JP2625356B2 - Image processing device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus,
In particular, the present invention relates to an image processing apparatus for compressing a gradation of an image.

[0002]

2. Description of the Related Art First, gradation compression of an image will be described with reference to FIG. FIG. 16 shows an example of a density histogram of an image represented by 14 gradations. In the graph of FIG. 16, the horizontal axis represents each density of 14 gradations, and the vertical axis represents the ratio of pixels of the corresponding density to all pixels. Here, a case where an image represented by 14 gradations is represented by 3 gradations will be described. In this case, it is necessary to divide the density of 14 gradations into three groups, and to associate one density with each group. Here, the density assigned to each group is called a group density. In the case of FIG. 16, the 14 gradations are G (1) to G (3).
, And each group has T (1)
To T (3).

[0003] Such gradation compression can be divided into three processes.

The first step is a step of creating a density histogram from an original image. In this process, it is required to create a density histogram at high speed.

The second step is a step of dividing the density into N groups based on the density histogram created in the first step. Dividing the density into N groups is the same as setting (N-1) thresholds on the horizontal axis of the density histogram. For example, in the case of FIG. 16, since N = 3, the density is divided into three by the first threshold value and the second threshold value. In this process, a threshold value must be set so that the compressed image looks as natural as possible. Specifically, in the image, a number of threshold values are densely set in a density section expressing a target to be watched. This is because the object is expressed with many gradations and does not feel unnatural even after compression. on the other hand,
The threshold is roughly set in the density section representing the background portion of the image. This is because the background portion is not watched, so that it does not seem unnatural even if it is displayed with a small number of gradations.

[0006] The third step is a step of giving a group density to the density divided in the second step.

Of the series of gradation compression processing described above, the second
Examples of the integral division method, which is a conventional technique of the process, are Takeshi Aiin,
This is described in pages 130-132, published by Morikita Publishing Co., Ltd. (hereinafter referred to as Reference 1), published by Masayuki Nakajima, "Image Information Processing", March 1992. The integral division method is a method of setting a threshold value so that the frequency of each group becomes equal. By setting the threshold value in this manner, the threshold value is set densely in the high-frequency density section, and the threshold value is set coarsely in the low-frequency density section. Since the high-frequency density section is often a portion to be watched in the image, the watched region is expressed by a number of gradations.

[0008]

Although the integral division method is easy to execute, it has a problem that improper gradation compression is performed depending on the type of image. This problem will be described by taking the case of compressing the image of FIG. 17 as an example. The image of FIG. 17 shows a small object 1 in a white background.
501 exists. FIG. 18 is a density histogram of the image of FIG. In FIG. 18, density 1 represents white, and density 14 represents black. Most of the image in FIG. 17 is occupied by a white background. For this reason, in the histogram of FIG. 18, the frequencies are concentrated on the density 1 to density 4, which are white sections. On the other hand, object 15
Since 01 is very small, the density 9 to density 11, which is the display section of the object 1501, occupies only a small frequency. Therefore, when the integral division method is applied to the frequency histogram of FIG. 18, the threshold values concentrate on the density 1 to density 4. For this reason, only one object 1501 is watched.
The image is displayed at the gradation density, and the image after compression becomes extremely unnatural.

Further, the optimum gradation compression varies depending on the state of the original image and the display device for displaying the compressed image. For this reason, it is preferable to create candidates by performing some gradation compression and select the most suitable one with the naked eye.
However, in the above-mentioned integral division method, since only one type of gradation compression can be performed, the optimum gradation compression cannot be selected from several candidates.

[0010]

In order to solve the above-mentioned problems, according to the present invention, a function is set which increases when a suitable threshold value is used and decreases when an inappropriate threshold value is used. The tone compression of the image is executed according to the threshold value to be converted. A method based on a genetic algorithm is used to find a threshold that maximizes a function. For details of the genetic algorithm, see Takeshi Yasui and Tomoharu Nagao, "Processing and Recognition of Images," Nov. 25, 1992, published by Shokodo Co., Ltd.

