JP2802629B2 - Image data compression device and image processing device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Overview] Adopting multi-tone adaptive block coding (CBTC),
Regarding an image data compression device and an image processing device for encoding a color image decomposed into a plurality of components, when encoding an image to a specific code amount, a threshold value for determining the magnitude of the code amount can be quickly determined. In the image data compression apparatus as described above, a gradation change amount detecting means for calculating a difference between a maximum gradation value and a minimum gradation value in a block; An appearance frequency calculating means for calculating an appearance frequency of the gradation change amount obtained by the gradation change amount detection means in the entire multi-valued image; and a statistical value of an appearance frequency for each gradation change amount obtained by the appearance frequency calculation means. A statistical value calculating means to be obtained and a threshold value determining means for determining the threshold value based on the statistical value obtained by the statistical value calculating means are provided.

[Industrial applications]

The present invention relates to an image data compression apparatus and an image processing apparatus, and in particular, divides a multi-valued image of a color image decomposed into a plurality of components into blocks each including a plurality of adjacent pixels, and for each block, a maximum level in the block. The difference between the tonal value and the minimum tone value is compared with the threshold value to determine one or a plurality of representative tones for displaying the image of this block. A multi-gradation adaptive block coding system for displaying a resolution value indicating the distribution of the representative gradation in each block by displaying the difference value with the gradation value, and coding the reference value, the difference value and the resolution component. The present invention relates to an image data compression device and an image processing device that employ the above.

[Conventional technology]

As a compression method using the high efficiency of image data, for example, there is a multi-tone adaptive block coding method (preprint 6 of the 1987 National Meeting of the Institute of Image Electronics Engineers of Japan).

This multi-tone adaptive block coding method (Generalized
Block Truncation Coding (hereinafter abbreviated to GBTC) is a technique for encoding a color image by converting, for example, luminance (Y) component image data, color difference (U) component image data, and color difference (V) component image data. The original image is divided into blocks of N × N pixels, and each pixel (X _ji ) is divided into the maximum / minimum gradation values within the block. ⁿ (n
= 0, 1, 2,...) Level, the quantization level of each pixel is expressed in a bit plane format, and gradation information and bit plane information are encoded.

The case where N = 4 and n = 2 will be described in detail. FIG. 3 shows the GBTC algorithm.
Each block has a difference between a maximum gradation value (L _MAX ) and a minimum gradation value (L _MIN ) in the block and encoding parameters T ₁ , T ₂ (T ₁ <
T ₂ ), as shown in FIG. 4, the coding mode is classified into the following three coding modes (mode A, mode B, and mode C).

Mode A: When L _MAX −L _MIN ≦ T ₁ , the pixels in the block are quantized to one level (P _o ).

Mode B: If T ₁ <L _MAX −L _MIN ≦ T ₂ , the pixels in the block are quantized to two levels (P ₁ , P ₂ ).

In the mode C: T ₂ <L _MAX −L _MIN , the pixels in the block have four levels (Q ₁
~ Q ₄ ).

The quantization levels are the reference level La of the block, the level interval Ld, and the level designation signal (φ ₁ ) _ij (φ ₂ ) _{ij for} each pixel.
Is described. If the average processing is AVE (),
As shown in the figure, each value required for encoding is calculated.

Mode A: P _o = AVE (X _ij ) = La (φ ₁ ) _ij = 0, (φ ₂ ) _ij = 0 (for all i, j) Mode B: P ₁ = AVE (X _ij ≧ ( L _MAX + L _MIN ) / 2) P ₂ = AVE (X _ij <(L _MAX + L _MIN ) / 2) La = (P ₁ + P ₂ ) / 2 Ld = P ₁ −P ₂ (φ ₁ ) _ij = 0: (However, X _ij ≧ (L _MAX + L _MIN ) / 2
In the case _{of) (φ 1) ij = 1} : ( _{However, X ij <(L MAX +} L MIN) / 2
(Φ ₂ ) _ij = 0: (for all i, j) Mode C: Q ₁ = AVE (X _ij ≧ (3L _MAX + L _MIN ) / 4) Q ₄ = AVE (X _ij <( _{L MAX + 3L MIN) / 4} ) La = (Q 1 + Q 4) / 2 Ld = 2 (Q 1 -Q 4) / 3 (φ 1) ij = 0, (φ 2) ij = 0: ( However, X _ij ≧ La + Ld / 2) (φ ₁ ) _ij = 0, (φ ₂ ) _ij = 1: (However, when La + Ld / 2> X _ij ≧ La) (φ ₁ ) _ij = 1, (φ ₂ ) _ij = 0: (when La> X _ij > = La−Ld / 2) (φ ₁ ) _ij = 1, (φ ₂ ) _ij = 1: (However, when La−Ld / 2> X _ij ) As shown in FIG. 5, the resolution components (φ ₁ , φ ₂ ) are connected between blocks and converted into two bitmaps, and each bitmap is converted into an MMR coding method which is a standard coding method for a binary image. Encode. Ld is variable-length coded after nonlinear quantization,
La nonlinearly quantizes the pre-difference (ΔLa) using DPCM coding and then performs variable-length coding.

