JP2896319B2 - How to make a color image mask - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for producing a mask used for cutting out a desired color portion in a color image.

[0002]

2. Description of the Related Art In a printing plate making process, it is sometimes desired to selectively perform various processes such as color change and sharpness enhancement on a specific color portion of a color image such as a photo original.
For example, there are cases where it is desired to change only the color of the clothes of the model wearing the clothes, or to emphasize the sharpness on the portion excluding the skin of the model. As described above, when the desired processing is performed (or not performed) only on a specific color portion of a color image,
A cutout mask indicating the color portion is created.

FIG. 1 is an explanatory diagram showing an example of a color image and a preferable mask. The color image shown in FIG. 1A is a photograph in which a black hair model is wearing clothes in which a rose flower pattern is printed on a yellow background. Rose flower
It has white and pink petals and green leaves around it. The mask shown in FIG. 1B is a mask showing a yellow portion and a green portion of the clothes.

As a method of creating a mask by analyzing the color of a color image, for example, Japanese Patent Laid-Open No. 1-298
There is one described in Japanese Patent No. 477. According to this method, the colors of a plurality of pixels in a portion (training area) to be a mask region are analyzed to obtain a principal component axis of the color. When the color of each pixel in the color image includes many components other than the principal component axis (that is, when the color is far from the principal component axis), the pixel is excluded from the mask area.

[0005]

However, in the above-described conventional technique, it was difficult to exclude the skin color from the mask area while using yellow as the mask area as shown in FIG. 1 (B). This is because the skin color contains a considerable amount of yellow component, and if yellow is specified as the color of the mask area, the skin color is also recognized as the mask area. Such a problem is a common problem when one of two colors having relatively similar hues is used as a mask region and the other is excluded from the mask region.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems in the prior art, and among a plurality of colors having relatively similar hues, a part of a color portion is used as a mask region.
It is an object of the present invention to provide a mask creating method capable of excluding other color portions from a mask area.

[0007]

In order to solve the above-mentioned problems, a method according to claim 1 of the present invention comprises:
(A) determining a plurality of representative colors based on a statistical distribution of colors in the color image, and displaying the plurality of representative colors on a display device; and (b) selecting the plurality of representative colors from among the plurality of representative colors. Specifying a target color of the mask; (c) specifying a non-target color of the mask from among the plurality of representative colors; and (d) a first color vector of the target color in a predetermined color space. And (e) obtaining a second color vector of the non-target color in the predetermined color space;
(F) obtaining a third color vector independent of the first and second color vectors; and (g) determining a color of each pixel in the color image by calculating a color of the first to third color vectors. Obtaining first to third coefficients for the first to third color vectors when represented by a linear combination;
(H) creating the mask according to the value of the first coefficient for each pixel in the color image.

According to the above-described method, the first coefficient is large and the second coefficient is small for a color close to the target color, while the first coefficient is small for the color close to the non-target color. Becomes larger. Therefore, even when the hue of the target color and the non-target color are close to each other, if a mask is created according to the first coefficient, a color part close to the target color is set as a mask area and a color part close to the non-target color is set as a mask area Can be excluded from In addition, multiple representative colors are displayed based on the statistical distribution of colors in the color image, and the target color and the non-target color are specified from among them. Therefore, the target color and the non-target color can be easily specified. can do.

In the method according to the second aspect, the step (a) includes a step of dividing the color space into a plurality of subspaces.
Obtaining a frequency of appearance of a color in the color image in each of the plurality of subspaces; and, in accordance with the frequency of appearance, a color representing at least a part of the plurality of subspaces is represented by the plurality of colors. And displaying as a representative color.

[0010] This makes it possible to easily designate a color having a high appearance frequency in the color image as a target color or a non-target color.

In the method according to the third aspect, the step of dividing the color space into a plurality of subspaces includes a step of obtaining a histogram for each component of a three-dimensional color system with respect to colors in the color image. And setting a boundary around the local minimum of the histogram of each component,
Partitioning the color space into the plurality of subspaces partitioned by the boundary.

In this way, since the subspace is divided according to the frequency of appearance of the colors in the color image, a plurality of representative colors representing the main colors existing in the color image can be obtained.

According to a fourth aspect of the present invention, the step (h) includes a step of creating mask data representing the mask by binarizing the first coefficient with a threshold value.

This makes it possible to easily create binary mask data indicating only the mask area.