More specifically, in order to solve the above-described problem, the image processing apparatus according to the present invention is configured such that the gene data holding a plurality of threshold values and the threshold value holding the gene data are desirable. Sometimes it produces large values,
Fitness generating means for generating a small value when the threshold value held by the gene data is undesirable, and optimal gene data for obtaining and outputting gene data which maximizes the value generated by the fitness generating means. Calculating means, a group density determining means for creating and outputting a density group and a group density of the density group from a threshold value held by the gene data output from the optimum gene data calculating means, and a group density determining means. Output means for tone-compressing and outputting the original image based on the density group to be output and the group density.

[0012]

Next, a first embodiment of the present invention will be described with reference to the drawings.

First, data referred to as "gene data" in the present embodiment will be described with reference to FIG. Referring to FIG. 3, XN pieces of gene data X (i) are held in the gene data storage area 121 of the pixel specifying means 120. Gene data X (i) includes SN segments S (i, 0) to S (i, S) indicating the position of the threshold.
N-1). Here, SN is the number of thresholds set by gradation compression. For example, FIG.
, SN = 2. Further, the segment S (i,
j) are BN bits B holding a value of 0 or 1
(I, j, 0) to B (i, j, BN-1). Here, the BN is prepared in a sufficient number to express the position of the threshold. For example, in the case of FIG. 14, BN = 4
It is. Thereby, each segment S (i, j) is 0 to
15 values can be expressed. In the following description, S
Each segment itself is indicated by the symbol (i, j), and the value of the number held by this segment is also expressed. By the way, in the case of FIG. 14, it is sufficient to represent a value of 0 to 12 to indicate the position of the threshold value. Therefore, when S (i, j)> 12, the segment S (i, j)
The value of j) becomes meaningless. Therefore,
When S (i, j)> 12, the segment S (i, j)
It is convenient to use a cyclic notation in which j) is regarded as indicating the number of (S (i, j) -12). In the following description, it is assumed that this cyclic notation is used.

Next, the structure of this embodiment will be described.

Referring to FIG. 1, according to a first embodiment of the present invention, an image storage means 140 for storing an original image and a gradation compression apparatus 110 for compressing the gradation of the original image stored in the image storage means 140 are provided. A random number generator 130 for supplying a random number to the gradation compressor 110; and a pixel designator 120 for storing gene data generated by the gradation compression process of the gradation compressor 110.
It is composed of The pixel designation means 120 includes a gene data storage area 121 for storing information of gene data.
And a work area 122 for temporarily storing the genetic data when updating the genetic data. Gradation compression device 11
0 is a density histogram creation unit 111 that creates a density histogram of the original image stored in the image storage unit 140
Gene data generating means 112 for randomly generating gene data and setting the gene data in the gene data storage area 121
A gene data updating means 113 for updating the gene data stored in the gene data storage area 121 in accordance with a procedure described later, and an optimal gene data selection for selecting the optimal gene data from the gene data updated by the gene data updating means 113 Means 114, group concentration determining means 115 for determining a group of concentrations and the group concentration based on the optimal gene data selected by optimal gene data selecting means 114, and group concentration determining means 115
And an output unit 116 for tone-compressing and outputting the original image stored in the image storage unit 140 on the basis of the group and the group density determined in (1).

Still referring to FIG. 2, the genetic data updating means 113 performs duplication and deletion of the genetic data in the genetic data storage area 121 in accordance with a procedure described later to perform genetic data selection. A normalizing means 203 for normalizing the number of gene data selected by the selecting means 202 to a certain number;
3. Crossover means 204 for randomly selecting two pieces of gene data normalized by 3 and exchanging a part of the information constituting the gene data to perform crossover of the gene data.
A mutation means 205 for randomly changing information of the gene data crossed by the crossing means 204, and a selection means 2
02, normalization means 203, crossover means 204, and mutation means 2
05 and an update control means 201 for controlling the operation of the system.

Next, the operation of this embodiment will be described. In this embodiment, a case will be described in which an original image having M gradations is gradation-compressed to M 'gradations. Therefore, the number of segments SN per gene data is M'-1, and the number of bits BN per segment is the minimum number required to represent the number of 0 to (M-2).

When the tone compression device 110 of this embodiment is started, first, the density histogram creating means 111 in the tone compression device 110 is operated. The density histogram creating means 111 obtains the density histogram at high speed by an approximate method. This is for compensating the operation time of the gene data updating means 113. Hereinafter, the operation of the density histogram creating unit 111 will be described with reference to the flowchart of FIG.