In GBTC, when a certain image is encoded, the ratio of three encoding modes (mode A, mode B, and mode C) is determined by the values of the encoding parameters T ₁ and T ₂ , and the code amount as a whole is also determined. It is determined. Therefore, the total code amount for one image data can be changed by changing the values of the encoding parameters T ₁ and T ₂ . That is, if the ratio of mode A is increased, the code amount is reduced, and if the ratio of mode C is increased, the code amount is increased.

The above-described processing is performed on all the image data of the luminance (Y) component image data, the dye (U) component image data, and the color difference (V) component image data, and the sum of the code amounts of all the image data is one color. This is the code amount for the image.

[Problems to be solved by the invention]

By the way, in the conventional image data compression apparatus and image processing apparatus, when it is necessary to adjust the total code amount to a predetermined value or less, it is necessary to encode each image data according to the apparent complexity of the image to be encoded. The value of the encoding parameter is changed until the total code amount reaches a predetermined value. However, reproduction is performed according to the ratio of each image data to all codes, that is, how much code amount is allocated to luminance (Y) component image data, color difference (U) component image data, and color difference (V) component image data. Image quality will be greatly affected. This is particularly noticeable when the amount of code distribution between the luminance (Y) component and the color difference (U, V) component is changed.

Therefore, there is a problem that it takes time to select an appropriate encoding parameter which is encoded with an appropriate total code amount and has a good quality of a reproduced image by trial and error, and it takes time. .

Therefore, the present invention provides an image data compression apparatus and an image processing apparatus capable of quickly determining a threshold value for obtaining good image quality within a specific code amount when encoding an image to a specific code amount. It is intended to provide a device.

[Means for solving the problem]

In the present invention, a means for solving the above problem is to divide a multi-valued image of a color image into blocks composed of a plurality of adjacent pixels, and for each block, a maximum gradation value and a minimum gradation value in the block. Means for comparing the difference with the tonal value with the threshold value, means for calculating the number of representative tones for displaying the image of the block based on the comparison result by the comparing means, and means for calculating based on the number of representative tones Means for representing a representative gradation value of each block by a reference value in each block and a difference value calculated based on a difference between the representative gradation values, and calculating a resolution component indicating a distribution of the representative gradation value in each block. And a means for encoding the reference value, the difference value, and the resolution component, wherein the maximum gradation value and the minimum gradation value in the block are provided for each of a plurality of components constituting the color image. Find difference with value A gradation change amount detecting means, an appearance frequency calculating means for calculating an appearance frequency of a gradation change amount for each component constituting a color image from the gradation change amount obtained by the gradation change amount detecting means, A statistic calculating means for calculating a statistic of the gradation change amount for each component obtained by the appearance frequency calculating means; and a predetermined calculation performed on the statistic of each component calculated by the statistic calculating means. Image data, comprising: a threshold value determining unit that compares a calculated value with a predetermined value and determines a threshold value for determining the number of gradations in a block for each of the components. It is a compression device.

