization method for uniformly classifying the signal and a so-called non-linear quantization method for unevenly dividing the quantization step into a coarse portion and a dense portion. It is also possible to classify the method into a scalar quantization method that executes quantization in only one-dimensional direction and a vector quantization method that clusters and quantizes vector information created by multidimensional information.

These quantization techniques are used in general digital signal processing in general. Mainly PC
Encoding represented by M communication, compression of digital information such as voice and image, and expression method when the number of gradations that can be expressed is limited in an image output device such as a copying machine, a facsimile, and a printer. Applied to.

A problem in quantization is handling of quantization error. If a large number of representative values used for quantization are set with respect to the information amount of the input signal, the quantization error is reduced accordingly. Regarding this quantization error, there are a method of simply discarding the generated error and a method of positively utilizing the information of the generated error for other quantization conditions. The latter includes a method of controlling the size of the next quantization step according to the size of the error, such as an adaptive quantizer.

If the quantization step is large,
In order to prevent the quantized representative value from continuing to be uniform, quantization is performed after the dither signal is superimposed on the input signal, or the enhanced gray scale method (B
isignami, W .; T. , Richards,
G. FIG. P. Whelan, J .; W. : The Imp
loved Gray Scale and the
Coarse-Fine PCM Systems, Tw
o New Digital TV Bandwidth
h Reduction, Proc. IEEE, 54,
3, pp. 376-390 (1966-3)), there is also a method of controlling so that the average luminance (density) for each small area of the received image matches the level of the original signal.

From a long time ago, in order to represent a halftone of a natural image or the like, a method of dispersing the quantized values and outputting an average (pseudo) gradation between visually quantized steps Is used. The pseudo gradation process in the image output device uses the use of these quantization errors.

A pseudo-halftone process which has been particularly popular in recent years is a method called an error diffusion method which has the same concept as the above-mentioned improved gray scale method. This method is described in reference R. Floyd & L. Steinb
erg, “An Adaptive Alogorit
hm for Special Gray Scal
e ", SID 75 Digest, pp 36-37, the multi-valued image data of the pixel of interest is binarized, and the error between the binarization level and the multi-valued image data is weighted with a predetermined weight to determine the neighborhood of the pixel of interest. It is added to the data of the pixel.

Further, the multilevel error diffusion method is an extension of this to some quantized representative values (multilevel). This is a method of diffusing an error generated during the number of output levels smaller than the number of input levels to peripheral pixels, and the processing between each level is almost equivalent to the processing of the binary error diffusion method.

FIG. 9 shows a block diagram for realizing this multilevel error diffusion method. In FIG. 9, 201 is an input terminal for inputting image information, 202 is an adder, 203 is an error memory,
Reference numeral 204 is a subtractor, 205 is a section determination unit, 206 is a representative value selection unit, and 207 is an output terminal for outputting image information.

In the general multi-valued error diffusion method that has been conventionally performed in the above-mentioned structure, first, the input terminal 201 is used.
The error information accumulated from the peripheral pixels for the target pixel stored in the error memory 203 is added to the image information input from the adder 202 (the information after the addition is hereinafter referred to as “correction information”). Subsequently, the correction information is input to the section determination unit 205, and it is determined to which section among the plurality of sections divided in advance.

The division setting of this section may be linear or in consideration of experimentally obtained nonlinearity.
When the section is determined, the representative value selection unit 206 outputs the quantized representative value expressing the determined section.

Since the quantization error is the difference between the selected quantized representative value and the correction information, the difference between the two is calculated by the subtractor, and predetermined weighting is applied to the error memory 2.
03, and the error propagation processing is repeated for the input information after the pixel of interest.

FIG. 10 shows an example of input / output allocation in the section determination unit 205 and the representative value selection unit 206 shown in FIG. In the following description, it is assumed that the quantized representative value that can be output is the four values (0, 85, 170, 255) shown in FIG.

In this case, the interval division is divided into four intervals (0 to 42, 42 to 42) including the quantized representative value so that the quantization error which is the difference between the signal of each interval and the quantized representative value is minimized.
128, 128-212, 212-255).