Referring to FIG. 4, in step 401, 0 is set in the density histogram H (i). Here, i is 0 to (M−) corresponding to the density of M gradation of the original image.
1). When step 401 ends, the process moves to step 402.

In step 402, the image storage means 140
EN pixels are extracted from the pixels of the original image held in. Pixels extracted at this time are selected at random based on random numbers supplied from the random number generation means 130 via the signal line A. When step 401 ends, the process moves to step 403.

In step 403, N (i) is updated. N (i) is the number of pixels of density i among the pixels extracted up to this point. The update of N (i) is performed by adding the number of pixels having the density i among the EN pixels extracted in step 402 to N (i). When step 402 ends, the process moves to step 404.

In step 404, a new density histogram H '(i) is calculated according to the following (1).

[0023]

When step 404 ends, the process moves to step 405.

In step 405, D is calculated according to the following equation (2).

[0026]

D is an index indicating the degree of change from the density histogram H (i) to the new density histogram H '(i). As D increases, the change from H (i) to H' (i) occurs. Indicates that it is big. When step 405 ends,
The process proceeds to step 406.

In step 406, the density histogram H
A new density histogram H ′ (i) is set in (i).
When step 406 ends, the process moves to step 407.

In step 407, the magnitude of D is determined. If D is greater than a predetermined value,
The process branches to step 402. On the other hand, if D is equal to or smaller than the predetermined value, the process proceeds to step 408.

In step 408, the density histogram H
(I) is output to the gene data updating means 113. Upon completion of step 408, density histogram creating means 1
11 stops the operation.

When the tone compression device 110 is started, the gene data generating means 112 is also started at the same time as the density histogram creating means 111. Gene data X in the gene data storage area 121 is generated by the gene data generating means 112.
A random value is set to each bit B (i, j, k) of (i). Hereinafter, the operation of the density histogram creating unit 111 will be described with reference to the flowchart of FIG.

In step 501, 0 is set to a variable w. When step 501 ends, the process moves to step 502.

In step 502, the random number generating means 130
, A random number r of 0 to 1 is generated. Step 502
Is completed, the process moves to step 503.

In step 503, among the bits of the gene data in the gene data storage area 121, (i × XN
× SN) + (j × SN) + k is a bit B (i,
j, k) is set to r. When step 503 ends, the process moves to step 504.

In step 504, 1 is added to the variable w. When step 504 ends, the process moves to step 505.

At step 505, the value of the variable w is determined. If the value of the variable w is smaller than (XN × SN), the process branches to step 502. On the other hand, when the value of the variable w becomes equal to (XN × SN), the entire process ends. When step 505 ends, the gene data generating means 112 stops operating. When the concentration histogram creating means 111 and the gene data generating means 112 end the processing, the gene data updating means 113 is activated.

The genetic data updating means 113 obtains, based on a genetic algorithm, genetic data that maximizes a function F () described later. That is, the genetic algorithm holds a group of genetic data. In the case of this embodiment, each gene data indicates a threshold. The selection means 202 repeatedly performs the operations of duplication, mutation, and extinction on these genetic data, thereby executing virtual natural selection. In the duplication and extinction operations, F
() Increases the chances of replication for large genetic data, and F () increases the chances of disappearance for small genetic data. Then, as the virtual natural selection process progresses, genetic data having a large F () occupies the population. In addition, new genetic data is successively generated in the population by the mutation operation performed by the crossing means 204 and the mutation means 205. Thus, in the initial state, F
Even if genetic data that maximizes () does not exist in the population, such genes will occur in the population during the selection process, and the genetic data will eventually occupy the population. Then, if the virtual natural selection is performed for a sufficiently long time, in the final state, the population is occupied by the genetic data that maximizes F (). As described above, genetic data that maximizes F () can be obtained by the genetic algorithm. The above is the outline of the gene data algorithm in the present embodiment. Specifically, when the gene data updating means 113 is activated, first, the selection means 202 operates. Hereinafter, the operation of the selection unit 202 will be described with reference to the flowchart of FIG.

In step 506, the variable w is set to 0. When step 506 ends, the process moves to step 507.