Further, the means of the present invention divides a multi-valued image of a color image into blocks composed of a plurality of adjacent pixels, and calculates a difference between a maximum gradation value and a minimum gradation value and a threshold value in each block for each block. Means for comparing, means for calculating the number of representative tones for displaying the image of the block based on the result of comparison by the comparing means, and a representative tone value calculated based on the representative tone value, Value, means for representing a difference value calculated based on a difference between the representative gradation values, means for calculating a distribution resolution component indicating the distribution of the representative gradation values in each block, And a means for encoding a resolution component.
For each of a plurality of components constituting a color image, a gradation change amount detecting means for obtaining a difference between a maximum gradation value and a minimum gradation value in a block, and a gradation change amount obtained by the gradation change amount detecting means An appearance frequency calculating means for calculating an appearance frequency of a gradation change amount for each component constituting the color image; and a statistic for calculating a statistic of the gradation change amount for each component obtained by the appearance frequency calculation means. A calculating unit that compares a value calculated by a predetermined calculation from the statistic of each component calculated by the statistic calculating unit with a predetermined value, and determines the number of tones in a block for each of the components. And a threshold value determining means for determining a threshold value for performing the processing.

[Action]

According to the investigation by the present applicant, the complexity of each component can be determined from statistical values such as the average value, standard deviation, and variance for each component obtained by calculating the gradation change amount in each block. .

Therefore, according to the present invention, the appearance frequency of the entire gradation change amount multi-valued image in a block is obtained, and further, a statistic related to the gradation change amount is calculated. Since the threshold value at which the quality can be obtained is determined, the threshold value can be quickly determined without trial and error.

Further, in the second invention of the present application, the invention can be applied to, for example, an image data compression apparatus employing a multi-tone provision type block encoding method. And the number of gradations can be determined quickly.

〔Example〕

An embodiment of an image data compression apparatus according to the present invention will be described below with reference to the drawings.

FIG. 2 shows an embodiment of the image data compression apparatus according to the present invention. In FIG. 1, reference numeral 1 denotes an input terminal of image data, and 2 denotes input image data as a plurality of image data, that is, luminance (Y) component image data, color difference (U) component image data, and color difference (V) component image data. A demultiplexer for separating and outputting, buffer memories for storing input image data 3, 4, and 5; a multiplexer 6 for selecting and outputting an image output from each buffer memory; 7 a maximum gradation of image data A gradation change amount detection unit 18 for detecting the value (L _MAX ) and the minimum gradation value (L _MIN ). A gradation number 18 determines the gradation number based on a threshold value from a threshold value calculation unit 16 described later. 4 shows a decision unit. In the same figure, reference numeral 19 denotes one or more representative gradation values, a reference gradation value (La), and a difference value (Ld) based on the maximum gradation value (L _MAX ), the minimum gradation value (L _MIN ), and the number of gradations. ), A reference tone value code generator 20 for encoding the determined reference tone value, a representative tone value buffer memory 21 for temporarily storing this value, and , 22 denotes a difference value code generator for coding the difference value, and 23 denotes a difference value buffer memory for storing the coded difference value. In addition, 24
Is a gradation value storage unit that stores the representative gradation value, 28 is a comparison unit that compares the image data with the representative gradation value, 25 and 29 are storage buffers corresponding to the two resolution components φ ₁ and φ ₂ , 26 , 30 is a resolution component encoding unit that encodes the resolution component, 27 and 31 are resolution component code buffers that store the encoded resolution component, and 32 is a reference tone value, a difference value, and a code of the resolution component. A multiplexer 33 is the demultiplexer 2 described above,
3 shows a control unit that executes input / output control of the multiplexer 6 and the multiplexer 32. These parts are compression parts of normal image data.

In this embodiment, as a part for determining a threshold value corresponding to a predetermined code amount, the above-mentioned gradation change amount detection unit 8 includes an appearance frequency calculation means for calculating the appearance frequency of the gradation change amount in the entire multi-valued image. , An appearance frequency table 9 for storing the appearance frequency, and a statistic calculation means for calculating the statistic of the appearance frequency, that is, the average value, the standard deviation, and the variance from the contents of the appearance frequency table. , A statistic memory for storing the statistic of each component calculated by the statistic calculation unit, and a statistic for comparing the statistic stored in the statistic memory A comparison unit 15 is provided. In the present embodiment, the above-described appearance frequency table 9 stores a combination of a cumulative frequency calculation unit 11 that calculates the cumulative frequency of appearance frequencies and a predetermined encoding parameter based on the relationship between the statistics of each image data. An encoding parameter memory 17, a statistic comparing unit 15, an accumulated frequency calculating unit 11, and a threshold calculating unit 16 for calculating a value of an encoding parameter based on an output of the encoding parameter memory 17 are provided. . Here, the statistic calculation unit 10, the statistic memories 12, 13,
14, the statistic comparison unit 15, the encoding parameter memory 17, and the threshold calculation unit 16 will be described in detail.