When it is determined in each section, a quantized representative value expressing the selected section is output. This is the same in any conventional technique using quantization, and when each section is considered as a set, the quantized representative value expressing the set is always a value included in the set.

In other words, the quantized representative value D representing the set I has a relationship of (DεI).

[0018]

However, the above-mentioned conventional technique has the following problems.

Assuming the case of the multi-valued error diffusion method described with reference to FIG. 9, the quantized representative value is output by the allocation shown in FIG. 10, and the difference between the quantized representative value and the correction information is propagated. Let's go. The assignment of FIG. 10 is replaced with the graph of FIG. 11 for clarity. In FIG. 11, the horizontal axis is the error-corrected input information, and the vertical axis is the output quantization information. A straight line 401 is a straight line showing input information = output information,
The difference between this straight line 401 and each quantized representative value becomes a quantization error.

Assuming that there is input information having continuous tones from 0 to 128, the difference between the straight line 401 and the quantized representative value becomes a quantization error, as described above, and the error occurs. Has a maximum value of 43. That is, in the transition from 0 to 85, the quantized representative value of 170 is not selected even if the maximum error is added, and the intermediate level is represented by the appearance ratio of the quantized representative values 0 and 85. . Similarly, in the transition from 85 to 128, the intermediate level is represented by the appearance ratio of the quantized representative values 85 and 170.

However, the problem here is when the input is at a level close to the quantized representative value. As is clear from the graph of FIG. 11, the quantization error is extremely small in the values near the quantized representative value. Therefore, there is a drawback that a single density is continuous and is perceived as a pseudo contour.

This is largely influenced by the difference in the followability depending on the density of error diffusion. For example, when the output value is considered to be a binary value of 0 and 255, the density of 128 can be almost expressed by two pixels, one ON and one OFF. However, the density of 1 is represented by one ON and 254 OFF.
Therefore, the number of pixels required to express a certain density varies depending on the density, and generally, the number of pixels required for output increases near the multi-value level that can be output.

To express a desired density, a plurality of pixels having the same value must be collected, but this is not always the case in an actual image. Therefore, in the vicinity of the multi-value level that can be output, one multi-value level color is almost continuous, which causes a pseudo contour. In the above example, when an input having a level near the quantization level value of 85 is performed, almost 85 pixels are continuously generated and are recognized as a single density portion.

Similar false contours are generated in the multi-value dither method which is another pseudo halftone process. That is, in order to express a level close to the quantized representative value, a large number of dither matrices must be used. After all, the level close to the quantized representative value is likely to be a pattern of one representative value color, unlike other gradation expressions.

The problem with the generation of this pseudo contour is that Ochi:
Image quality improvement by layering the multi-valued error diffusion method ", is also mentioned in the Institute of Image Electronics Engineers of Japan, Proceedings 94-01-05, pp17-20. However, Ochi's method is that the quantized image is layered. This is a method in which molding is repeatedly performed by means of a chemical method, and the processing is extremely complicated and requires a great deal of processing time.

The problem is exactly the same in any other technique using general quantization. Also in the method of quantizing after superimposing the dither signal on the original signal before the quantization processing, when the original signal is at a level close to the quantized representative value, the output of the quantized level close to the value is naturally large. That is, it can be effective only when the quantization representative value is switched.

In summary, in the signal processing method using quantization, the problem is caused by the fact that the magnitude of the quantization error is mixed up. That is, the fundamental cause is that the signal having a large quantization error and the signal having a small quantization error have the same processing form. Due to that cause, a portion in which a plurality of quantized values are generated and a portion in which a single quantized value is continuous are mixed, which causes a problem.

[0028]

The present invention has been made for the purpose of solving the above-mentioned problems, and is provided with, for example, the following structure as one means for achieving the above-mentioned objects.

That is, a signal processing device which uses a plurality of different quantized values for an input signal to generate and output a value finer than the quantization interval according to the appearance ratio, and outputs the input signal according to the signal value. A classifying unit for classifying into a plurality of sets, and a quantizing unit for quantizing an input signal included in the set I classified by the classifying unit using a quantized representative value D which is not (DεI). It is characterized by

Further, for example, it is characterized by further comprising calculation means for calculating the quantization error and feedback means for feeding back the calculation error calculated by the calculation means to the quantization condition of another input signal.