In step 507, the selection means 202 calculates a numerical value called fitness. Here, the fitness is an index for executing duplication and deletion of each gene data. The function F (X (i)) for determining the fitness is
It is set in the fitness generating means 206 in advance. F (X
(I)) is set so as to take a large value when the threshold value indicated by X (i) is suitable, and to take a small value when the threshold value indicated by X (i) is inappropriate. Need to be done. Here, in order to solve the above-mentioned problem of the integral division method, the function F (X (i)) is calculated by the following equation (3).
Is defined.

[0040]

In equation (4), S ′ (i, j) is S
When (i, 0) to S (i, SN-1) are arranged in ascending order, this is a value indicating the (j + 1) th one. That is, among the SN thresholds represented by S (i, j), the threshold value is the (j + 1) th threshold when counted from the left. FIG. 7 shows the relationship between S (i, j), S ′ (i, j) and each threshold. Also, C1 and C2
Is an appropriate constant. The first term on the right side of the above equation (4) becomes maximum when the threshold value indicated by X (i) follows the integral division. On the other hand, the second term on the right side of the equation (4) is X
It becomes maximum when the threshold values shown in (i) are at equal intervals. Therefore, F (X (i)) increases due to the effect of the first term when the threshold value indicated by X (i) approaches the integral division. However, when the threshold value is concentrated in a specific section as shown in FIG. 18 as a result of the integral division, the threshold value is reduced by the effect of the second term. And F (X (i))
Has a maximum value in that the effects of the first and second terms are antagonized. Although the function F (X (i)) is a non-linear function, processing is performed based on a genetic algorithm in the present embodiment, so that gene data that maximizes F () can be obtained. When F (X (w)) is calculated in step 507, the process proceeds to step 508. (4)
The significance of the expression is described in more detail below. The first term including the minus sign on the right side of equation (4) is f (S ′ (i,
j + 1)) and f (S ′ (i, j)) match, that is, when the threshold value indicated by X (i) matches the integral division, the threshold value departs from the integral division. The more the value decreases. On the other hand, the second term including the minus sign on the right side of equation (4) is S ′ (i, j + 1) and S ′ (i,
When j) coincides, that is, when the threshold value indicated by X (i) is at equal intervals, the maximum value is obtained, and the value decreases as the threshold value moves away from the equal intervals. The first and second terms
Since it is extremely rare that there is a threshold value that maximizes both terms, F ′ (X
(I)) usually takes the maximum value at a moderate threshold, which is neither the threshold maximizing the first term nor the threshold maximizing the second term. Whether this moderate threshold value is closer to integral division or equal division can be adjusted by the values of the coefficients C1 and C2. By adjusting C1 and C2, the threshold value for maximizing F () can be set to have the following properties. That is, the threshold value thus set usually approaches the integral division due to the effect of the first term of the equation (4), but if the threshold interval is concentrated in a specific section as a result of the integral division, ( 4) Due to the effect of the second term of the equation, integral division is separated and approaches equal division. In this way, it is possible to prevent the threshold value from being concentrated on a specific section.

In step 508, the random number generator 130 generates a random number r of 0 to (2 × F (X (w))). When step 508 ends, the process moves to step 509.

In step 509, X
R copies of (w). When step 509 ends, the process moves to step 510.

In step 510, w is incremented by one. When step 510 ends, the process moves to step 511.

In step 511, the value of w is determined. If w is less than XN, the process branches to step 507. On the other hand, when w = XN, that is, when the processing of steps 507 to 509 is completed for all the gene data X (i) in the gene data storage area 121, the update control means 201 stops operating. . Gene data generated in the work area 122 by the selection means 202 is represented by X ′ (i). X '(i) is F
Those with large (X '(i)) occupy a large number, and F (X'
Those in which (i) is small are decimal numbers. In addition, the total number of the genetic data X '(i) is random and not constant.
Therefore, it is necessary to normalize the total number to XN by the normalizing means 203.

When the processing of the selection means 202 is completed, the normalization means 203 is activated. Hereinafter, the operation of the normalizing means 203 will be described with reference to the flowchart of FIG.

In step 601, the work area 122
From the number XN 'of the gene data X (i) replicated in X
The number obtained by subtracting N is set as a variable w. When step 601 ends, the process moves to step 602.