In this embodiment, the statistic is to obtain the standard deviation σ of the gradation change amount of each image data, and the calculated standard deviations σ _Y , σ _U , and σ _V are used as the statistic memories 12 and 13 of the respective components. , 14. The statistic comparison unit 15 calculates the average value σ _UV = (σ _U + σ _V ) / 2 of the standard deviations of the two color difference components, and calculates the ratio σ _Y / σ between this value and the standard deviation σ _Y of the luminance (Y) component. _Ask for _UV .

On the other hand, the encoding parameter memory 17 stores the above-described σ _Y /
coding parameters (闘値) T _1Y T _2Y applied to the luminance components and chrominance components corresponding to the value of sigma _UV, set of values of T _1UV T _2UV is stored. This coding parameter is a value indicating the percentage of the total data where the position of the threshold value in the cumulative frequency of each component is set, and the larger the value of σ _Y / σ _UV , the more the amount of code of the luminance component is distributed. Is large. For example, in this embodiment, there are two encoding parameter sets as shown in Table 1 where the value of σ _Y / σ _UV is 3 as a boundary. That is, the parameter set 1 is selected when σ _Y / σ _UV ≧ 3, and the parameter set 2 is selected when σ _Y / σ _UV <3.

Specifically, the standard deviations σ _Y , σ as shown in Table 2
_U, parameter set 1 is the image A having a sigma _V
, And the parameter set 2 is used for the image B.

Based on the cumulative frequency calculated by the cumulative frequency calculating unit based on such values, the threshold value in the image is output so as to be a predetermined percentage.

Next, the operation of the embodiment of the image data compression apparatus according to the present embodiment will be described.

In the present embodiment, a description will be given on the assumption that the gradation level of image data is 0 to 255.

First, the multi-valued image data input from the terminal 1 is demultiplexed by the demultiplexer 2 into buffer memories 3, 4,.
5 is stored as an original image. And the original image is 4 × 4
The image data is divided into blocks each composed of pixels, and image data X _ij of one block is sequentially read from the buffer memory. Gradation change amount detector 7, from the inputted multi-valued image data X _ij detects the maximum gradation value L _MAX and the minimum gradation value L _MIN output.

In the appearance frequency calculation unit 8, based on the maximum gradation value L _MAX and the minimum gradation value L _MIN in the block from the gradation change amount detection unit 7,
The difference value L _MAX -L _MIN is obtained, and the frequency of occurrence of the difference value L _MAX -L _MIN is counted.

Here, since the gradation level of the image data is 0 to 255, the difference value L _MAX -L _MIN is also distributed in the range of 0 to 255.

These processes are repeated for one screen, and the appearance frequencies of 256 types of difference values in one screen are calculated and stored in the appearance frequency table 9.

After calculating the appearance frequency, the statistic calculation unit 10 calculates the standard deviation of the difference value of the multi-valued image data based on the appearance frequency data in the appearance frequency table 9. The calculated standard deviation is stored in the Y component statistics memory 12, U
It is stored in either the component statistics memory 13 or the V component statistics memory 14.

The above processing is performed for the three components (Y component, U component, and V component). After the processing of each component is completed, the Y component statistics memory 12, the U component statistics memory 13, and the V component statistics memory 14
Are compared by the statistic comparison unit 15 as described above. According to the value of each statistic, the coding parameter (percentage to the cumulative frequency) of each component in the coding parameter memory 17 is selected.

After the above processing, the original image (one of the component images) is selected again according to the instruction of the control unit 33, and divided into blocks composed of 4 × 4 pixels as shown in FIG. ,
One of the blocks of image data X _ij is selected from the buffer memories 3, 4, and 5 of each component in accordance with an instruction from the control device 33, and is sequentially read.

The gradation change amount detection unit 7 calculates the maximum value L from the image data _Xij.
MAX and the minimum value L _MIN are detected and output.

The appearance frequency calculation unit 8 calculates the maximum gradation value L _MAX and the minimum gradation value L _MIN in the block from the gradation change amount detection unit 7,
The difference value L _MAX -L _MIN is calculated, and the appearance frequency of the difference value L _MAX -L _MIN is counted.

The above processing is repeated for one-component multivalued image data, and the appearance frequency in each component is calculated and stored in the appearance frequency table 9.