In the method according to the fifth aspect, the step of binarizing the first coefficient includes the step of creating a histogram indicating the frequency of occurrence of the first coefficient, and the step of binarizing the first coefficient. Setting a coefficient of 1 as the threshold value.

This makes it possible to automatically determine the threshold for binarization according to the histogram.

In the method described in claim 6, the third color vector is a vector representing white.

In this way, the third color vector can be easily determined.

In the method described in claim 7, the step (f) converts the first and second color vectors into fourth and fifth color vectors in a hue / saturation / brightness space, respectively. And if the saturation component of at least one of the fourth and fifth color vectors exceeds a predetermined value,
A vector representing white in the predetermined color space is selected as the third color vector, and when at least one of the fourth and fifth color vectors has a saturation component less than the predetermined value, the third color vector is selected. Of the fourth and fifth color vectors, for the color vector having a larger saturation component, the hue / saturation /
Generating a sixth color vector obtained by rotating the hue by a predetermined angle in the lightness space, and generating the third color vector by converting the sixth color vector into a vector in the predetermined color space; Is provided.

This makes it possible to easily determine a third color vector independent of the first and second color vectors. In addition, since the first to third color vectors each represent an actual color, the first to third color vectors correspond to the color of each pixel.
Is an index that actually represents the component of the target color.

In the method described in claim 8, the first to third color vectors are unit vectors.

In this way, the first to third color vectors are vectors representing only the hues, so that a mask can be created according to the hues of each color image.

[0023]

【Example】

A. First Embodiment FIG. 2 is a block diagram showing the configuration of a mask creation system for creating a cutout mask by applying an embodiment of the present invention. This mask making system
A CPU 20, an original image memory 22 for storing a color original image, a vector value memory 24 for storing color vector components, a matrix memory 26 for storing matrix components described later, and a multi-level memory for storing multi-tone mask data. This is a computer system including a gradation mask memory and a binary mask memory 30 for storing binary mask data.
The system further includes a keyboard 32 and a mouse 34 as input means or designation means, a reading scanner 36 as image input means, a color CRT 38 as display means, and a hard disk device 40 as external storage means. I have.

The CPU 20 includes a matrix operation unit 50,
It has a function as a multi-tone mask calculation unit 52 and a binarization calculation unit 54. These components are realized by the CPU 20 executing a software program stored in a RAM (not shown). Note that these functions will be further described later.

FIG. 3 is a flowchart showing the overall procedure for producing a cutout mask in the embodiment. Step S
In step 1, the image data of the color original is read by the reading scanner 36 and stored in the original image memory 22. This color original is, for example, a color photograph as shown in FIG. The original image memory 22 is a frame memory, and the image stored in the original image memory 22 is a color CRT.
38 is displayed.

In step S1a, the color original is scanned again by the reading scanner 36 to create a color histogram, and a plurality of representative colors are determined from the histogram. Then, the determined representative colors are displayed on the color CRT 38. FIG. 4 is an explanatory diagram showing an example of a method of dividing a color space into subspaces when creating a color histogram.
In the example of FIG. 4, the RGB space is divided by a predetermined unit length L along each coordinate axis, and is divided into a cubic unit space Sp. In step S1a, first, each unit space Sp
Is obtained, a histogram Hst (i, j, k) of a three-dimensional array indicating the frequency of appearance of colors in the color original is obtained. Here, i, j, and k are coordinate axes whose directions are equal to the R, G, and B coordinates, and whose unit coordinate value is equal to the unit length L. For example, if each of the colors R, G, and B is expressed in 255 gradations, and the unit section L is 5, i,
The j and k coordinates take values in the range of 0 to 51, respectively. now,
The luminance value (R) of each of the RGB colors of the nth pixel in the color original
When n, Gn, Bn) is (30, 120, 200), the corresponding i, j, and k coordinate values are 6 (= 30/5), 24 (= 120/5), 40 (= 2
00/5), the histogram value Hst (6, 24,
1 is added to 40). At the same time, (i, j,
Average value Rave of each color component in unit space at position k)
(I, j, k), Gave (i, j, k), Bave (i,
j, k) is updated and calculated according to the following equation 1.

[0027]

(Equation 1)

By performing the above-described processing for all the pixels of the color original, a three-dimensional histogram Hst (i, j, k) indicating the frequency of appearance of the color in each unit space and the color of each unit space are obtained. Average values Rave (i, j, k), Gave (i,
j, k) and Bave (i, j, k).