In step 602, the value of w is determined. If w is 0, the process branches to step 611. If w is smaller than 0, the process branches to step 607. Step 6 when w is greater than 0
The process is shifted to 03.

In steps 603 to 606, X '(i)
, W genetic data is extracted at random and duplicated every two. Thereby, the work area 122
Is normalized to XN. When step 606 ends, the process moves to step 611.

In steps 607 to 610, X '
From (i), w genetic data are randomly selected and deleted. However, the random number generating means 130 is controlled so that the same gene data is not deleted twice. In steps 607 to 610, the number of gene data in the work area 122 is normalized to XN. Step 6
When 10 is completed, the process moves to step 611.

At step 611, the work area 122
XN gene data X ′ (0) to X ′ (XN−
1) is set in the gene data storage area 121. When step 611 ends, the normalizing means 203 stops operating and the crossover means 204 is activated.

Hereinafter, the operation of the crossover means 204 will be described with reference to the flowchart of FIG.

In step 701, INT (XN
× CR) is set. Here, CR is a constant called a crossover rate. INT () is a function for rounding a decimal number and converting it to an integer. When step 701 ends, the process moves to step 702.

In steps 702 to 704, the random number generation means 130 generates random numbers r1 and r1 of 0 to (XN-1).
2, and a random number r3 of 0 to (SN-1) are generated.
When step 704 ends, the process moves to step 705.

In step 705, S (r1, r3) 〜
S (r1, SN-1) and S (r2, r3) to S (r2,
SN-1) are exchanged. An outline of this operation is shown in FIG. When step 705 is completed, step 7
The process is shifted to 06.

In step 706, w is reduced by one. When step 706 ends, the process moves to step 707.

In step 707, the value of w is determined. If w is 1 or more, the process branches to step 702. On the other hand, when w is 0, that is, at step 70
When the processing of steps 2 to 705 has been completed INT (XN × CR) times, the crossover means 204 stops operating and the mutation means 20
5 is activated.

Hereinafter, the operation of the mutation means 205 will be described with reference to the flowchart of FIG. In step 901, the number INT (XN × MR) is set to w. Here, MR is a constant called a mutation rate. When step 901 ends, the process moves to step 702.

In step 902, the random number generation means 130
, A random number r of 0 to (XN × SN−1) is generated. When step 902 ends, the process moves to step 903.

In step 903, (i × XN × SN)
The value of B (i, j, k) that satisfies + (j × SN) + k = w is inverted. When step 903 ends, step 9
The process is shifted to 04.

In step 904, w is reduced by one. When step 904 ends, the process moves to step 905.

In step 905, the value of w is determined. If w is 1 or more, the process branches to step 902. On the other hand, when w is 0, that is, at step 90
When the processing of steps 2 to 903 has been completed INT (XN × MR) times, the mutation unit 205 stops the processing, and the update control unit 201 determines the configuration of the gene data X (i) in the gene data storage area 121. . Hereinafter, the operation of the update control unit 201 will be described with reference to the flowchart of FIG.

The update control means 201 which has sequentially activated the selection means 202 to the mutation means 205 in steps 1001 to 1004 determines the configuration of the gene data X (i) in the gene data storage area 121. Gene data storage area 1
21 is the specific genetic data X (i).
If it is occupied by a predetermined ratio or more by the maximum, the update control unit 201 outputs Xmax to the group density determination unit 115. At this time, the difference of X (i) is S ′ (i,
j) is determined. That is, even when S (i, j) is different, it is determined that the genetic data having the same S ′ (i, j) is the same.

On the other hand, when the gene data storage area 121 is not occupied by the specific gene data at a predetermined ratio or more, the process branches to step 1001. By repeatedly executing steps 1001 to 1004, the gene data in the gene data storage area 121 changes sequentially. Finally, most of the gene data storage area 121 is occupied by the gene data X (i) that maximizes F (), and the processing of the update control means 201 ends. When the processing of the update control means 201 ends,
The group density determining means 115 is activated. The group density determination means 115 determines the groups G (0) to G ′ of M ′ gradation densities based on the threshold indicated by the optimal gene data Xmax output from the gene data update means 113.
(M'-1) and the group density T (0) of this group
~ T (M'-1).