Further, there is provided a signal processing device which uses a plurality of different quantized values for an input signal and generates and outputs a value finer than the quantization interval according to the appearance ratio, wherein the input signal is output in accordance with the signal value. A classifying unit that classifies into a plurality of sets, and quantized representative values D and D ′ that are neither (DεI) nor (D′ εI) for the input signals included in the set I sorted by the classifying unit. And dithering means for performing dithering using.

Further, for example, it is characterized by further comprising calculation means for calculating the quantization error and feedback means for feeding back the calculation error calculated by the calculation means to the dithering condition of another input signal.

Furthermore, for example, the quantized representative value is a quantized representative value included in a set adjacent to the set I.

[0034]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings.

[First Example of Embodiment of the Invention] FIG. 1 is a conceptual diagram showing a basic concept of a signal processing method in an example of the embodiment of the present invention.

In FIG. 1, reference numerals 100 and 101 denote two sets (sets I and I '), which are the results obtained by evaluating signal values for the purpose of quantizing an input signal and classifying them into several sets. Is shown. Naturally, since it is a classification of sets for quantization, each set has a quantized representative value representing each set. That is, set I100, set I '
In 101, a representative value D ″ (10
2) and D (103).

In the above-mentioned conventional technique, for the purpose of reducing the quantization error, in order to represent each signal included in the set I100, the representative value D ″ is represented and each signal included in the set I′101 is represented. To a representative value D. This is the same in a method in which a quantization error is calculated and the error value is used when quantizing another signal.

On the other hand, the present embodiment is characterized by not using the quantized representative value included in the set. That is, the characteristic feature is that the representative value D, which is a representative value originally representing another set, is used to represent each signal included in the set I100.

That is, regarding the relationship between the set I and the quantized representative value D, the condition is that the quantized representative value D is not in the set I, and the condition is "not (DεI)".

The method of the present example in which the quantization is performed under the above conditions is particularly effective when the quantization error is calculated and the error value is used when quantizing another signal, or when the dither signal is used in combination. is there.

Naturally, in this example, although there is a representative value with a small quantization error, this is a method of intentionally selecting a representative value with a large error. As it deteriorates. However, in a system in which a plurality of pixels represent a signal value on an average (pseudo) basis, there is little problem even if the quantization error of one pixel becomes large.

According to this example, the bias of the quantization error can be reduced, and the size of the quantization error, especially the extremely small portion of the quantization error can be eliminated.

The case where the above-mentioned quantum method of the present example is applied to a multivalued error diffusion method will be described below as an example. FIG. 2 is a block diagram of a quantizer according to the multilevel error diffusion method in this example. In FIG. 2, the same components as those in FIG. 9 described above are designated by the same reference numerals, and detailed description thereof will be omitted.

In FIG. 2, 201 is an input terminal for inputting image information, 202 is an adder, 203 is an error memory, 2
Reference numeral 04 is a subtractor, 5 is a section determination unit, 6 is a representative value selection unit, 2
An output terminal 07 outputs image information. In this example, the configurations of the section determination unit 5 and the representative value selection unit 6 are significantly different from the configuration of FIG.

3 and 4, the section determination unit 5 in the multi-valued error diffusion method in this example having the configuration of FIG. 2 described above.
An example of input / output allocation in the representative value selection unit 6 is shown. Note that FIG. 4 is a graph of FIG.

Now, assume that the quantized representative value that can be output is four values (0, 85, 170, 255).
In this example, the interval division by quantization is divided into four intervals (0 to 6).
3,64-127,128-191,192-255)
Divided into In the following description, the number of divisions will be described by taking four cases as an example, but the present invention is not limited to the above division example, and it goes without saying that the division can be made into any number.

In the following description, for ease of explanation, a case where the divided sections are four sections will be described as an example.

These four sections are referred to as sections A, B, C and D.
Here, in this example, for example, the quantized representative value of each section is set as follows. This setting state is shown in FIG.