After calculating the appearance frequency, the cumulative frequency calculation unit 11 calculates the cumulative frequency based on the appearance frequency data in the appearance frequency table 9. Further, the threshold calculation unit 16 calculates the threshold from the output of the cumulative frequency calculation unit 11, the output of the statistic comparison unit 15, and the contents of the encoding parameter memory 17 as described above.

The number-of-tones determination unit 18 determines the difference value and the threshold value T ₁ (the Y component, T _1Y ,
If the UV component is _T1UV ), it is determined whether the representative gradation number is 1 or 2 or more. If the number of representative gradations is determined to be 2 or more, the difference value is compared with a _second threshold value T2 for determining the number of representative gradations, and it is determined whether the number of representative gradations is 2 or 4. .

According to the representative number of gradations determined by the number-of-gradations determining unit 18,
The representative tone value determination unit 19 obtains the tone in the block by linear or non-linear quantization.

When the number of representative gradations is 1, the average value in the block is set as the representative gradation value and the reference value La as shown in FIG.
Further, 0 is assigned to all pixels as the resolution components φ ₁ and φ ₂ as shown in FIG.

When the representative number of gradations is 2, an intermediate value between the maximum gradation value L _MAX and the minimum gradation value L _MIN in the block is calculated as shown in FIG. determination of the mean value P ₁ and the average P ₂ of the gradation values of less than between values intermediate. Further, an average value (P ₁ + P ₂ ) / 2 of P ₁ and P ₂ is calculated as the reference value La. As resolution component phi, assign a 0 for the pixels of the representative gradation value P _1, the representative gradation value is assigned a 1 to the pixels of P _2. phi ₂ assigns a 0 to all the pixels (FIG. 5 (2)). Further, the difference between P ₁ -P ₂ of the P ₁ and P ₂ is the difference value Ld.

When the number of representative gradations is 4, as shown in FIG.
The maximum gradation value L _MAX and the minimum gradation value L _MIN in the block are divided into four equal parts, and the average value Q ₁ and the lower 1/4 of the gradation values in the upper 1/4 range (including the boundary line) Request range average Q ₄ of the gradation values become (not including the border). Reference value La calculates the Q _1, average value of _{Q 2 (Q 1 + Q 4} ) / 2. Further, above for Q ₁ in, Q ₂ from 2
Let (Q ₁ −Q ₄ ) / 3 be the difference value Ld. Furthermore, the distance between Q ₁ and Q ₄ is 3
Calculate gradation values to be equally divided. The gradation value equally divided into Q ₁ and Q ₄ is set as a representative gradation value of the block. This is called Q ₁ , Q ₂ , Q ₃ , Q
₄ and set as shown in Fig. 4 (3). The resolution component is
Assignment is performed as follows according to the representative gradation value.

In the pixel representative gradation values Q _1, φ _{1 =} 0, and phi _{2 =} 0, in the pixel representative gradation value _{Q 2, φ 1 = 0,} φ 2 = 1, the representative gradation value Q ₃ In the pixel of, φ ₁ = 1, φ ₂ = 0, and in the pixel whose representative gradation value is Q ₄ , φ ₁ = 1, φ ₂ = 1.

As described above, after the representative gradation value, the reference value, and the difference value in the block are determined by the representative gradation value determination unit 19, the representative gradation value is stored in the gradation value storage unit 24.

The comparison unit 28 reads out one block of multi-valued image data X _ij one pixel at a time and compares it with the representative gradation value in the gradation value storage unit 24. With this, it is determined which of the representative gradation values is approximated, and the resolution components φ ₁ and φ ₂ corresponding to the representative gradation values are output. Stored in 29. The contents of the resolution component storage buffers 25 and 29 are read out by the encoding units 26 and 30 and encoded by a known MMR encoding, and the resolution component storage units 27 and 30 are read.
Stored in 31.

The reference value La calculated by the representative gradation determination unit 19 and the difference value Ld
Are variable-length coded by the code generators 20 and 22 and stored in the buffer memories 21 and 23, respectively.

As described above, after the completion of processing multi-tone image of 1 component, La, Ld, .phi.1, phi ₂ of the encoded data sequentially read via the multiplexer (MPX) 32, is sent.

Therefore, according to the present embodiment, the code amount of each image data is determined based on the statistic of the gradation change amount.
When encoding color image data within a predetermined code amount, a good image can be obtained, and the code amount of each image data can be distributed in a short time.