Specifically, when 0 <i <(SN−1), G (i) is (S′max (i) +1) to S′ma
x (i + 1). Where S'm
ax (i) is S ′ of the genetic data Xmax.
The definition of S 'has been described above.

When i = 0, G (i) is 0 to
It is constituted by the density of S'max (0). And i
= SN-1, G (i) is S'max (SN-
1) It is composed of +1 to (M-1) concentrations.

On the other hand, there are several methods for determining the group density. Here, two methods will be described.

The first method is a case where an intermediate density in each group G (i) is set as a group density T (i). A group density T (i) determined by this method is shown at 1102 in FIG.

The second method is a method in which the most frequent density in each group G (i) is set as the group density T (i). The reference numeral 1101 in FIG. 13 shows the group density T (i) determined by this method.

The determined group G (i) and group density T (i) are output to the output means 116. When the group G (i) and the group density T (i) are output, the operation of the group density determining means 115 stops, and the output means 116 is activated.

The output means 116 compresses the gradation of the original image in the image storage means 140 according to the group G (i) and the group density T (i) determined by the group density determining means 115, and outputs it to the outside. Since the gradation compression of an image when the group G (i) and the group density T (i) have already been determined is a commonly practiced technique, a detailed description thereof will be omitted. When the tone-compressed image 117 is output by the output unit 116, the tone compression device 110 stops operating.

Next, the effect of this embodiment will be described.

In this embodiment, a function F () is set which increases as the threshold approaches the integral division and decreases when the thresholds are excessively dense, and maximizes the function F (). The gradation compression of the image is performed by the threshold value. Therefore, it is possible to solve the problem of the integral division method in which threshold values are concentrated in a specific section, and to obtain a more natural gradation compressed image. At this time, since the function F () becomes non-linear, the threshold value is obtained by a genetic algorithm.

In the present embodiment, F () is defined by equation (3), but the scope of the present invention is not limited to this. F () may be a function that increases when the threshold value indicated by the gene data is favorable and decreases when the threshold value is undesirable, and realizes various gradation compression depending on the setting method of F (). can do.

Next, a second embodiment of the present invention will be described.

Referring to FIG. 14, in the second embodiment of the present invention, in addition to the configuration of the first embodiment, output means 116 is provided.
Display means 1202 and second display means 1203 for respectively displaying the two images output from the printer, and input means 120 for inputting information for selecting the two images
4 and 2 according to the information input from the input means 1204.
Selecting means 12 for selectively outputting one of the two images
01.

Next, the operation of this embodiment will be described. Also in the present embodiment, the operations of the density histogram creation unit 111, the gene data generation unit 112, the selection unit 202, the normalization unit 203, the crossover unit 204, and the mutation unit 205 are not different from those in the first embodiment. .
The feature of this embodiment lies in the operation of the update control means 201 and the processing after the group density determination means 115.

FIG. 15 shows a flowchart of the operation of the update control means 201 in the present embodiment. FIG.
In step 5, steps 1301 to 1304 are not different from those of the first embodiment. On the other hand, step 1305 executes an operation different from that of the first embodiment.

In step 1305, the update control means 201 determines that the genetic data in the genetic data storage area 121 is the two specific genetic data Xmax1 and Xmax.
It is determined based on x2 whether or not a predetermined ratio or more is occupied. If not occupied, the process branches to step 1301. On the other hand, if the gene data storage area 121 is occupied by Xmax1 and Xmax2 at a predetermined ratio or more, the process proceeds to step 1306. Here, the property of the genetic algorithm is used, in which if the processing is stopped at an appropriate time, a plurality of variable candidates that maximize the objective function can be obtained.

At step 1306, Xmax1 and Xmax2 are output to group density determining means 115. Upon completion of step 1306, the gene data updating means 113 stops operating and the group concentration determining means 11
5 is activated.

The group density determining means 115 outputs Xmax1 and Xmax output from the genetic data updating means 113.
The first group, the first group density, the second group, and the second group density are determined based on max2, and output to the output unit 116. The method of determining the group and the group density is not different from that of the first embodiment.