The unit length L for dividing the unit space
Is not limited to 5, and any value can be adopted. However, if the unit length L is a value of a power of 2, the upper few bits of each coordinate axis of RGB represented by a plurality of bits become the coordinate values (i, j, k) of the histogram. There is an advantage that it is easy.

Further, it is not necessary to perform the above-described processing for all the pixels of the color original, and the above-described processing may be performed every few pixels by thinning out the image of the color original. In this case, the pre-scan in step S1a may be realized by scanning the color original again, or may be realized by thinning out the image data read in step S1.

Histogram H for the entire color original
When st and the average colors Rave, Gave, Bave are obtained, the average color (Rave, Gave, Bave) at the position (i, j, k) where the appearance frequency (the number of pixels) is equal to or more than a predetermined threshold value is representative. Color is determined. The plurality of representative colors determined in this way are displayed on the color CRT 38 in the order of appearance frequency. FIG. 5 is an explanatory diagram illustrating a display example of a plurality of representative colors. On the right half of the monitor screen, color patches indicating a plurality of representative colors are arranged, and on the left half, a thinned-out image of a color document painted with the representative color is displayed.

The average color of the unit space may be simply displayed as the representative color in the descending order of the appearance frequency without determining whether the appearance frequency is equal to or higher than the predetermined threshold value. Also in this case, the user can easily select the representative color having a high appearance frequency as the target color or the non-target color.

In steps S2 and S3, the user designates a target color to be clipped and a non-target color from a plurality of representative colors displayed on the color CRT 38, respectively. For example, yellow as the background color of the clothes in FIG. 1A is designated as the target color, and the skin color of the model is designated as the non-target color. The designation of a color is performed by selecting a representative color patch displayed on the color CRT 38 using a pointing device such as the mouse 34. in this way,
By simultaneously displaying the color patches of a plurality of representative colors and the color image painted with the representative colors, the task of designating the target color and the non-target color can be easily performed without requiring skill. Non-target colors can be specified. In addition, since the representative color is a color included in the color image, there is an advantage that the possibility of mistakenly specifying the target color or the non-target color is low.
Furthermore, when designating the target color and the non-target color, it is sufficient to display the thinned image painted with the representative color, and there is an advantage that it is not necessary to display the actual image without thinning as it is.

Instead of calculating the average color of each unit space, a color corresponding to a predetermined position in each unit space may be used as a color representative of each unit space. For example, a color corresponding to the center of each unit space or a color corresponding to a predetermined corner of each unit space can be used as a color representing the unit space. However, if the average color of each unit space is calculated, there is an advantage that a correct color included in a color document can be obtained as a representative color.

Since the routines of steps S1a to S9 are repeatedly executed, if it is necessary to distinguish between the target color and the non-target color in each cycle of this routine, it is specified in the first cycle. "1st target color" and "1st non-target color"
Call.

FIG. 6 is an explanatory diagram showing the positions of target colors and non-target colors on the Munsell hue circle. The yellow specified as the first target color and the skin color specified as the first non-target color are located close to each other on the hue circle. According to the present invention, even when the hues of two colors are close to each other, it is possible to set one color as a target color of the mask and the other color as a non-target color of the mask. In the following, unless otherwise specified, the case where yellow and flesh color are designated as target colors and non-target colors will not be specified, and will be generally described.

In step S4, the matrix operation unit 50
(FIG. 2) calculates the color vector of the target color and the color vector of the non-target color, and performs a process of obtaining a matrix (matrix) described later.

FIG. 7 is a flowchart showing a detailed procedure of step S4. In step S11, a first color vector is obtained by normalizing the color vector of the target color. FIG. 8 is a graph showing the first and second color vectors. Each of the RGB components of the target color Csub is represented by (R1, G
1, B1), the vector representing the target color Csub is represented as Vsub (R1, G1, B1). FIG.
For simplicity of illustration, the G axis in the RGB space is omitted.
In step S11, the matrix operation unit 50 determines that the unit vector V1 (r1, g1,
b1), and the components (r1, g1, b1) are stored in the vector value memory 24. Here, the unit vector is a vector having a length of one. Similarly, in step S12, the color vector Vnsub (R2, G2, B2
) Is determined, and a unit vector V2 (r2, g2, b2)
The component is stored in the vector value memory 24.

