JP2002290763A - Image processing method and image processing apparatus - Google Patents

Abstract

(57) Abstract: An image processing method and an image processing apparatus for dividing a color image into a plurality of small areas and extracting a representative color that approximates each small area with high accuracy. An input image is divided into a plurality of small areas. The color data in each small area is classified into the current group, and the target component selection unit 23 selects some components having a large variance from the representative color obtained from each group in descending order. Then, the group dividing unit 24 divides the color data in the target group into a plurality in the selected component direction, and the representative color extracting unit 25 obtains the representative color of each group. By performing such sequential group division and representative color extraction processing until a predetermined number of representative colors is obtained, highly accurate compression approximation of an image is realized.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a method for compressing color data,
The present invention relates to an image processing method and an image processing apparatus for restoring.

[0002]

2. Description of the Related Art A color image has a huge amount of data about colors. Therefore, when displaying or transferring an image, it is common practice to reduce the data amount by approximating the number of colors of the original color data with a smaller number of colors.

As a compression method of such a color image,
For example, a method described in JP-A-61-252792 is known. In this method, the red signal R, the green signal G, and the blue signal B are handled independently. An outline of this technique will be described with reference to FIGS. In an image processing apparatus employing this method, as shown in FIG. 29, input means 2300 is independently provided for each color of image information.
a, 2300b and 2300c are provided. Each color data input through the input units 2300a, 2300b, and 2300c is converted into the pre-encoding units 2301a and 2301a.
1b, 2301c. Pre-stage encoding means 230
Each of 1a, 2301b, and 2301c calculates an average value of color data. The specific block of the color image is divided into two regions based on the average value. Furthermore, two representative values are calculated by averaging the data of the pixels included in each divided area. Through the above processing, each color in the block is divided into two regions, and two representative values are obtained for each color. By the combination of three colors of RGB, a total of 2 to the third power = 8 are obtained. Then, eight representative colors representing the eight divided areas are obtained. The latter-stage encoding unit 2302 selects two colors as representative colors by detecting the appearance frequency of the eight colors in the block, and sets the two colors as representative colors of the color data in this block. The obtained two representative colors and the positional information for approximating the pixel data in the image with the representative colors are sent to the decoding unit 2303 via a communication medium such as the Internet or an interface. The decoding unit 2303 creates a restored image by embedding a representative color in predetermined pixel coordinates, and outputs the restored image to the output unit 2.
It is output by 304a, 2304b, and 2304c.

As another method, a method of obtaining approximate data by performing a principal component analysis of an RGB signal (for example,
Japanese Patent Application Laid-Open No. 1-264092) is known. this is,
A color (principal component color) representative of a specific block is determined in consideration of the correlation of the RGB signals, and the block is divided based on the principal component color, and the color in this block is determined by a predetermined number of colors. This is a method that approximates the following.

[0005]

However, Japanese Patent Application Laid-Open No.
In the technique as described in 1-225792,
Since the RGB signals are handled independently of each other and the correlation of each color is not considered at all, distortion in the restored image is likely to occur, and the S / N ratio (Signal to Noise Rati) for quantitatively evaluating the degree of distortion is considered.
o) was often pointed out that good results were not obtained.

The principal component analysis used in the above-mentioned Japanese Patent Application Laid-Open No. 1-264092 requires a multi-dimensional matrix operation to obtain the correlation between the RGB signals. There is a risk that the scale of the wear will increase. Further, in the case of software processing of a CPU or the like, there is a problem that a long processing time is required.

SUMMARY OF THE INVENTION The present invention has been made in view of such problems in the prior art, and provides an image processing method and an image processing apparatus for approximating color data in a small area with high accuracy and high speed. It is the purpose.

[0008]

In order to achieve the above object, the present invention employs the following means.

In the present invention, the area dividing means divides an input image into a plurality of small areas. The reference color table set has a reference color set prepared in advance, and the color data classifying means determines a predetermined number of color data in the small area based on the color data in the small area and the reference color in the table. Divided into groups. The representative color determining means obtains a representative color of the obtained color data in each group. Then, the target area approximating means approximates the color data of each small area using the obtained predetermined number of representative colors.

When a predetermined number of representative colors are obtained in this way, a predetermined number of groups are obtained based on the color data in the small area and the reference colors in the table. The approximation is fast.

The representative color determining means obtains a representative color of the color data in the group using, for example, the calculation result of the total variance calculating means. The total variance calculating means calculates the total variance from the average value of the color data in each group obtained by the color data classifying means. In this case, approximation accuracy is ensured by obtaining a representative color having a small total variance.

However, with the reference color set prepared in advance, there is a possibility that the approximation accuracy cannot be sufficiently secured. For this reason, a sequential representative color determining means is provided separately from the representative color determining means, a sequential division determination value calculating means calculates a division determination value, and the sequential division determining means calculates a sequential representative color based on the division determination value. The determination unit may determine whether to determine the representative color.

The sequential division judgment value calculation means calculates a division judgment value from the average value of the color data in each group obtained by the color data classification means and the reference color used by the color data classification means. The sequential division determining unit determines whether the representative color is determined by the sequential representative color determining unit based on the division determination value.

When the representative color is determined by the sequential representative color determining means according to the determination of the sequential division determining means, the target area approximating means is obtained by the sequential representative color determining means instead of the representative color determining means. The color data of each small area is approximated using the predetermined number of representative colors.

The sequential representative color determining means divides the target group into a plurality of groups based on a criterion relating to the component of interest selected based on the distribution of the color data in the target group. Is repeated until a predetermined number of representative colors are obtained. Therefore, approximation suitable for the target area is performed, and the accuracy of approximation is improved. Since the process of sequentially obtaining the representative color is performed only at a portion where the approximation accuracy cannot be sufficiently ensured by the reference color set prepared in advance, the high-speed and high-precision approximation is performed as a whole.

Also, it is possible to speed up approximation and increase accuracy without using a reference color. For example, the initial setting means sets an initial group for the small area, classifies all color data in the small area into an initial group, and obtains a representative color of the initial group. The target component selection means selects a plurality of target components of interest at the time of division into the target group based on the distribution of the color data belonging to the target group. The group dividing means divides the target group into a plurality of groups in accordance with the obtained plurality of divided components of interest, and distributes the color data belonging to the target group to the groups obtained after the division. In this case, the representative color extracting unit obtains a representative color of the color data belonging to each group obtained by the group dividing unit. The end determination means determines whether a predetermined number of representative colors have been obtained. The extraction of the representative color is repeated from the selection of the divided component of interest until the predetermined number of representative colors is obtained, and when the end determination unit determines that the predetermined number of representative colors has been obtained, the target area approximation unit outputs The color data of each small area is approximated using a predetermined number of representative colors.

Even when a predetermined number of representative colors are sequentially obtained as described above, a group is divided according to a plurality of components of interest, so that the number of repetitions is reduced accordingly. As a result, the approximation is performed with high accuracy and high speed.

When the focused component selecting means selects a plurality of divided focused components, for example, the component-specific variance calculating means calculates the variance of the representative color of the color data and the components of the color data in the target group. The divided component determining means selects the component of interest in order from the component having the largest variance value obtained by the component-specific calculating means among the components.

Since the group division is performed sequentially on the components having large variance values, the approximation accuracy is high.
However, the user does not always seek a high-precision approximation. For example, a user may require speed over accuracy, or may require only enough accuracy to identify an image, such as a thumbnail image. In this case,
Representative colors obtained sequentially can be used.

For this purpose, the holding judging means judges whether or not a plurality of representative colors set in a predetermined number have been obtained. The step encoding holding means, when it is determined that the representative color is obtained by the holding means, uses the representative color in each step and position information for approximating the color data in the small area using the representative color. Hold. The user image quality selecting means is used by the user to select any one of a plurality of image quality modes. The appropriate representative color reading means calls the representative color and the position information according to the selection mode from the step encoding holding means according to the selection result using the user image quality selecting means.
In this case, the end determining means determines whether or not the maximum predetermined number of representative colors has been obtained, and the target area approximating means determines whether or not the end determining means has determined that the maximum predetermined number of representative colors has been obtained. The color data of each small area is approximated using the representative color obtained by the reading means.

Thus, for example, the thumbnail mode,
Approximation is performed in accordance with an image quality mode specified by a transmission destination user, such as a high-speed mode or a high-definition mode. As a result, high user flexibility is ensured.

Further, when the representative colors are sequentially obtained, a second group dividing means may be provided in order to increase the accuracy and speed of approximation. In this case, the target component selection means does not need to select a plurality of divided target components.

The second group dividing means further divides each group into a plurality of groups based on the color data in each group obtained by the group dividing means,
The color data in each group is allocated to the group obtained after the division.

The representative color extracting means obtains a representative color of the color data belonging to each group obtained by the second group dividing means instead of the group dividing means.

In this case, since each group is further divided into a plurality of groups based on the color data in each group obtained by the group dividing means, the number of repetitions is reduced accordingly. Therefore, the approximation is performed with high accuracy and at high speed. In some cases, the division by the second group dividing unit is performed only once by the group dividing unit.