The output means 116 outputs the original data stored in the image storage means 140 based on the first group, the first group density, the second group, and the second group density input from the group density determining means 115. Executing gradation compression of the image, and selecting the first image and the second image by the selecting means 1
201, a first display means 1202 and a second display means 1203.

The first display means 1202 is connected to the output means 11
6 is displayed.

The second display means 1203 is connected to the output means 11
6 is displayed.

The input unit 1204 is, for example, a keyboard or the like. The input unit 1204 inputs which of the first image and the second image is to be selected, and outputs the selected image to the selection unit 1201.

The selecting means 1201 selects one of the first image and the second image based on the information input from the input means 1204, and outputs it as a gradation-compressed image 117. When the gradation-compressed image 117 is output, the gradation compression device 110 stops operating.

Next, the effect of this embodiment will be described.

In the present embodiment, an image is displayed by executing two types of gradation compression from two candidates of the optimal gene data, and one of them is selected by the selection means. For this reason, gradation compression suitable for a display image or a display device can be executed. In the present embodiment, the property of the genetic algorithm is utilized in which a plurality of variable candidates for maximizing the objective function can be obtained by stopping the processing at an appropriate time.

[0089]

In the first embodiment of the present invention, a function F () is set which increases when the threshold value is favorable and decreases when the threshold value is inappropriate, and maximizes this function. The gradation compression of the image is executed by the threshold value. Therefore, desired gradation compression can be performed flexibly. Also,
By employing the equation (3) as the function F (), it is possible to solve the problem of the integral division method in which the threshold value concentrates in a specific section depending on the state of the original image.

In the first embodiment of the present invention, when creating a density histogram, pixels are extracted in a divided manner, and the process is terminated when the change in the density histogram falls below a certain level. Therefore, the creation of the density histogram can be executed at high speed.

Further, in the second embodiment of the present invention, a plurality of gradation compression candidates are created and a suitable one is selected by the selection means. Natural gradation compression can be performed.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a first embodiment of the present invention.

FIG. 2 is a block diagram showing a detailed structure of a gene data updating means 113.

FIG. 3 Gene data X, segment S and bit B
FIG.

FIG. 4 is a flowchart showing the operation of a density histogram creation unit 111;

FIG. 5 is a flowchart showing the operation of the gene data generating means 112.

FIG. 6 is a flowchart showing the operation of the gene data updating means 113.

FIG. 7 is a diagram showing a relationship between S and S ′.

FIG. 8 is a flowchart showing the operation of the normalizing means 203.

FIG. 9 is a flowchart showing the operation of the crossover means 204.

FIG. 10 is a diagram showing the operation of the crossover means 204.

FIG. 11 is a flowchart showing the operation of the mutation means 205.

FIG. 12 is a flowchart showing the operation of the update control means 201.

FIG. 13 is a diagram showing the operation of the group gradation determining means 115.

FIG. 14 is a block diagram showing a second embodiment of the present invention.

FIG. 15 shows an update control unit 201 according to the second embodiment of the present invention.
5 is a flowchart showing the operation of the embodiment.

FIG. 16 is a diagram showing gradation compression of an image.

FIG. 17 is a diagram showing an example of an original image.

FIG. 18 is a diagram showing an example of integral division.

[Explanation of symbols]

 Reference Signs List 110 tone compression device 111 density histogram creating means 112 gene data generating means 113 gene data updating means 114 optimal gene data selecting means 115 group density determining means 116 output means 121 gene data storage area 122 work area 130 random number generation means 140 image storage means 201 update control means 202 selection means 203 normalization means 204 crossover means 205 mutation means 1201 selection means 1202 first display means 1203 second display means 1204 input means 1501 object

Claims (8)