Interval A (0 to 63) ... Quantized representative value ... 0 Interval B (64 to 127) ... Quantized representative value ... 170 Interval C (128 to 191) ... Quantum Quantized representative value ... 85 Section D (192-255) ... Quantized representative value ... 255 As is clear from this setting, the quantized representative value of section B is not included in the set of section B. It is a value. The same applies to section C.

Input / output allocation in this example will be described with reference to FIG. In FIG. 4, the straight line 605 is a straight line indicating (input information = output information) as in FIG. 11 (the input information in this case is the information after addition correction from the error memory). In the figure, 601, 602, 603, and 604 are representative values of the respective sections.

That is, in this example, by adopting the representative value of the adjacent set as the representative value, in the present example,
For signal values other than 0 and 255, there is no representative value at which the straight line 605 intersects. That is, there is no signal in which the quantization error is 0 in the sections B and C. The fact that there is no part with a small quantization error means that there is no part that is represented by one quantized representative value, and that any signal value represents a signal value by the appearance ratio of multiple quantized representative values. Become.

Now, assume that the input information has a continuous gradation of 0 to 128. When the value is small, the quantized representative value is expressed by the appearance ratio of 0 and 170. As the value increases, the quantized representative value is not only 0, 170 but also the quantized representative value due to cumulative addition of the quantization error. Value 8
5 and 255 are also included so that pseudo gradation expression is performed. Even if a signal value close to the quantized representative value is input, since it is always expressed by a plurality of quantized representative values, the occurrence of the pseudo contour as described in the conventional example does not occur.

As described above, according to this example, it is possible to eliminate the bias of the quantization error, especially the phenomenon that the quantization error is extremely small.

Therefore, in the pseudo representation method using the appearance ratio of a plurality of quantized representative values, it is possible to prevent a single quantized representative value from continuously occurring at any input signal value. It is possible to reduce the occurrence of pseudo contours in the pseudo gradation expression of image processing, and it is possible to obtain good image quality.

[Second Example of Embodiment of the Invention] FIG. 5 is a principal block diagram showing a second example of the embodiment of the present invention. In the second example, an example in which the signal processing device is represented by multivalued dither will be described. In FIG. 5, reference numeral 701 is an input terminal for inputting multi-tone image information. 702
Is a section determination unit, which determines which section belongs to the signal value of the input information. 703 is a section determination unit 70
2 is a dither signal generating means for generating a dither signal suitable for the section based on the section information transmitted from No. 2, and 704 is an output value determining means for determining an output value by comparing the input information with a threshold value of the dither signal. The above configuration can be easily configured by using a LUT (lookup table).

The operation of the second example having the above configuration will be described below with reference to FIG. FIG. 6 is a diagram for explaining an example of gradation expression for an input in the second example. FIG. 6A shows how to assign inputs and outputs of a multi-valued dither method according to a conventional technique, and FIG. Shows the input / output allocation method of the multi-valued dither method in the second example. The characteristic of the second example having the above configuration is that the dither signal is generated according to the section.

For example, in the conventional example shown in (A), assuming that the quantized representative value that can be output has four values of A, B, C, and D, the interval is divided into three, and the signal value is A. In the section from B to B, the quantization representative values A and B are represented by binary dither. Similarly, in the section from B to C, the binary dither representation from B and C, and from C to D in the section from C to D. Binary dither expression is performed.

The problem in this case is the level represented by only a single quantized value. When this multi-valued dither is used for image processing, even though the whole is expressed by pseudo gradation, the halftone expression method mixes the pseudo gradation part and the flat part with a single density. Then, as described above, the pseudo contour is generated.

Therefore, in the second example, a process of eliminating a portion having a small quantization error is performed. That is, in order to eliminate the single density portion as much as possible, the gradation portion represented by only the quantized representative value B or C shown in FIG.
As shown in (B), the dither expression is set by using the quantized representative value with which the quantization error becomes large.

As shown in (B), the gradation expression near the signal value B is a dither expression using quantized representative values A and C without using the quantized representative value B. In addition, the gradation representation with a signal value near C is not represented by using the quantized representative value C, but is represented by quantized representative values B and D in the sections before and after the quantized section in the conventional case. To do.