The second group dividing means includes, for example, an auxiliary dividing reference calculating means and a re-group dividing means. The auxiliary division criterion calculation means calculates the Euclidean distance from the origin of the color data in each group obtained by the group division means. The regrouping means further subdivides the group obtained by the grouping means on the basis of the calculation result of the auxiliary division criterion calculating means and distributes the color data in the target group to the group obtained after the division.

By utilizing the Euclidean distance from the origin of the color data in each group, approximation accuracy is ensured.

Even when the representative colors are sequentially obtained as described above, if the representative colors are sequentially obtained only for some of the small areas, the approximation is performed with high accuracy and high speed as a whole.

For this purpose, the color data statistic calculation means includes:
The statistical distribution of the color data in the small area is obtained. The already-represented color use determining unit compares the statistical distribution of the color data with the statistical distribution obtained from the extracted area from which the representative color has already been extracted, and determines whether to use the representative color of the extracted area. I do. The similar area determining means selects an extracted area having the highest similarity to the small area. The color data classification means divides the color data in the small area into a predetermined number of groups based on the color data in the small area and the representative color in the selected extracted area.
The simple representative extracting means obtains a representative color of the color data belonging to each group obtained by the color data classifying means. If the predetermined representative color use determining means determines that the representative color of the extracted area is to be used, each small area is determined using a predetermined number of representative colors obtained by the simple representative extracting means instead of the representative color determining means. Approximate the color data.

As described above, if there is an extracted area having a similar statistical distribution, a predetermined number of representative colors can be obtained by using the representative colors of the extracted area. As a result, the approximation is performed with high accuracy and high speed as a whole.
When regions having similar statistical distributions are continuous, the approximation speed is significantly increased.

The statistical distribution of the color data in the small area is represented by a histogram, for example.

The extracted area may be set appropriately. For this purpose, the sequential division target determination means
An area in which the representative color is determined by the representative color determining means is selected from the small areas. For the area selected by the sequential division target determination means, representative colors are sequentially obtained. The area selected by the sequential division target determining means is an extracted area.
The close similar area determination means compares the statistics obtained from the extracted area close to the small area not selected by the sequential division target determination means with the statistics of the small area, and determines the most similar area. Select a highly extracted area. In this case, the simplified representative extracting means divides the color data in the target small area into a predetermined number of groups based on the representative colors of the extracted areas selected by the close similar area determining means, and Ask for. Then, the target area approximating means approximates the color data of each small area using the predetermined number of representative colors obtained by the simple representative color extracting means for the small areas not selected by the sequential division target determining means.

As described above, a representative color is sequentially obtained for a predetermined area, and for other areas, a representative color is obtained using a representative color of an adjacent extracted area. Is performed with high precision and at high speed. Note that, in order to suppress variations in accuracy, it is preferable to sequentially and uniformly select areas for obtaining the representative colors from the image.

[0034]

Embodiments of the present invention will be described below with reference to the drawings. In each of the embodiments, the color data represents a color signal of a pixel, and in the RGB color system, represents data including three components of red, green, and blue. In the drawings of the configuration of the image processing apparatus in each embodiment, the same parts are denoted by the same reference numerals.

(First Embodiment) First, the first embodiment of the present invention will be described.
An image processing method and an image processing apparatus according to the embodiment will be described. As shown in FIG. 2, in the image processing apparatus, the area dividing means 21 divides the input image 20 into minute block areas. The initial setting unit 22 performs an initial group setting when the color data in the target block area is sequentially divided into groups. Attention component selection means 23
Determines the target component at the time of division of this group based on the distribution of color data within the group to be divided. The group dividing means 24 divides the target group into a plurality of groups according to the target component, and allocates the color data belonging to the target group to the groups obtained after the division. The representative color extracting unit 25 obtains a representative color of the color data belonging to each group obtained by the group dividing unit 24. The end determination unit 26 determines whether or not a predetermined number of representative colors has been obtained. The target area approximating means 27 approximates the color data in each block area using the obtained predetermined number of representative colors.

The initial setting means 22 includes a distance calculating means 29 for obtaining a Euclidean distance from the origin of the color data belonging to the target block area, and all color data in the area based on the Euclidean distance into a plurality of groups. It comprises an initial group dividing means 30 for classifying, and an initial representative extracting means 31 which uses the average color of the color data belonging to each group obtained by the above means as a representative color of the group.
The target component selecting unit 23 includes a component-specific variance calculating unit 32 for calculating the variance of each component of the color data belonging to the representative color of the color data belonging to the target group, and a variance value obtained by the component-specific variance calculating unit. It is composed of divided component determining means 33 which selects the components in order from the larger components and sets them as the components of interest for division at the time of group division.

The operation of the image processing apparatus and the image processing method according to the first embodiment configured as described above will be described with reference to FIG. As shown in FIG. 1, the input color image 20 is divided into small block areas by the area dividing means 21 (S1). This is based on the fact that the number of color data in a minute area is limited to a certain extent, so that even if expressed with a small amount of color data, the deterioration in accuracy will not be so large. The subsequent representative color extraction processing is executed for each block area. In this processing method, as described below, in order to realize high accuracy of an approximate image, grouping is sequentially performed based on a statistical distribution of color data in a target block area, and a representative is represented by each group. Color extraction is performed.

First, the initial setting means 22 sets an initial group and its representative color when performing the division process sequentially (S2). As a setting method, for example, as shown in FIG. 5A, one initial group is set assuming that all data in the target block belongs to one group, or as shown in FIG. There is a method of obtaining the Euclidean distance from the origin of all the color data in the target block, and classifying and setting it into a plurality of initial groups. The block configuration shown in FIG. 2 is based on the method shown in FIG. The color data cIj (rj, gj,
The Euclidean distance length_k [j] from the origin of bj) (j = 0,..., pxl_k) is calculated. Initial group division means 3
0 divides the initial group into two groups according to (Equation 1). Here, ave_length_k is the target block
Indicates the average value of distance length_k [j] in k, and t_k_pix is k
Indicates the total number of pixels in the box. {CIj} represents the length (norm) of the vector when the color data cIj is regarded as a vector.

[0039]

(Equation 1)

In the initial representative extracting means 31, this group
The average value of each color data belonging to 0 and 1 is obtained as a representative color of this group. The initial group dividing means 3
At 0, the initial group is set to two, but this number is not unique and can be more. In order to divide data in a block into groups with high accuracy, it is preferable to use about two. In addition, as shown in FIG. 5A, one initial group can be set. In this case, the processing of the distance calculating unit 29 and the initial group dividing unit 30 in FIG. 2 can be omitted. At this time, since the number of sequential processes until a predetermined number of representative colors is obtained increases, the processing time does not decrease. However, the accuracy of the representative color extraction and the accuracy of the approximate image obtained by using the final color are ultimately shown in FIG. It becomes higher than the method shown in FIG.

Based on the initial group, the target component selecting means 23, the group dividing means 24, the representative color extracting means 25,
The end determination unit 26 performs a process of sequentially dividing the color data in the target block into groups and extracting a representative color. FIG. 3 schematically shows a specific example of the processing. In this specific example, it is assumed that the division target group S is divided into four.

FIG. 3A shows a target group S, a representative color RIs, and input pixel color data Is [j] in a space formed by the r, g, and b axes. Here, the current division target group S is set to group 1 (S = 1), and the representative color RIs (Rrs, Rgs, Rbs) representing the current division target group S is set to the representative color RI ₁ ^k. And The current representative cluster vector number num is num = 1, and the temporary variable tnum
Is set to tnum = 1.

The component-based variance calculation means 32 calculates Ns pixel color data Is [j] (rs) belonging to the current group S to be divided.
[j], gs [j], bs [j]), as shown in FIG. 3B, the sum (variance) of deviation absolute values for the representative color RIs independently for each of the r, g, and b axes. δrs, δgs, δbs are obtained (S3).

The divided component determining means 33 calculates the sums of the absolute differences δrs, δgs,
δbs are arranged in descending order, and an axis Amax1 having the maximum value and an axis Amax2 having the next largest variance value are selected (S4). FIG.
In the example of (b), the axis Amax1 having the maximum value is the r axis,
The axis Amax2 having the next largest variance is the g-axis.

FIG. 3 (c) is a view similar to FIG. 3 (a) and FIG. 3 (b).
It is viewed from the b axis. As shown in FIG. 3C, the group dividing unit 24 divides the Ns pixel color data Is [j] in the current group S into two groups in the Amax1 direction, and
It is also divided into two groups in the x2 direction (S5). As a result, the group S is divided into a total of four groups.

The processing of group division is performed using Amax1 of the representative color RIs.
Amax1 of the component and each pixel color data Is [j] belonging to group S
The Amax2 component of the representative color RIs and the Amax2 component of each pixel color data Is [j] belonging to the group S are compared, and the comparison is performed based on the magnitude relation. In the example of FIG. 3C, group division is performed as in (Equation 2).