In the present embodiment, an example in which the standard deviation is used as the statistic used for the distribution of the code amount has been described. However, the same determination can be made by using the variance and the average as the statistic. Also,
In the present embodiment, the encoding parameter has a predetermined value of 2
Although a set is provided, other values may be employed, or three or more sets may be provided.

Further, in this embodiment, the appearance frequency table 11 is common to each component. However, by preparing an appearance frequency table for each component, the number of times of calculation of the appearance frequency can be reduced and the speed can be increased.

〔The invention's effect〕

As described above, according to the present invention, the image data compression apparatus calculates the appearance frequency of the gradation change amount of the gradation value in the block for each image component of the color image when encoding the color image. Calculate the statistic of this change,
Since the code amount is distributed to each component based on the statistic, there is an effect that the optimal code amount can be distributed within a predetermined code amount in a short time.

Further, the image processing apparatus according to the second aspect of the present invention can be applied to an image data compression apparatus employing a multi-tone provision type block encoding method, and performs encoding in such an apparatus. In this case, the threshold value and the number of gradations can be quickly determined.

[Brief description of the drawings]

FIG. 1 is a principle diagram of the present invention, FIG. 2 is a block diagram of an embodiment of an image data compression apparatus according to the present invention, FIG. 3 is an explanatory diagram of a GBTC algorithm, and FIG. FIG. 5 is an explanatory diagram of a resolution component. 41: gradation change amount detecting means 42: appearance frequency calculating means 43: statistical amount calculating means 44: threshold value determining means

The continuation of the front page (72) Inventor Shigeru Yoshida 1015 Uedanaka, Nakahara-ku, Kawasaki-shi, Kanagawa Prefecture Within Fujitsu Limited (58) Fields investigated (Int.Cl. ⁶ , DB name) H04N 1/41-1/419 H04N 7/24-7/68 H04N 11/04

Claims (2)

(57) [Claims] 1. A multi-valued image of a color image is divided into blocks each including a plurality of adjacent pixels, and a difference between a maximum gradation value and a minimum gradation value in each block is compared with a threshold value for each block. Means for calculating the number of representative gradations for displaying the image of the block based on the result of comparison by the comparing means; and a reference gradation value calculated based on the number of representative gradations as a reference value for each block. Means for expressing a difference value calculated based on the difference between the representative gradation values, means for calculating a resolution component indicating the distribution of the representative gradation values in each block, and the above-mentioned reference value, difference value and resolution An image data compression apparatus comprising: means for encoding components; and a gradation change amount detecting means for calculating a difference between a maximum gradation value and a minimum gradation value in a block for each of a plurality of components constituting a color image. And the gradation From the gradation change amount obtained by reduction amount detection means,
An appearance frequency calculating means for calculating an appearance frequency of a gradation change amount for each component constituting the color image; and a statistic calculation means for calculating a statistic of the gradation change amount for each component obtained by the appearance frequency calculation means. And comparing a value calculated by performing a predetermined operation on the statistic of each component calculated by the statistic calculation means with a predetermined value, and for each of the components, the number of gradations in a block. And a threshold value determining means for determining a threshold value for determining the threshold value. 2. A multi-valued image of a color image is divided into blocks each including a plurality of adjacent pixels, and a difference between a maximum gradation value and a minimum gradation value in each block and a threshold value are compared for each block. Means, means for calculating the number of representative gradations for displaying the image of the block based on the comparison result by the comparing means, and a representative gradation value calculated based on the representative gradation value, a reference value for each block, Means for representing a difference value calculated based on the difference between the representative gradation values, means for calculating a distribution resolution component indicating the distribution of the representative gradation values in each block, and the reference value, the difference value, and the resolution component. Image data processing apparatus comprising: a gradation change amount detection unit that obtains a difference between a maximum gradation value and a minimum gradation value in a block for each of a plurality of components constituting a color image. , Before From the gradation change amount obtained by the gradation change amount detecting means,
An appearance frequency calculating means for calculating an appearance frequency of a gradation change amount for each component constituting the color image; and a statistic calculation means for calculating a statistic of the gradation change amount for each component obtained by the appearance frequency calculation means. And comparing a value calculated by a predetermined calculation from a statistic of each component calculated by the statistic calculation means with a predetermined value to determine the number of tones in a block for each of the components And a threshold value determining means for determining a threshold value of the image.