In the second example, by setting the quantized representative values distant from each other in this way, even in any halftone expression, a pseudo gradation expression is produced by the appearance ratio of a plurality of quantized representative values, which gives a feeling of strangeness. It is possible to obtain the image quality without.

In the case of signal processing using dither, there are a method of using a threshold value of a dither signal and a method of superposing a dither signal on an input signal and quantizing it with a fixed threshold value, but they are equivalently the same. .

Further, in the second example, the system for outputting one pixel of the output signal for one pixel of the input signal has been described, but in the so-called density pattern method of outputting a plurality of outputs for one pixel of the input signal, The invention is effective.

Further, the scalar quantization in the one-dimensional direction has been described above, but this can be similarly used in the case of quantizing multidimensional vector information.

For example, a case where there are some cluster-classified vector information as shown in FIG. 7 will be described. Now, it is assumed that the quantized representative value of each cluster is vector information described by a cross.

Attention is paid to the cluster A surrounded by the thick line in FIG. According to the conventional quantization method, the signal value included in the cluster A outputs the quantized representative value a included in the cluster A as shown in FIG. 8B, the shaded area in the cluster A shown in FIG. 8B is quantized using the quantized representative value a, and the quantized representative that is the area including the quantized representative value a surrounded by the shaded area. The signal value in the vicinity of the value a is quantized by selecting the quantized representative value of another cluster in the peripheral portion without using the quantized representative value a.

Also in this case, it is preferable that the quantization error is fed back to another input signal so that the quantization error becomes small as a whole.

As described above, according to the second example, this signal processing method can be applied to the multilevel error diffusion method,
Good image quality can be obtained with a slight modification compared to the conventional method. In particular, if the signal processing is performed by software, the hardware configuration can be made common with the conventional one,
It can be easily realized by only changing the settings. Furthermore, this can be applied to the multi-valued dither method, and good image quality can be obtained by a simple method.

The present invention can be used not only in the field of image processing but also in any quantization technique when there is a problem in the phenomenon that a certain quantization value appears continuously.

Even when the present invention is applied to a system composed of a plurality of devices (eg, host computer, interface device, reader, printer, etc.), a device composed of one device (eg, copier, facsimile) Device).

Further, an object of the present invention is to supply a storage medium having a program code of software for realizing the functions of the above-described embodiments to a system or apparatus, and to supply the computer (or CPU) of the system or apparatus.
And MPU) read and execute the program code stored in the storage medium.

In this case, the program code itself read from the storage medium implements the functions of the above-described embodiment, and the storage medium storing the program code constitutes the present invention.

As a storage medium for supplying the program code, for example, a floppy disk, a hard disk, an optical disk, a magneto-optical disk, a CD-ROM, a CD
-R, a magnetic tape, a nonvolatile memory card, a ROM, or the like can be used.

When the computer executes the readout program code, not only the functions of the above-described embodiment are realized, but also the OS (Operating System) running on the computer based on the instruction of the program code. ) May perform some or all of the actual processing, and the processing may realize the functions of the above-described embodiments.

Furthermore, after the program code read from the storage medium is written in the memory provided in the function expansion board inserted into the computer or the function expansion unit connected to the computer, based on the instruction of the program code. Needless to say, this also includes the case where the CPU or the like provided in the function expansion board or the function expansion unit performs a part or all of the actual processing and the processing realizes the functions of the above-described embodiments.

[0076]

As described above, according to the present invention, the error in the quantization error, especially the phenomenon in which the quantization error is extremely small, is reduced.

Therefore, in the pseudo representation method using the appearance ratio of a plurality of quantized representative values, a single quantized representative value is prevented from continuously occurring at any input signal value. If the present invention is used for the pseudo gradation expression of the image processing, it is possible to reduce the occurrence of pseudo contours and obtain good image quality.