[0047]

(Equation 2)

As shown in FIG. 3D, the representative color extracting means 25 outputs the four groups S, num + 1, num
+2 and num + 3 as the new representative colors RIs (Rrs, Rgs, Rb
s), RInum + 1 (Rrnum + 1, Rgnum + 1, Rbnum + 1), RInum + 2
(Rrnum + 2, Rgnum + 2, Rbnum + 2), RInum + 3 (Rrnum + 3, R
gnum + 3, Rbnum + 3) (S6). And the representative color number n
um is set to num = num + 3.

The representative color extracting means 25 temporarily retains the obtained representative color and positional information indicating which representative color approximates the pixel in the block. This representative color is used for the variance value calculation of each component by the component-specific variance calculation means 32 when dividing again. FIG. 4 shows information held at that time.

If S has not reached tnum and the required representative color has not been extracted (S7), the number S representing the current division target group is set to S = S + 1, and the target group is changed. (S8) The processing from step S3 is repeated.

Num is a preset representative number of colors c_nu
It is determined whether m has been reached (S9), and if not, S = 1 and tnum = num are set. Then, step S2
Is performed. The end determination means 26 performs such an end determination.

The predetermined number c_num extracted by the above processing
, And positional information indicating which of the representative colors approximates the color data of the pixels in the block. Then, using the held information, the target area approximating means 27 at the approximation / restoration destination approximates and restores the pixel data in each block area (S10). This is, as shown in FIG.
i, j) corresponds to the pixel coordinates (i, j) in the target block, and an approximate image is generated by embedding the representative color embedded in the position information coordinates in the corresponding pixel coordinates. Can be obtained. Here, the example has been described in which the division target group S is divided into four.
It is also conceivable to divide only the axis Amax1 having the maximum variance (divide into two). In this case, the accuracy of the extracted representative colors is expected to be higher than that in the case of division into four, and this also leads to higher accuracy of the finally obtained output image. However, since the number of processes for sequential division increases,
Processing speed is slightly slower. Also, the number of groups to be divided is 8
In this case, the processing of the component-specific variance calculation unit 32 and the division component determination unit 33 can be omitted, and high-speed representative color extraction can be performed. However, the extraction accuracy may deteriorate. Therefore, it is conceivable to add a means for controlling the number of groups to be divided according to the purpose. It is also conceivable that the distribution of the color data in the target block area is obtained and the number of divisions of the target group is controlled in accordance with the distribution variation.

As described above, according to the present embodiment, the average value of the color data in the subdivided area and the color direction in which the variance from the average value is large are sequentially divided into a plurality of groups, and each group is represented as a representative. Therefore, it is possible to perform high-precision and high-speed area division and extract a representative color representative of each area.

Note that these processes can be performed by software processing using a central processing unit (CPU), a digital signal processor (DSP), or the like used in a computer or the like according to the image processing method according to the first embodiment of the present invention. It can be realized similarly. Further, these processes can be similarly realized by using a semiconductor chip prepared according to the image processing method according to the first embodiment of the present invention.

(Second Embodiment) Next, a second embodiment of the present invention will be described.
An image processing method and an image processing apparatus according to the embodiment will be described. As shown in FIG. 7, in the image processing apparatus, the second group dividing unit 70
The plurality of groups obtained in step 4 are further divided into a plurality of groups by color data, and the color data in the target group is allocated to the groups obtained after the division.

Then, the second group dividing means 70 calculates an auxiliary division criterion calculating means 71 for calculating the Euclidean distance from the origin of the pixel color data in the group obtained by the group dividing means 24, Originally, the group obtained by the group dividing unit 24 is further divided into groups, and the regrouping unit 72 is configured to distribute the color data in the target group to the group obtained after the division.

The operation of the image processing apparatus and the image processing method according to the second embodiment configured as described above will be described with reference to the flowchart of FIG. FIG.
Schematically shows the state of group division in this embodiment. Similar to the image processing method and the image processing apparatus according to the first embodiment of the present invention, after dividing the input image 20 into a plurality of blocks (S1), the initial setting unit 22
Set some initial groups and their representative colors (S
2). After that, the component-based variance calculation means 32 executes the process shown in FIG.
8 (b) for the group S to be divided as shown in FIG.
As shown in (3), the variance values δrs, δgs, δbs from the representative colors of the color data in the group S are calculated (S3). The division component determination means 33 selects an Amax component having the maximum value of the variance on each of the r, g, and b axes as a division target component (S
21). In the example of FIG. 8B, the r-axis is selected as the component of interest for division. As shown in FIG. 8C, the group dividing unit 24 divides the target group into two along the Amax component, and the Amax component of the representative color and the Amax of each pixel color data Is [j] belonging to the group S. A process is performed to compare the components and divide the two groups according to the magnitude relationship (S22). The groups obtained here are again referred to as group S and group num + 1. Note that num indicates the number of groups before division.

Next, the second group dividing means 70
As shown in (d), the group S and the group num + 1 are each further divided into two groups. At this time, first, the group S and num
The Euclidean distance len from the origin of the pixel color data belonging to +1 is obtained (S23). Then, the regrouping unit 72 divides the group S into the group S and num + 2 based on the average value of the distance of the pixel color data in the group S, and distributes each color data. Similarly, group num + 1 is divided into groups num + 1 and num + 3, and group n
The color data of um + 1 is classified (S24). By doing so, you can reduce the number of times to split the group,
High-speed group division can be performed.

Then, the representative color extracting means 25 finds the average value of the color data in the group obtained at this time and sets it as the representative color of the group (S6). End determination means 26
Determines whether or not a predetermined number of representative colors has been extracted (S7). If not, the target group is changed (S8). The process is repeated. On the other hand, if it is determined that the representative color extraction has been completed, and if the processing of all blocks has been completed (S9), at the time of image approximation,
An approximate image of the input image is created based on the position information when the target area approximating means 27 approximates the obtained representative color and each pixel color data, and an output image 28 is obtained (S10).

As described above, according to the present embodiment, after the average value of the color data in the subdivided region and the color direction having the largest variance from the average value are divided into two groups, By performing a plurality of divisions again based on the distance from the origin of the color data in the group to which the group belongs, the number of times of performing the group division can be reduced.

Note that these processes can be performed by software processing using a central processing unit (CPU) and a digital signal processor (DSP) used in a computer or the like according to the image processing method according to the second embodiment of the present invention. It can be realized similarly. Further, these processes can be similarly realized by using a semiconductor chip prepared according to the image processing method according to the second embodiment of the present invention.

(Third Embodiment) Next, an image processing method and an image processing apparatus according to a third embodiment of the present invention will be described. FIG. 10 shows a configuration of an image processing apparatus according to the third embodiment. In the image processing apparatus shown in FIG. 10, the reference color table set 100 has a prepared reference color set. The color data classification means 101 divides the color data in the small area into a predetermined number of groups based on the color data in the target block area and the reference color set in the reference color table set. The sum variance calculation means 102 calculates the sum of variance from the average value in each group. The representative color determination unit 103 selects one division result from the plurality of group division results obtained by the color data classification unit 101, and obtains a representative color of the in-group color data obtained at that time.

The operation of the image processing apparatus and the image processing method according to the third embodiment configured as described above will be described with reference to the flowchart of FIG.

The input image 20 is divided into a plurality of small block areas by the area dividing means 21 (S1). Then, the representative color is extracted for each block area as in the first embodiment of the present invention. However, in the third embodiment, the process of first setting the initial group in the target block and repeating the group division sequentially is not performed. In the third embodiment, an appropriate set of n reference colors is selected from a table set prepared in advance, and the pixel color data in the target block is classified into n groups based on the reference colors. Then, a representative color representing each group is extracted.

To create the reference color table set 100,
For example, the image processing in the first embodiment is used. In this case, image processing according to the first embodiment is performed in advance on many image samples, and a predetermined number n of sets of representative colors are obtained. The KK colors may be tabulated as KK reference color sets. It is also conceivable to create a set having LL n reference colors by applying a statistical method such as a vector quantization method to the vectors composed of the KK n representative colors. Instead of the statistical method, it is also possible to apply a method such as Kohonen's self-organizing neural network.

In order to prepare a set of representative colors, an average image of characteristic images such as textures, characters, and relatively flat images may be used. In this case, after the average image of each image is generated, the image processing according to the first embodiment is performed on each average image, and a set of representative colors is prepared. Also in this case, the representative color set may be tabulated as the same number of reference color sets, or may be tabulated as a different number of reference color sets by a vector quantization method or the like. Further, it is also possible to use a representative color of a block already processed for the same image for a reference color set without preparing a separate image.

In order to prepare a set of representative colors for each classification target such as a person's skin, a plant, and the sky, a block obtained by subdividing a plurality of images prepared in advance may be used. In this case, each block is classified according to the ratio occupied by each classification target, and representative color extraction processing is performed on the classified blocks. The set of representative colors obtained in this manner may be tabulated as it is as a reference color for the corresponding classification target, or as in other examples,
The data may be tabulated using a vector quantization technique or the like.