The first and second color vectors V1 and V2 obtained in this manner constitute two of the three vectors constituting the basis for expressing the color in the RGB space. As the third color vector V3 for forming the basis, a unit vector independent of the first and second color vectors V1 and V2 is selected. That is, the third color vector is the first color vector.
Is a unit vector that cannot be expressed by a linear combination of the second color vector V1 and the second color vector V2. For example, the third color vector V3
Is selected as the unit vector VWH representing white.
The component of the white unit vector VWH is (1 / {3,1 / √)
3,1 / √3).

FIG. 9 shows the first to third color vectors V
FIG. 3 is an explanatory diagram illustrating a relationship between 1, V2, and V3 and an RGB space. In this embodiment, as shown in the following equation 2, three linearly independent color vectors V1, V2, and V3 are used as a basis, and an arbitrary color E (Re, Ge, Be) in the RGB space is used.
Is represented by a linear combination (linear combination) of three color vectors V1, V2, and V3.

[0041]

(Equation 2)

As will be described later, mask data representing a mask is created based on a coefficient k1 for a first color vector V1 (ie, a color vector for a target color). By the way, when one of the first and second color vectors V1 and V2 coincides with the third color vector V3, the three color vectors are not linearly independent.
An arbitrary color cannot be represented by the three vectors V1 to V3. Also, the first and second color vectors V1, V
Even when one of the two is sufficiently close to the third color vector V3, the coefficient k1 for the first color vector V1 may become excessively large or small. Therefore, in step S13 of FIG.
It is checked whether one of the color vectors V1 and V2 is sufficiently close to the white unit vector VWH (1 / √3, 1 / √3, 1 / √3). If they are sufficiently close, in step S14, a vector other than the white unit vector VWH is set as the third color vector V3.

First, in step S13, the first and second color vectors V1 and V2 are converted into an HSV coordinate system (hue / saturation / lightness coordinate system) according to the following equation (3).

[0044]

(Equation 3)

In equation 3, the operators “max” and “m”
"in" indicates an operation that takes the maximum value and the minimum value of the values in parentheses, respectively. According to Equation 3, the first color vector V1
(R1, g1, b1) and the second color vector V2 (r2,
g2, b2), the hue value H, lightness value V, and saturation value S are obtained.

FIG. 10 is an explanatory diagram showing the HSV space. As can be seen from FIG. 10, the white unit vector VWH exists on the coordinate axis of the lightness V. In other words, the white unit vector VWH is a vector having no saturation S and no hue H. The HSV space is a cylindrical coordinate system (zrθ coordinate system)
Where the lightness V corresponds to the vertical coordinate z, the saturation S corresponds to the diameter r, and the hue H corresponds to the angle θ.

A color vector having a small saturation value S is close to a white unit vector VWH (1 / √3, 1 / √3, 1 / √3). Accordingly, the first and second color vectors V1, V
If at least one of the saturation values S is equal to or less than a predetermined value, it can be determined that the color vector is sufficiently close to the white unit vector VWH. As the predetermined value, for example, each component is 8-bit data, that is, 0 to 255
10 (0A in hexadecimal)
h) The degree is appropriate.

Instead of the above method, when the ratio S / V between the saturation value S and the lightness value V is equal to or less than a predetermined value, it is determined that the color vector is sufficiently close to the white unit vector VWH. You may make it.

If at least one of the first and second color vectors V1 and V2 is sufficiently close to the white unit vector VWH, the white unit vector V3 is set as the third color vector V3 in step S14. Vectors other than VWH are determined. FIG. 11 is an explanatory diagram showing a method of setting a vector other than the white unit vector VWH as the third color vector V3. First, a color vector farther from the white unit vector is selected from the first and second color vectors V1 and V2. In the example of FIG. 11, the saturation value S
Is selected, the first color vector V1 is relatively large. Then, 90 is subtracted from the hue value H of the selected color vector V1. That is, the color vector V2 is rotated clockwise by 90 ° in the HSV space. The vector (H-90, S, V) obtained in this way is inversely transformed into an RGB coordinate system, and the RGB components are normalized, whereby a third color vector V3 as a unit vector is obtained. FIG. 11 illustrates the third color vector V3 obtained by such an operation. Note that the rotation angle can be set to any angle other than 90 °.