Furthermore, according to the present invention, the present signal processing method can be applied to the multi-valued error diffusion method, and good image quality can be obtained with a slight modification as compared with the conventional method. In particular, if the signal processing is performed by software, the hardware configuration can be made common to the conventional one, and it can be easily realized by only changing the setting. Furthermore, this can be applied to the multi-valued dither method, and good image quality can be obtained by a simple method.

[0079]

[Brief description of drawings]

FIG. 1 is a conceptual diagram showing a basic concept of a signal processing method in an example of an embodiment of the present invention according to the present invention.

FIG. 2 is a block configuration diagram of a quantizer according to a multilevel error diffusion method in the present example.

FIG. 3 is a diagram for explaining setting of input / output allocation in a section determination unit and a representative value selection unit shown in FIG.

4 is a graph showing the setting of input / output allocation in FIG.

FIG. 5 is a block configuration diagram of a quantizer according to a multilevel error diffusion method in a second example of the embodiment of the present invention.

FIG. 6 is a diagram for explaining an example of gradation expression for an input in the second example.

FIG. 7 is an explanatory diagram of clustered vector information in the case of quantizing multidimensional vector information according to the second example.

FIG. 8 is a diagram for explaining the quantization in the example of FIG. 7 according to the second example.

FIG. 9 is a block diagram of a conventional quantizer using a multi-level error diffusion method.

10 is a diagram for explaining setting of input / output allocation in the section determination unit and the representative value selection unit shown in FIG. 9 of the conventional example.

11 is a graph showing the setting of input / output allocation in FIG.

[Explanation of symbols]

 5, 205, 702 Section determination unit 6, 206 Representative value selection unit 201, 701 Input terminal 202 Adder 203 Error memory 204 Subtractor 207, 705 Output terminal 703 Dither signal generation means 704 Output value determination means

Claims (10)

[Claims] 1. A signal processing device for generating a value finer than a quantization interval according to an appearance ratio by using a plurality of different quantized values for an input signal and outputting the quantized value. A classifying unit for classifying into a plurality of sets, and a quantizing unit for quantizing an input signal included in the set I classified by the classifying unit by using a quantized representative value D which is not (DεI) A signal processing device characterized by the above. 2. The method according to claim 1, further comprising calculation means for calculating a quantization error, and feedback means for feeding back the calculation error calculated by the calculation means to a quantization condition of another input signal. The signal processing device described. 3. A signal processing device which uses a plurality of different quantized values for an input signal and generates and outputs a value smaller than a quantization interval according to an appearance ratio, wherein the input signal is output in accordance with the signal value. Classifying means for classifying into a plurality of sets, and quantized representative values D and D ′ which are neither (DεI) nor (D′ εI) for the input signals included in the set I sorted by the classifying means. And a dithering unit for performing dithering using the signal processing apparatus. 4. The method according to claim 3, further comprising calculation means for calculating the quantization error, and feedback means for feeding back the calculation error calculated by the calculation means to the dithering condition of another input signal. The signal processing device described. 5. The signal processing apparatus according to claim 1, wherein the quantized representative value is a quantized representative value included in a set adjacent to the set I. 6. A signal processing method in a signal processing device, wherein a plurality of different quantized values are used for an input signal to generate and output a value smaller than a quantization interval according to an appearance ratio, wherein the input signal is a signal. A signal processing method characterized by classifying into a plurality of sets according to a value, and quantizing an input signal included in the classified set I using a quantized representative value D that is not (DεI). 7. The signal processing method according to claim 6, further comprising calculating a quantization error at the time of quantization and feeding back the calculated calculation error to a quantization condition of another input signal. 8. A signal processing method in a signal processing device, wherein a plurality of different quantized values are used for an input signal to generate and output a value finer than a quantization interval according to an appearance ratio, wherein the input signal is a signal. Quantized representative values D and D ′ that are classified into a plurality of sets according to values and are neither (DεI) nor (D′ εI) for the input signals included in the classified set I.
A signal processing method characterized by performing dithering using. 9. The signal processing method according to claim 8, further comprising calculating a quantization error and feeding back the calculated calculation error to a dithering condition of another input signal. 10. The signal processing method according to claim 6, wherein a quantized representative value included in a set adjacent to the set I is used as the quantized representative value.