The number of representative colors in one set does not need to be uniform n, but can be variable. In that case, KK representative color sets of n stages obtained from many image samples are similarly prepared, and a plurality of reference color sets are obtained from the KK representative color sets in each stage by the above-described method. The reference color table set 10 is obtained by converting the table into the required table
0 is prepared.

The color data classifying means 101 first selects the m-th n reference color sets dR from the reference color table set 100.
It_m (t = 0,..., N-1) is selected (S31). next,
As shown in FIG. 11, k in the block
Of pixel data I_k [j] are classified into n groups (S
32). The classification is, for example, a reference color set dRIt_m (t =
0,..., N-1) and the pixel data I_k [j], and the pixel data I_k [j] is classified into the reference color group t having the shortest distance. Thereafter, an average value in the n groups is obtained (S33). The variance value from the average value in each group is obtained, and the sum δ_m of the variance values from the average value in each group when the m-th reference color set is used is calculated (S34). When this processing is performed for all the reference color sets (S35, S36), the representative color determination unit 103 determines the average value obtained from the reference color set m that minimizes the sum of the variance values as the target block area. It is determined as a representative color obtained for k (S37). Then, the position information for the pixel data in the block at this time is also held, and if the processing of all the blocks is completed (S
9) By using the representative color and the position information at the time of image restoration, the target area approximation unit 27 generates an output image 28 (S10).

By doing so, unlike the image processing method and the image processing apparatus according to the first and second embodiments of the present invention, the color in the target area can be changed at high speed without sequentially performing group division. Data division and representative color extraction are possible. By preparing a plurality of high-precision reference color sets prepared from a large number of image samples in advance, accurate representative color extraction can be performed without being greatly affected by the input image, and the accuracy of the restored approximated image can be improved. Deterioration can be suppressed.

Note that these processes can be performed by software using a central processing unit (CPU) and a digital signal processor (DSP) used in a computer or the like according to the image processing method according to the third embodiment of the present invention. It can be realized similarly. Further, these processes can be similarly realized by using a semiconductor chip created according to the image processing method according to the third embodiment of the present invention.

(Fourth Embodiment) Next, an image processing method and an image processing apparatus according to a fourth embodiment of the present invention will be described. In the image processing apparatus shown in FIG.
A division determination value is calculated from the average value in each group obtained in step (1) and the reference color used by the color data classification means 101. The sequential division determination unit 131 determines whether to select a representative color from the average value obtained here based on the division determination value. The sequential representative color determination unit 132 determines that the sequential division determination unit 131 determines that the sequential division is to be performed.
The color data in the small area is sequentially divided into groups, and a representative color of the color data belonging to each group is obtained.

In the image processing apparatus and image processing method according to the fourth embodiment configured as described above, processing is performed as shown in the flowchart of FIG. Third of the present invention
As in the embodiment, the input image 2
0 is divided into block regions (S1). For the divided block region k, the color data classification means 101 prepares an m-th reference color set dRIt_m (t = 0,.
n-1) to perform group division, and obtain an average value eRIt_m (t = 0,..., n-1) in each group (S31, S31).
32, S33).

[0074]

(Equation 3)

Next, the sequential division judgment value calculation means 130
The sum b_thr of deviations between the m-th reference color set and the average value eRIt_m of each group obtained in 101 as in (Equation 3)
es_m is obtained (S41). This process is performed for all the prepared KK reference color sets (S35, S36),
The sequential division determination means 131 compares the minimum value of the sum b_thres_m with a preset reference value BThreshold (S4).
2). If the value is smaller than the value, the representative color determining unit 103 calculates the total sum b_t of shifts for the KK reference color sets.
The average value eRIt_m in each group obtained by the reference color set mb that obtains the minimum value of hres_m is set as a representative color obtained from the block (S37). Note that the representative color determining means 1
For 03, the representative color can be obtained as follows.
First, the average value eRIt_m of the group t when divided by the m-th reference color set is compared with the nt belonging to the group t.
The sum of the absolute deviation values of the pixel value data It [j] is obtained on the r, g, and b axes, and the mean square value of the results on each axis is calculated. And
This value is summed up for all the groups t to obtain a value (variance from the average value) δ_m, and group division is performed using a reference color set mbd that gives the minimum value of the variance δ_m. The average value eRIt_mb in the group is set as a representative color obtained from the block.

On the other hand, if the minimum value of b_thres_m is larger than the reference value BThreshold preset by the sequential division determination means 131, the sequential representative color determination processing by the sequential representative color determination means 132 is performed (S43).

The sequential representative color determining means 132 includes, for example, the initial setting means 2 according to the first embodiment of the present invention.
2. A target component selecting unit 23, a group dividing unit 24, a representative color extracting unit 25, and an end determining unit 26 are provided. The sequential representative color determining unit 132 performs the same processing as that of the first embodiment. Further, the sequential representative color determining means 132
Secondly, the initial setting means 2 according to the second embodiment of the present invention
2. A target component selecting unit 23, a group dividing unit 24, a second group dividing unit 70, a representative color extracting unit 25, and an end determining unit 26 are provided, and the sequential representative color determining unit 132 performs the same operation as in the second embodiment. May be performed.

When the processing of all blocks is completed (S9),
Finally, an approximate output image 28 is obtained by using the representative color obtained by the representative color determining means 103 or the sequential representative color determining means 132 and its position information in the target area approximating means 27 (S10). .

As described above, in the fourth embodiment, when approximating the pixel color data in the block area obtained from the input image, when the reference color in the prepared table relatively matches. Can easily use the reference color to divide a group and find a representative color in each group, which leads to a reduction in processing. On the other hand, when the reference color at the time of the table is used uniformly as in the third embodiment, when the reference color and the average value in the group based on the reference color are greatly shifted, the representative color is used. Although the extraction accuracy may be reduced, in such a case as in the present invention, representative grouping is performed by performing group division according to the distribution of color data in an actual area as in the first or second embodiment of the present invention. The color extraction accuracy can be improved.

Note that these processes can be performed by software processing using a central processing unit (CPU) and a digital signal processor (DSP) used in a computer or the like according to the image processing method in the fourth embodiment of the present invention. It can be realized similarly. Further, these processes can be similarly realized by using a semiconductor chip prepared according to the image processing method according to the fourth embodiment of the present invention.

(Fifth Embodiment) Next, an image processing method and an image processing apparatus according to a fifth embodiment of the present invention will be described. FIG. 15 illustrates a configuration of an image processing apparatus according to the fifth embodiment, and FIG. 14 illustrates a flowchart of processing in an image processing method according to the fifth embodiment. In the image processing apparatus shown in FIG. 15, the color data statistic calculation unit 150 obtains a statistical distribution of the color data in the target area. The already-represented color use determining unit 151 compares the color data statistic obtained by the color data statistic calculation unit 150 with the statistic obtained from the already-extracted representative color region, and uses the representative color of the processed region. Is determined.
The similar area determination means 152 is the representative color use determination means 15
If it is determined to be used in step 1, the processed area having the highest similarity is selected. The simple representative extracting unit 153 obtains a representative color of the in-group color data obtained when the color data classifying unit 101 performs the group division based on the representative colors selected by the similar region determining unit 152.

The operation of the image processing method and the image processing apparatus according to the fifth embodiment configured as described above will be described with reference to the flowchart of FIG. In the fourth embodiment of the present invention, a comparison is made with a plurality of reference color set tables prepared in advance. By performing comparison with the obtained representative colors, the group division processing is simplified.

The color data statistic calculation means 150 converts the color data I_k [i, j] of k in the target block into a region which has already been processed.
A color data statistic Len for comparison with the color data I_p [i, j] in p is obtained (S51). Examples of the statistics include an average square error between pixel color data corresponding to each other and an average of absolute difference values of histogram frequencies of color data in a block. However, this statistic is not uniquely determined by this value, and many other methods are possible. (Equation 4) represents the mean square error between the pixel color data, and (Equation 5) represents the average of the absolute difference values of the histogram frequencies of the in-block color data.

[0084]

(Equation 4)

[0085]

(Equation 5)

The average square error between the pixel color data is the statistic Le
When used for n, as shown in FIG. 16, color data in the comparison target block corresponding to the color data in the target block is prepared, and the degree of similarity is determined by averaging the square error of both with the number of pixels. . When the average of the absolute difference values of the histogram frequencies of the color data in the block is used as the statistic Len, a histogram of the color data in the target block and a histogram of the color data in the comparison target block are prepared. The degree of similarity is determined by a value obtained by averaging the absolute values with the number of pixels. In the former case, the color data value itself is the reference, and in the latter case, the frequency as the color is the reference. When more accuracy is required, the former reference value is more preferable, but the latter reference value is considered to be sufficient because the number of color data in a minute block is limited.