Since the third color vector V3 obtained by the above procedure is independent of the first and second color vectors V1 and V2, the three color vectors V1 to V3 can be used as a basis. In other words, any color in the RGB color space is represented by the three color vectors V1
ＶV3. The third color vector V3 is a vector indicating a color that actually exists because the third color vector V3 is obtained by rotating the hue of the first or second color vector. Since all three color vectors V1 to V3 are vectors indicating actual colors, the coefficient k1 for each color vector
The value of ~ k3 will have a realistic meaning representing the mixture ratio of the three colors.

Since a vector independent of the first and second color vectors V1 and V2 may be selected as the third color vector, the third color vector may be set by various methods other than those described above. Can be. For example, the outer product V1 × V2 of the first and second color vectors V1 and V2 is converted to the third color vector V3.
It is also possible to use as. However, this cross product V
In some cases, 1 × V2 does not represent a real color. In this case, the values of the coefficients k1 to k3 for each color vector have no practical meaning. Therefore, as described above, the first
And by rotating the hue of the second color vector
It is preferable to obtain the color vector of

When the third color vector V3 is determined in this way, in step S15 (FIG. 7), a 3 × 3 matrix M composed of three color vectors V1 to V3 is used.
(See Equation 2) is created and stored in the matrix memory 26.

The coefficient K (k1, k2, k3) for an arbitrary color E (Re, Ge, Be) is given by the following equation (4).

[0054]

(Equation 4)

Therefore, if the inverse matrix M ^-1 of the matrix M is obtained, the coefficient K (k1, k2, k) can be obtained by multiplying the arbitrary color E (Re, Ge, Be) by the inverse matrix M ^-1.
3) can be obtained. Therefore, in step S16, the matrix calculation unit 50 calculates the inverse matrix M ⁻¹ ,
It is stored in the matrix memory 26.

When the processing in step S4 of FIG. 3 is completed, a multi-tone mask is created in step S5. FIG. 12 is a flowchart showing a procedure for creating a multi-tone mask. In step S21, the multi-tone mask calculation unit 52 reads out the pixel value E (Re, Ge, Be) of each pixel of the color image in the original image memory 22, substitutes the value into the right side of Expression 4 to obtain its coefficient K (k1 , K2, k3).
In step S22, the obtained coefficient K (k1, k2, k
The coefficient k1 of the first color vector V1 is extracted from 3) and stored in the multi-tone mask memory 28 in step S23.

The first coefficient k1 indicates the intensity of the first color vector V1 in the color of each pixel. In other words,
The value of the first coefficient k1 increases as the color of each pixel includes more components of the first color vector V1. Therefore, the bitmap data representing the first coefficient k1 for all the pixels in the color image can be used as a multi-tone mask indicating a color region similar to the target color.

When the value of the first coefficient k1 is negative, the value of the coefficient k1 may be replaced with 0 because the pixel has a color opposite to the color to be clipped.

The first coefficient k1 of each pixel is set to the first coefficient k1a (= R1 / r1 = G1 / g1) for the target color.
= B1 / b1) divided by kk (= k1 / k)
It is also possible to use 1a) as a multi-tone mask. At this time, if the value of kk is 1 or more, it may be replaced with 1, and a negative value may be replaced with 0. By doing so, the value of the bitmap data representing the multi-tone mask can be contained in the range of 0 to 1, so that 2
Setting of the threshold value in the binarization process is facilitated.

Alternatively, three coefficients k for each pixel
The values of 1, k2, and k3 may be normalized according to the following Expression 5, and the bitmap data of the normalized coefficient ku1 may be used as a multi-tone mask.

[0061]

(Equation 5)

Note that the coefficient ku1 also has a negative value of 0
Is preferable. The coefficient ku1 obtained in this way can also be set within the range of 0 to 1, so that
Setting of the threshold value in the binarization process is facilitated.

An application of the multi-tone mask is, for example, a watermark synthesizing process. In this case, the intensity of the target color can be used as a synthesizing ratio. In addition, by referring to the gradation value of the multi-tone mask, it is possible to weight or weaken the image processing performed on the cutout target portion. For example, when performing a contour emphasis process on a target portion of an image, a strong contour emphasis can be applied to a portion having a high tone value of the mask, and a weak contour emphasis can be applied to a portion having a low tone value.

When the multi-tone mask is created in this way, in step S6 in FIG. 3, the binarization calculation section 54 binarizes the multi-tone mask to create a binary mask. As a method of setting a threshold value used for binarization, the following various methods can be considered.