(57) [Claims] 1. Gene data holding a plurality of threshold values, and when the threshold value held by the gene data is desirable, a large value is generated, and the threshold value held by the gene data is undesirable. A fitness value generating means for generating a small value when the value is obtained, an optimal gene data calculating means for obtaining and outputting gene data which maximizes a value generated by the fitness value generating means, and an output of the optimal gene data calculating means. Group density determining means for creating and outputting a density group and a group density of the density group from the threshold value held by the gene data to be processed, and the density group and the density based on the group density output by the group density determining means. An output unit for outputting an image after gradation-compressing the image. 2. The method according to claim 1, wherein the value generated by said fitness generating means is a first term that increases when a threshold value held by said gene data approaches an integral division, 2. The image processing apparatus according to claim 1, further comprising a second term that decreases when the density is excessively high. 3. Gene data storage means for storing gene data holding a plurality of threshold values, and when the threshold value held by the gene data is desirable, a large value is generated to hold the gene data. A fitness generating means for generating a small value when the threshold value to be performed is not desirable; and gene data for duplicating or extinguishing the genetic data in the genetic data storage means according to the value generated by the fitness generating means. A selection means; a gene data crossing means for selecting two gene data from the gene data in the gene data storage means, exchanging information of the two gene data, and storing the information again in the gene data storage means; Gene data mutation means for changing the information of the gene data in the gene data storage means; Activating the genetic data selection means, the crossover means, and the genetic data mutation means until the genetic data is occupied by the first genetic data in a predetermined ratio or more, and the genetic data in the genetic data storage means is changed to the first gene data. A control means for outputting the first gene data when the data is occupied by a predetermined ratio or more; a density group and a group density of the density group based on a threshold value held by the first gene data output from the control means; And a group density determining unit for generating and outputting the image, and an output unit for performing a gradation compression of the original image based on the density group and the group density output from the group density determining unit. Processing equipment. 4. The method according to claim 1, wherein the value generated by the fitness generating means is a first term that increases when a threshold value held by the gene data approaches an integral division, The image processing apparatus according to claim 3, further comprising a second term that decreases when the density is excessively high. 5. Genetic data holding a plurality of threshold values, and a large value is generated when the threshold value held by the genetic data is desirable, and the threshold value held by the genetic data is undesirable. A fitness generating means for generating a small value when the value is an optimal gene data calculating means for obtaining and outputting a plurality of gene data candidates which maximize the value generated by the fitness generating means; Group concentration determining means for creating and outputting a plurality of concentration groups and group concentrations of the concentration groups from threshold values held by the plurality of gene data output by the calculating means; and a plurality of groups output by the group concentration determining means Output means for outputting a plurality of gradation-compressed images by gradation-compressing the original image based on the extended group density; The image processing apparatus characterized by having a selection means for selectively outputting the plurality of gradation compression image force. 6. A value generated by said fitness generating means is a first term which increases when a threshold value held by said genetic data is an integral division, and when a threshold value held by said gene data is excessive. 6. The image processing apparatus according to claim 5, further comprising: a second term that decreases when the density is high. 7. Gene data storage means for storing gene data holding a plurality of threshold values, and when the threshold value held by the gene data is desirable, a large value is generated to hold the gene data. A fitness generating means for generating a small value when the threshold value to be performed is not desirable; and gene data for duplicating or extinguishing the genetic data in the genetic data storage means according to the value generated by the fitness generating means. A selection means; a gene data crossing means for selecting two gene data from the gene data in the gene data storage means, exchanging information of the two gene data, and storing the information again in the gene data storage means; Gene data mutation means for changing the information of the gene data in the gene data storage means; Start with the genetic data mutation means, the cross unit and the genetic data selection means to the genetic data is occupied predetermined proportion or more in the first gene data gene data and the second gene data,
Control means for outputting the first gene data when the gene data in the gene data storage means is occupied by the first gene data and the second gene data in a predetermined ratio or more; A first group and a first group concentration are created and output from the threshold value held by the first genetic data, and the second group output by the control means is output.
Group density determining means for generating and outputting a second group and a second group density from threshold values held by the gene data of the first and second groups, and the first group and the first group output by the group density determining means. The original image is gradation-compressed based on the group density and a first gradation-compressed image is output, and the second group and the second group density output from the group-density determining means are used to generate the first gradation-compressed image. Output means for outputting a second gradation-compressed image by gradation-compressing the original image; and outputting one of the first gradation-compressed image and the second gradation-compressed image output from the output means. An image processing apparatus comprising: a selection unit for selecting and outputting. 8. A value generated by the fitness generating means is a first term that increases when a threshold value held by the genetic data approaches an integral division, and a value generated by the fitness value generating means includes a threshold value held by the genetic data. The image processing apparatus according to claim 7, further comprising: a second term that decreases when the density is excessively high.