The color data statistic calculation means 150 obtains a value for measuring the similarity Len of the distribution of the color data in one area selected from the target block area k and the processed area p (S52). ). Already-representative color use determining means 151
Determines whether to use the already extracted representative color based on this value (S53). For this determination, the statistic obtained earlier is calculated for each processed block, and the smallest value of the statistic is compared with a predetermined determination value PThreshold. If it is smaller than that value, the representative color obtained from the processed area is used. At this time, the selection of the processed area and the representative color to be used is performed by the similar area determination unit 152. Here, the processed block area PP having the minimum value of the color data statistics is selected (S5).
4), using the representative color obtained from the block,
As in the embodiment, the color data classification means 101 divides the color data in the target block area into groups. The simple representative extracting unit 153 obtains the average value of the color data belonging to each group obtained by receiving the result of the color data classifying unit 101, and sets the average value as the representative color of the group (S5).
5).

On the other hand, when the already-represented color use determining means 151 determines that the color data statistic between the target block and the processed area is larger than a predetermined determination value PThreshold, the representative color use determination means 151 has been used until that time. There is a possibility that a block area different in color data distribution from the area subjected to the color extraction processing is targeted. Accordingly, in this case, the group is sequentially divided and the representative color of each group is obtained by using the sequential representative color determining means 132 (S43). By doing so, it is possible to suppress a decrease in accuracy in representative color extraction from such an area.

By employing such a processing configuration, unnecessary sequential group division can be avoided as in the fourth image processing method of the present invention, and higher-speed image approximation becomes possible. Moreover, since the processed area of the image itself is used for the table without preparing a table including a reference color set previously obtained from a plurality of images as in the third and fourth image processing methods of the present invention, the sequential processing is performed. There is an effect that it is possible to extract a representative color with higher precision even from an area where no color is extracted.

Note that these processes can be performed by software processing using a central processing unit (CPU) and a digital signal processor (DSP) used in a computer or the like according to the image processing method according to the fifth embodiment of the present invention. It can be realized similarly. Further, these processes can be similarly realized by using a semiconductor chip prepared according to the image processing method according to the fifth embodiment of the present invention.

(Sixth Embodiment) Next, an image processing method and an image processing apparatus according to a sixth embodiment of the present invention will be described. FIG. 19 illustrates a configuration of an image processing apparatus according to the sixth embodiment. FIG. 18 is a flowchart illustrating an image processing method according to the sixth embodiment.

In the image processing apparatus shown in FIG. 19, the sequential division target determining means 180 selects a plurality of areas to be subjected to the sequential division processing from the plurality of block areas subdivided by the area dividing means 21. Based on the color data statistic obtained by the color data statistic calculation unit 150, the proximity similar region determination unit 181 determines a plurality of regions that are not selected by the sequential division target determination unit 180 based on the color data statistic. , A processed area having the highest similarity is selected from the sequentially divided target areas.

In this embodiment, first of all, a plurality of block areas obtained by blocking the entire image are sequentially represented by the representative color, as in the first or second embodiment of the present invention. An extraction process is performed. Thereafter, when extracting a representative color from the remaining block area, a plurality of blocks that are close to the target block are selected from the blocks that have been subjected to the sequential representative color extraction processing. And
The block area most similar to the color data distribution in the target block area is selected from the adjacent sequential processed block areas, and the color data in the target block area is divided into groups by the representative color at that time. , Representative color extraction is performed.

An operation and an image processing method of the image processing apparatus according to the sixth embodiment having the above-described configuration will be described with reference to the flowchart of FIG. First, the input image 20 is divided into a plurality of minute regions by the region dividing means 21 (S1). Next, the sequential division target determining means 180 determines an area for performing the sequential representative color extraction processing as described in the first or second embodiment of the present invention from the plurality of block s areas (S61). In this case, any selection is possible. However, it is preferable to select as uniform as possible over the entire area of the image in order to maintain the representative color extraction accuracy in an area where sequential representative color extraction is not performed. Therefore, a number is assigned to the area divided by the area dividing means 21 from the top of the image, and the odd number is sequentially assigned to the division target determining means 18.
0 is selected as the target area. Sequential division target determination means 18
0 for the area selected in step 0
32 is performed (S43,
S62). Here, when the method described in the first embodiment is used, the sequential representative color determining unit 132 includes the initial setting unit 22, the target component selecting unit 23, the group dividing unit 24, the representative color extracting unit 25, An end determination unit 26 is provided, and the same processing as in the first embodiment is executed. When the method described in the second embodiment is used, a second group dividing unit 70 is further added to the sequential representative color determining unit 132.

On the other hand, for the block area not selected by the sequential division target determining means 180, the statistic indicating the color data distribution in the block is calculated by the color data statistic means 150 (S63, 64). As this value, the amount described in the fifth embodiment (FIGS. 16 and 17) can be used, or another reference amount may be used. Then, the proximity similar area determination unit 181 selects an appropriate processed block area from the statistics calculated from the target block area and the block area adjacent to the block and subjected to the sequential representative color extraction processing (S65). The color data classifying unit 101 performs group division on the color data in the target block area based on the selected representative color. The simple representative extracting unit 153 sets the average value of the color data belonging to each group as the representative color of the group (S66). After the processing of the block area not selected by the sequential division target determination unit 180 is completed (S67), the image restoration is the same as in the previous embodiments, and thus the description is omitted.

As described above, by adopting such processing and configuration, it is not necessary to prepare a table including a reference color set previously obtained from a plurality of images. Further, compared to the fifth embodiment, there is an effect that the trouble of searching for a processed area to be a table can be reduced.

Note that these processes are similarly realized by software processing using a central processing unit (CPU) and a digital signal processor (DSP) used in a computer or the like in accordance with the image processing method in the sixth embodiment. can do. Further, these processes can be similarly realized by using a semiconductor chip prepared according to the image processing method according to the sixth embodiment of the present invention.

(Seventh Embodiment) An image processing method and an image processing apparatus according to a seventh embodiment of the present invention will be described. FIG. 21 shows the configuration of the image processing apparatus according to the seventh embodiment, and FIG. 20 shows a flowchart for explaining the image processing apparatus according to the seventh embodiment.

In the image processing apparatus shown in FIG. 21, the holding judgment means 200 judges whether or not the obtained current division number satisfies the holding division number. The stage encoding holding unit 201 holds the representative color at this stage and the position information when approximating the inside of the small area with this representative color, when the holding determination unit 200 determines to hold. The maximum end determination means 202 determines whether the maximum number of obtained representative colors has been obtained, and if not, the processing moves to the division criterion calculation means. The user image quality selection unit 203 selects an image quality mode when an image is approximated by the user.
The appropriate representative color reading means 204 is provided by the maximum end determining means 20.
If it is determined that the processing has been completed in step 2, the step encoding / holding unit 201 calls the representative color and position information corresponding to the user selection mode according to the result of the user image quality selecting unit 203. In the seventh embodiment, when a representative color is extracted, the maximum number of representative colors and the number of representative colors for each stage up to that are set, and the representative color for each of the set stages is set. Position information is held. Then, in accordance with the mode specified by the user, image approximation is immediately performed from the stored data using the number of representative colors and position information corresponding to the mode.

The operation of the image processing method and the image processing apparatus according to the seventh embodiment will be described with reference to the flowchart of FIG. First, the input image 20 is divided into blocks (S1), and the inside of each block area is sequentially divided into groups.
The process of extracting the representative color is the same as in the first embodiment (S2 to S6). What is different here is that not only the finally obtained representative color is stored, but also the representative color and position information at several stages up to that point are stored. In this case, for example, when the number of steps up to the final representative color number is three, the step encoding holding unit 201
The data held in is as shown in FIG. The retention judging means 200 checks whether or not the set number of representative colors has been obtained until the maximum number of representative colors is reached (S71), and when the number of representative colors at that stage is reached. Is stored in the stage encoding storage unit 201 (S7).
2).

On the other hand, when the maximum number of representative colors set by the maximum end determination means 202 has been obtained for all blocks (S73, S8, S9), all held information is transmitted to the user. As the image quality mode to be handled, for example, a high-speed mode that does not care much about the image quality, a high-definition mode, an intermediate mode, and the like are prepared, and the user selects an image quality mode from these by the user image quality selection unit 203 (S
74). According to the selection result, for example, in the case of the high-speed mode, as few representative colors as possible may be used, so that the number of representative colors at the lowest level is selected and image restoration is performed (S
75). On the other hand, in the high definition mode, the maximum number of representative colors is selected and image restoration is performed. The appropriate representative color reading means 204 takes charge of the reading and the determination of the representative color.

Although the qualitative display mode has been described here, it is also possible to directly select the approximate number of representative colors. Also, when all of the images have been completed by the maximum end determination unit 202, the image data is sent to the user, and based on the selection by the user image quality selection unit 203 at that time, the appropriate representative color reading unit 2
Although the description has been made so that the number of representative colors is selected at 04, the present invention is not limited to this. Step encoding holding means 201
May be transmitted to the user at all times, and while the user observes the image restored at the time of transmission, it may be possible to request display in a further representative number of colors. In addition, if the maximum number of representative colors is not reached in the current target block but the current image quality is sufficient, it may be configured to instruct to stop the extraction of the representative colors in the block.

[0103] The image thus compressed is usually restored once based on the data sent to the user, and then image processing such as filtering is performed again according to the desired image quality and resolution. Many. This is especially true when dealing with JPEG-compressed image data. However, since the representative colors in multiple stages and the position information when performing the region approximation with the representative colors are held as in this embodiment, the instructions of the plurality of image quality modes are not clear beforehand at the time of image encoding. (For example, image distribution on the Internet). further,
For such a restored image of the irreversibly compressed image,
Performing a filtering process or the like may increase the distortion generated during compression, but here, since all data obtained from the original image is used, such an influence can be suppressed.

Note that these processes can be performed by software processing using a central processing unit (CPU) and a digital signal processor (DSP) used in a computer or the like according to the image processing method in the seventh embodiment of the present invention. It can be realized similarly. Further, these processes can be similarly realized by using a semiconductor chip created according to the image processing method in the seventh embodiment of the present invention.

Also, in the first to seventh embodiments of the present invention, when finely dividing, each block size does not need to be fixed, and may be variable according to the image. Also, the number of representative colors extracted from each block area does not need to be fixed, and even if the number is variable, the contents described in the present invention similarly hold. The image processing method and the image processing apparatus according to the first to seventh embodiments of the present invention are applied to a device that exchanges images in real time. Such applications include, for example, image reading devices such as mobile scanners, and systems via network cameras such as surveillance cameras, TV conference systems, and telemedicine monitoring systems. Further, the image processing method and the image processing apparatus according to the first to seventh embodiments of the present invention can be used for high-speed display using a fixed number of colors on a mobile terminal such as a mobile phone or a PDA, or from a personal computer. It can also be used for fixed length compression of print image data sent to a printing device such as an ink jet printer or a color laser printer.

Incidentally, the image processing for approximating color data using representative colors as described above is based on human visual characteristics. In consideration of human visual characteristics, the size of a block and the number of representative colors are important for securing approximation accuracy.

An appropriate block size and the number of representative colors are as follows.
Varies by image. For example, if the relationship between the resolution of the image and the size of the block is not appropriate, the accuracy of the approximation deteriorates and the image quality of the restored image deteriorates. When an image is supplied to a personal computer from an image input device such as a scanner, the resolution of the image may be converted to another resolution on the personal computer after approximation is performed on the image input device side. . In such a case, the image quality of the image output on the personal computer side may be degraded. Here, the resolution conversion includes one accompanying the enlargement or reduction of the image.

For this reason, it is preferable that the size of the block is determined according to the resolution after the image is converted to a predetermined resolution. When the predetermined resolution is represented by the conversion ratio between before and after the resolution conversion, the block size may be changed according to the setting of the resolution conversion ratio.
Visual degradation can be suppressed by targeting large blocks when the resolution is high and targeting small blocks when the resolution is low. Such image processing will be described with reference to FIGS. 23, 24, and 25.

In the image processing apparatus shown in FIG. 23, a resolution conversion circuit 1002 is used as a resolution conversion means for converting a color image into a predetermined resolution. For example, the resolution conversion circuit 1002 converts the image fetched from the input circuit 1003 from a high resolution to a low resolution. A resolution setting circuit 1004 performs resolution setting by a signal 1401 according to a resolution conversion circuit 1002 and a representative color extraction circuit 1001.
Give to. The representative color extracting circuit 1 includes, as area determining means for determining the size of a small area from which a representative color is to be extracted in accordance with a predetermined resolution, and representative color extracting means for extracting a plurality of representative colors from the determined small area. Used.

For example, if the resolution setting is changed from 400 DPI to 200 DPI, as shown in FIG. 24, the resolution conversion circuit 2 converts adjacent pixels into the same value in units of two pixels. The representative color extraction circuit 1 sets the block size to half according to the setting of the resolution. At 400 DPI, the block size is 4 pixels x 4
In the case of pixels, the size of the block is determined to be 2 pixels × 2 pixels at 200 DPI. As a result, the processing accuracy is increased as much as the change in the document density becomes coarse, and image quality deterioration can be suppressed. The representative color extraction circuit 1 extracts the representative color of the determined block, for example, as described above in each embodiment, and
7 (hereinafter referred to as I / F 1007).

Here, the resolution conversion processing will be described in detail with reference to FIG. As shown in FIG.
When changing the DPI to 200 DPI, change the value of either the signal (data) 300 or the signal (data) 301 to
Copy to adjacent pixels. For example, the signal 301 is converted to the signal 302
Is copied to the signal 303. With this, 400 DPI
The black and white pattern disappears due to a decrease in resolution.
In the actual processing, since the signal 302 and the signal 303 have the same value, the thinning processing is performed so as to leave one of them. This is a reduction process. That is, the resolution conversion and the enlargement / reduction are handled as processing (or the same processing) that is close in terms of the circuit. Therefore, the resolution conversion is often used when the number of pixels does not change before and after the processing, but here, resolution conversion includes enlargement and reduction accompanying the change in the number of pixels.

When the enlargement processing or the reduction processing is performed,
What is necessary is just to set the magnification instead of the resolution. In this case, the resolution conversion circuit 1002 is used as scaling processing means for enlarging or reducing an image in a predetermined case. The resolution setting circuit 1004 sets the magnification. According to this magnification, the resolution conversion circuit 1002
The number of pixels is increased or decreased by using a pixel interpolation method, a nearest neighbor method, a bilinear method, a bicubic method, an orthogonal transformation method (DCT transformation method, wavelet transformation method), and the like, and enlargement processing and reduction processing are executed.

Referring to FIG. 25, a description will be given of a case where a document image is subjected to enlargement processing and reduction processing. For example, as shown in FIG. 25C, when one pixel is inserted between adjacent pixels of an original image of 3 pixels × 3 pixels to enlarge the original image twice, the number of pixels becomes 9 to 2
Increase to 5. In this case, as the number of pixels increases,
Even if a representative color is extracted from a larger block, image deterioration is small. Therefore, the representative color extraction circuit 1001 increases the size of the block according to the enlargement ratio. Since the representative colors are extracted from the larger area, the compression ratio can be increased. On the other hand, as shown in FIG. 25A, in the case of the reduction processing, the number of pixels decreases. Therefore, since the spatial frequency of the pattern on the original image is increased, processing with a higher resolution is required. Therefore, it is desirable to set the size of the block smaller. That is, the representative color extraction circuit 1001 sets the block size to be small according to the reduction ratio. As a result, a pattern having a high spatial frequency can be held, so that image quality can be improved.

As described above, by changing the size of the region from which the representative color is extracted according to the resolution, even if there is a high-frequency pattern, it is possible to maintain the high-frequency pattern and suppress the image quality deterioration. Further, the compression rate can be improved while maintaining the image quality.

For example, a photo original, a text original, a character
The appropriate resolution differs depending on the characteristics of the original such as a photographic original. Therefore, it is possible to classify an appropriate block size and the number of representative colors according to the characteristics of the document. When the feature of the document is set as the processing mode, the representative color extraction circuit 1
In the case of 001, the size of the block or the number of representative colors may be determined according to the processing mode representing the characteristics of the document instead of the resolution.

In the image processing apparatus shown in FIG. 26, a processing mode setting circuit 1005 is used as mode setting means for setting a processing mode according to a target document.
The processing mode setting circuit 1005 gives the processing mode to the representative color extraction circuit 1001.

For example, as shown in FIG. 27, it is assumed that “photo mode” and “text mode” are prepared as processing modes. In this example, each processing mode is subdivided into two setting modes according to the compression ratio. Each setting mode is associated with an appropriate block size and the number of representative colors. The representative color extraction circuit 1001 determines the size of a block and the number of representative colors using, for example, a table as shown in FIG. If the setting mode set by the processing mode setting circuit 1005 is 1,
The size of the block is 4 pixels x 4 pixels, and the number of representative colors is 2
Is determined.

As a result, the required processing resolution is automatically secured for each document, and high quality image quality can be obtained.

As described above, the image processing device that determines the size of a block can be realized not only by dedicated hardware but also by software processing by a CPU or a DSP.

When the processing is realized by software processing, a computer having a CPU and a DSP performs a step of converting an image to a predetermined resolution, and, in accordance with the predetermined resolution, a size of a small area for extracting a representative color. Is determined, and then a step of extracting a plurality of representative colors from the determined small area is performed.

If the image is to be enlarged or reduced to a predetermined magnification, after the step of enlarging or reducing the image to the predetermined magnification is performed, the computer sets the image according to the predetermined magnification.
The step of determining the size of the small region from which the representative color is extracted is performed.

Further, when using the processing mode,
After the step of setting the processing mode according to the target document is performed, the computer performs the step of determining the size of the small region from which the representative color is extracted or the number of representative colors according to the processing mode.

In the example shown in FIG. 27, the "text mode" and the "photo mode" are prepared as the processing modes, but the present invention is not limited to these modes. For example, another processing mode such as a “text / photo mode” of a mixed image, a “map mode”, or an “enlargement / reduction mode” may be prepared.

[0124]

As described above, according to the present invention, a predetermined number of groups can be obtained based on the color data in the small area and the reference colors in the reference color set table. As a result, the approximation of the color data in the small area is performed with high accuracy and at high speed. Further, in a region where an appropriate reference color set table is not prepared, if representative colors are obtained by sequential group division, it is possible to perform high-speed approximation of the entire image while suppressing deterioration in accuracy.

When a plurality of divided components of interest are selected and group division is performed based on the criterion relating to the selected plurality of divided components of interest, the number of repetitions is reduced even if group division is performed sequentially. Therefore, the approximation is performed with high accuracy and high speed.

Also, after the group division is performed based on the criterion relating to the selected divided component of interest, each group is further divided into a plurality of groups based on the color data in each group. The approximation is performed with high accuracy and high speed because the number of times of performing the division is reduced.

Further, even when the group division is performed sequentially, if the representative color of the small area is obtained by using the representative color of the extracted area from which the representative color has already been extracted, the group division is performed sequentially. Is only a part of the small area, and the approximation is performed with high accuracy and high speed as the whole image.

[Brief description of the drawings]

FIG. 1 is a flowchart for explaining an image processing method according to a first embodiment of the present invention;

FIG. 2 is a block diagram illustrating a configuration of an image processing apparatus according to the first embodiment of the present invention.

FIG. 3 is a conceptual diagram schematically illustrating processing by a representative color extracting unit of the image processing apparatus according to the first embodiment of the present invention.

FIG. 4 is a conceptual diagram showing information necessary for approximating each small area in the image processing apparatus according to the first embodiment of the present invention.

FIG. 5 is a conceptual diagram schematically showing processing by an initial setting unit of the image processing apparatus according to the first embodiment of the present invention.

FIG. 6 is a flowchart illustrating an image processing method according to a second embodiment of the present invention.

FIG. 7 is a block diagram illustrating a configuration of an image processing apparatus according to a second embodiment of the present invention.

FIG. 8 is a conceptual diagram schematically showing processing by a second group dividing unit of the image processing apparatus according to the second embodiment of the present invention.

FIG. 9 is a flowchart illustrating an image processing method according to a third embodiment of the present invention.

FIG. 10 is a block diagram illustrating a configuration of an image processing apparatus according to a third embodiment of the present invention.

FIG. 11 is a conceptual diagram schematically showing processing by a color data classification unit of an image processing apparatus according to a third embodiment of the present invention.

FIG. 12 is a flowchart illustrating an image processing method according to a fourth embodiment of the present invention.

FIG. 13 is a block diagram illustrating a configuration of an image processing apparatus according to a fourth embodiment of the present invention.

FIG. 14 is a flowchart illustrating an image processing method according to a fifth embodiment of the present invention.

FIG. 15 is a block diagram illustrating a configuration of an image processing apparatus according to a fifth embodiment of the present invention.

FIG. 16 is a diagram illustrating an example of color data statistics handled by an image processing apparatus according to a fifth embodiment of the present invention.

FIG. 17 is a diagram illustrating an example of color data statistics handled by the image processing apparatus according to the fifth embodiment of the present invention.

FIG. 18 is a flowchart illustrating an image processing method according to a sixth embodiment of the present invention.

FIG. 19 is a block diagram illustrating a configuration of an image processing apparatus according to a sixth embodiment of the present invention.

FIG. 20 is a flowchart illustrating an image processing method according to a seventh embodiment of the present invention.

FIG. 21 is a block diagram illustrating a configuration of an image processing apparatus according to a seventh embodiment of the present invention.

FIG. 22 is a conceptual diagram showing held information held by a stage encoding holding unit of the image processing apparatus according to the seventh embodiment of the present invention.

FIG. 23 is a block diagram illustrating a simple configuration of an image processing apparatus including a resolution conversion circuit.

FIG. 24 is a diagram illustrating a resolution conversion process.

FIG. 25 is a diagram for explaining enlargement / reduction processing.

FIG. 26 is a block diagram illustrating a simple configuration of an image processing apparatus including a mode setting unit.

FIG. 27 is a diagram for explaining a processing mode;

And FIG. 28 is a flowchart illustrating a conventional image processing method.

FIG. 29 is a block diagram illustrating a configuration of a conventional image processing apparatus.

[Explanation of symbols]

 Reference Signs List 20 input image 21 area dividing means 22 initial setting means 23 target component selecting means 24 group dividing means 25 representative color extracting means 26 end determining means 27 target area approximating means 28 output image 29 distance calculating means 30 initial group dividing means 31 initial representative extracting Means 32 Component-specific variance calculation means 33 Divided component determination means 70 Second group division means 71 Auxiliary division criterion calculation means 72 Regroup division means 100 Reference color table set 101 Color data classification means 102 Total variance calculation means 103 Representative color determination means 130 Sequential division judgment value calculation means 131 Sequential division judgment means 132 Sequential representative color determination means 150 Color data statistic calculation means 151 Existing representative color use judgment means 152 Similar area determination means 153 Simple representative extraction means 180 Sequential division target determination means 181 Proximity Similar Area determining means 200 Holding determining means 201 Step coding holding means 202 Maximum end determining means 203 User image quality selecting means 204 Appropriate representative color reading means 2300a Input means r 2300b Input means g 2300c Input means b 2301a Previous coding means r 2301b Previous coding Encoding means g 2301c pre-stage encoding means b 2302 post-stage encoding means 2303 decoding means 2304a output means r 2304b output means g 2304c output means b

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Yasuhiro Kuwahara 1006 Kazuma Kadoma, Osaka Prefecture Matsushita Electric Industrial Co., Ltd. 72) Inventor Yusuke Monobe 1006 Kadoma Kadoma, Kadoma-shi, Osaka Matsushita Electric Industrial Co., Ltd. (72) Inventor Hirotaka Oku 1006 Kadoma Kadoma, Kadoma-shi Osaka Pref. CA12 CA16 CB01 CB08 CB12 CB16 CC01 CE17 CH07 CH08 5C077 LL17 MP08 PP32 PP37 PP46 PP68 PQ12 PQ18 PQ19 PQ23 5C079 HB01 LA26 LB00 MA04 MA11 NA10 5L096 AA02 FA15 FA32 FA33 GA19 JA11

Claims (24)

[Claims] 1. An input image is divided into a plurality of small areas, and color data in the small areas is divided into a predetermined number of groups based on the color data in the small areas and reference colors in a table prepared in advance. An image processing method of performing a representative color extraction process on each of the obtained groups and approximating color data of a small area using the obtained predetermined number of representative colors. 2. After dividing an input image into a plurality of small areas, a reference color is determined based on color data in the small area and a judgment value obtained from a reference color in a table prepared in advance. It is determined whether or not to divide the color data in the small area into a predetermined number of groups based on the above.If the color data in the small area is not divided into the predetermined number of groups based on the reference color according to the above determination, Then, the component of interest is selected from the distribution of the color data in the small area, the small area is divided into a plurality of groups based on a predetermined criterion, and the representative color is extracted for each of the obtained groups. 2. The method according to claim 1, wherein a predetermined number of representative colors approximating the color data of the small area are obtained by repeatedly performing a process from the selection of the divided target component to the extraction of the representative color until a predetermined number of the representative colors are obtained. Image processing method. 3. An input color image is divided into a plurality of small regions, a plurality of divided target components are selected from a distribution of color data in the small regions, and a reference is made based on the selected plurality of divided target components. Then, the color data in the small area is divided into a plurality of groups, a representative color is extracted for each of the obtained groups, and until the predetermined number of the representative colors is obtained, the selection of the divided target component is performed. An image processing method in which the processing up to the extraction of the representative color of the group is repeated, and the color data of the small area is approximated using the predetermined number of representative colors obtained above. 4. A representative color representing a group is an average value of color data in the group, and a plurality of divided target components are selected in order from a component having a large variance with respect to the average value among the components of the color data. The image processing method according to claim 3. 5. The method according to claim 1, wherein the predetermined number is set in a plurality of stages, and a process from the selection of the divided target component to the extraction of the representative color of each group is repeatedly performed until the maximum predetermined number of the representative colors is obtained. Using the representative color of the stage and the information for approximating the color data in the small area using the representative color, when approximating the color data of each small area using the obtained representative color, 4. The image processing method according to claim 3, wherein when the user selects any one of the plurality of image quality modes, color data of a small area is approximated by selectively using representative colors in each stage according to the selected image quality mode. . 6. An input image is divided into a plurality of small regions, a component of interest to be divided is selected from the distribution of color data in the small region, and a component within the small region is selected based on a criterion for the selected component of interest. Is divided into a plurality of groups, each group is further divided into a plurality of groups based on the color data in each group, and a representative color extraction process is performed for each of the obtained groups. Until a predetermined number of representative colors are obtained, the processes from the selection of the divided target component to the extraction of the representative color of each group are repeated, and the color data of the small area is approximated using the obtained predetermined number of representative colors. Image processing method. 7. Among the components of the color data in the group,
7. The component having the largest variance with respect to the average value of the color data is selected as the component of interest for division, and each group is further divided into a plurality of groups based on the Euclidean distance value from the origin of each color data in the small area. Image processing method. 8. The color data distribution in the target small area is compared with the distribution in the extracted area in which the representative color has already been extracted, and the representative color of each group is selected from the selection of the component of interest for the target small area. It is determined whether or not to perform the sequential representative color extraction processing that repeats the processing up to the extraction of the color data distribution within the target small area if the sequential representative color extraction processing is not performed for the target small area according to the above determination. The color data in the small area is divided into a plurality of groups based on the representative colors representative of the selected extracted areas, and the process of extracting the representative colors of each of the obtained groups is performed. 7. The image processing method according to claim 3, wherein color data approximation of a small area is performed. 9. The image processing method according to claim 8, wherein the determination is made based on a similarity between a histogram of the color data in the target small area and a histogram of the color data in the extracted area. 10. A sequential representative color extraction process that repeats the process from the selection of the component of interest to the extraction of the representative color until a predetermined number of representative colors is obtained for only the small regions selected on the basis of a preset reference. For the small area where the selection is not performed, the color data in the small area is divided into a plurality of groups based on the representative colors of the extracted areas where the representative colors have already been extracted, and the obtained 7. The image processing method according to claim 3, wherein a representative color of each group is extracted, and color data approximation of a small area is performed by the obtained representative color. 11. The one to which all color data in a small area belongs
7. The image processing method according to 3 or 6, wherein one group is set as an initial group, and the initial group is divided into a plurality of groups based on a criterion regarding a component of interest. 12. A plurality of groups determined based on a Euclidean distance from the origin of each color data in a small area are set as an initial group, and the initial group is divided into a plurality of groups based on a criterion regarding a component of interest. Claim 3 or 6
The image processing method described in the above. 13. An area dividing means for dividing an input image into a plurality of small areas, a reference color table set having a reference color set prepared in advance, color data in the small area and reference in the table. Color data classifying means for dividing the color data in the subdivision into a predetermined number of groups based on the color; representative color determining means for obtaining the representative color of the color data in each of the obtained groups; An image processing apparatus comprising: target area approximating means for approximating color data of each small area using a number of representative colors. 14. A total variance calculating means for calculating a total variance from an average value of color data in each group obtained by said color data classifying means, wherein said representative color determining means comprises said total variance calculating means. 14. The image processing apparatus according to claim 13, wherein a representative color of the color data in the group is obtained using the calculation result of (1). 15. A sequential division judgment value calculation means for calculating a division judgment value from an average value of color data in each group obtained by the color data classification means and a reference color used by the color data classification means, The process of dividing the target group into a plurality of groups based on the criterion regarding the component of interest selected based on the distribution of the color data within the group and determining the representative color of each obtained group A sequential representative color determining unit that repeats until a predetermined number is obtained, and a sequential division determining unit that determines whether to determine a representative color by the sequential representative color determining unit based on the division determination value, When the representative color is determined by the sequential representative color determining means according to the determination of the sequential division determining means, the target area approximating means determines the position obtained by the sequential representative color determining means. 14. The image processing apparatus according to claim 13, wherein color data of each small area is approximated using a constant representative color. 16. An area dividing means for dividing an input color image into a plurality of small areas, an initial group is set for the small areas, and all color data in the small areas are classified into an initial group. Initial setting means for obtaining a representative color of the initial group; and focused component selecting means for selecting a plurality of divided focused components to be focused upon when the target group is divided based on the distribution of color data belonging to the target group. According to the plurality of divided target components obtained above, the target group is divided into a plurality of groups, and the color data belonging to the target group is divided into the groups obtained after the division. Representative color extracting means for obtaining a representative color of color data belonging to each group, and an end determining means for determining whether a predetermined number of the representative colors have been obtained. Selection of the target component of interest by the component-of-interest selection means, group division by the group-division means, and extraction of the representative color by the representative-color extraction means are repeated until a predetermined number of the representative colors are obtained. An image processing apparatus comprising: target area approximation means for approximating color data of each small area using the obtained predetermined number of representative colors when it is determined that a predetermined number of the representative colors have been obtained. 17. The component-of-interest selection means includes: component-specific variance calculating means for calculating a variance of a representative color of color data in a target group and each component of the color data; and component-specific variance calculating means of each component. 17. The image processing apparatus according to claim 16, further comprising: a split component determining unit that determines a component of interest in order from a component having a large variance value. 18. A storage judging means for judging whether or not the representative colors of a plurality of stages set to the predetermined number have been obtained, and each stage when the holding judging unit judges that the representative colors have been obtained. And a stage encoding holding unit for holding position information for approximating the color data in the small area using the representative color and the representative color, and a user selects one of the image quality modes from among the plurality of image quality modes. And a suitable representative color reading means for calling the representative color and the position information according to the selection mode from the step encoding holding means in accordance with the selection result using the user image quality selecting means. The end determination means determines whether the maximum number of the representative colors has been obtained. The target area approximation means determines whether the maximum number of the representative colors has been obtained by the end determination means. The image processing apparatus according to claim 16 wherein approximating the color data of each small region with the representative colors obtained by the appropriate representative color reading means. 19. An area dividing means for dividing an input image into a plurality of small areas, an initial group is set for the small areas, and all color data in the small areas are classified into the initial groups. Initial setting means for obtaining a representative color of an initial group; and focused component selecting means for selecting a focused component of interest when dividing the target group based on the distribution of color data belonging to the target group. In accordance with the obtained divided component of interest, the target group is divided into a plurality of groups, and the color data belonging to the target group is divided into groups obtained after the division. Each group is further divided into multiple groups based on the color data of, and the color data in each group is divided into groups obtained after the division. Second group dividing means for dividing, representative color extracting means for obtaining a representative color of the color data belonging to each group obtained by the second group dividing means, and determining whether a predetermined number of the representative colors have been obtained. Until the predetermined number of representative colors are obtained, selection of the component of interest by the component selection unit, group division by the group division unit and the second group division unit, and representative color extraction by the representative color extraction unit. If the extraction is repeated and the end determination means determines that the predetermined number of representative colors have been obtained, the target area approximation means for approximating the color data of each small area using the obtained predetermined number of representative colors An image processing apparatus comprising: 20. An auxiliary division criterion calculating means for calculating a Euclidean distance from an origin of color data in each group obtained by the group dividing means, and a second group dividing means, 20. A regrouping means for further subdividing the group obtained by the grouping means on the basis of the calculation result and allocating the color data in the target group to the group obtained after the division. Image processing device. 21. A color data statistic calculation means for calculating a statistical distribution of color data in the small area, and a statistical distribution of the color data obtained from an extracted area in which a representative color has already been extracted. A representative color use determining unit that determines whether to use the representative color of the extracted region in comparison with the distribution; a similar region determining unit that selects an extracted region having the highest similarity to the small region; Color data classification means for dividing the color data in the small area into a predetermined number of groups based on the color data in the small area and the representative color in the selected extracted area; Simplified representative extraction means for obtaining a representative color of the color data belonging to each group, wherein the target area approximating means is used when the representative color use determining means determines that the representative color of the extracted area is to be used. Is Serial image processing apparatus according to claim 16 or 19, wherein using a predetermined number of representative colors obtained by the simplified representative color extracting means for approximating the color data of each small area. 22. A sequential division target determining means for selecting an area of the small area in which the representative color is determined by the representative color determining means, and a color data statistic calculation for obtaining a statistic of color data in the small area. Means and a statistic of an extracted area in which a representative color has already been extracted and which is obtained from an extracted area close to a small area which is not selected by the sequential division target determination means, and a statistic of the small area. Means for selecting an extracted area having the highest similarity, and color data in the target small area based on the representative color of the extracted area selected by the close similar area determining means. Simple representative extraction means for dividing into a predetermined number of groups to obtain representative colors belonging to each group, wherein the target area approximating means performs the simple substitution for small areas not selected by the successive division target determining means. The image processing apparatus according to claim 16 or 19, wherein using a predetermined number of representative colors obtained by the extraction means for approximating the color data of each small area. 23. A distance setting means for calculating a Euclidean distance from an origin of color data in a small area, and an initial setting means for classifying all color data in the area into a plurality of groups based on the Euclidean distance. Group dividing means;
20. The image processing apparatus according to claim 16, further comprising an initial representative extracting unit that sets an average value of the color data in each group obtained by the initial group dividing unit as a representative color of the group. 24. A representative color representing a group is an average value of color data in the group.
The image processing apparatus according to any one of the preceding claims.