FIG. 13 is a flowchart showing a procedure for creating a binary mask while manually inputting a threshold value. When the user inputs a threshold value through the interactive processing (step S31), the binarization calculation unit 54 binarizes the multi-tone mask using the threshold value to create a binary mask (step S32). ). The user has access to color C
Observe the binarized mask displayed on the RT 38 and determine whether or not the mask is satisfactory (steps S33 and S3).
4). Note that when displaying the binary mask, FIG.
It is preferable to display a color image and a binary mask side by side as in 1 (B). If the binary mask is not satisfactory, the process returns to step S31, inputs a new threshold value, and executes the binarization process again. On the other hand, if the binary mask is satisfactory, the binarization process ends.
Creating a binary mask according to the procedure of FIG. 13 has an advantage that a mask desired by a user can be easily created.

It is also possible to automatically set the threshold for binarization based on the distribution of the multi-tone mask. FIG.
Is a histogram of multi-tone mask data. FIG.
Is the multi-tone mask value of the target color. In a normal color image, the number of pixels having a color close to the target color to be clipped is usually quite large, and therefore, a peak of the histogram appears near the mask value of the target color.
Further, since the background portion of the image has a dark color, a peak also appears in a portion where the mask value is small. In such a case, as shown in FIG. 14, the appearance frequency shows the minimum value at a value x smaller than the mask value Kst of the target color. Therefore, it is smaller than the mask value Kst for the target color, and
If a mask value x whose appearance frequency indicates a minimum value is obtained and used as a threshold value for binarization, a binary mask representing a color portion close to the target color can be created.

As shown in FIG. 15, when the maximum value of the multi-tone mask value is set to a predetermined value Kmax and all values above Kmax are replaced by Kmax, the mask value with which the appearance frequency becomes the minimum value is obtained. x can be determined more easily. When the multi-tone mask value is represented by 8-bit data, a value of about 128 is appropriate as the value Kmax.

As a method for obtaining the threshold value for binarization,
Based on the histogram of the multi-tone mask values, it can be determined using a so-called non-hierarchical clustering method (also called a rearrangement method or k-mean method). If a cluster in the histogram is obtained by this method and the lower limit value of the cluster of the target color is used as a threshold value, a binary mask indicating a color portion similar to the target color can be obtained.

FIG. 16 is a graph showing a color region cut out by the binary mask. The color in the region (Ra + Rb) is a color in which the value of the coefficient k1 of the first color vector V1 is equal to or greater than Ta. The color in the region Rb is a color in which the value of the coefficient k1 of the first color vector V1 is equal to or larger than Tb. Therefore, if the binarization threshold is set to Ta, the region (R
It is possible to create a binary mask indicating the color portion in (a + Rb). At this time, the region R adjacent to the left side of the region Ra
The color in c is excluded from the mask area because the component of the target color is small. Note that the threshold value for binarization is T
If it is set to b, a binary mask indicating the color portion in the region Rb can be created.

FIG. 17 is an explanatory diagram showing the effect of the first mask created by executing steps S1a to S7 of FIG. 3 once. Many circles in FIG. 17 indicate the color distribution in the color image. This color image has a first set Ob1 of pixels of a color close to the target color Csub.
And a second set Ob2 of pixels of colors close to the non-target color Cnsub. The first set Ob1 has a color of a portion (so-called highlight area) where light rays are reflected and shined during photographing. For example, in the color original shown in FIG. 1A, a shining portion of yellow clothing corresponds to the highlight area. According to the above-described embodiment, by setting the threshold value to Ta in the figure, a binary mask including the first set Ob1 including the highlight region and excluding the second set Ob2 is obtained. be able to. Since both the highlight region in the first set Ob1 and the second set Ob2 have a large B component, the conventional method using the color vector in the RGB space uses the highlight region and the second set Ob2. It is impossible to separate On the contrary,
In the method according to this embodiment, the highlight area close to the target color can be easily separated from the second set Ob2.

In the color image shown in FIG.
If the color yellow of the cloth is specified as the first target color Csub and the skin color of the model is specified as the first non-target color Cnsub, the mask shown in FIG. 1B showing the yellow portion of the cloth can be obtained. it can. As can be seen from the Munsell hue circle shown in FIG. 6, green is a hue on the opposite side of the first non-target color when viewed from the first target color. Therefore, the first target color C
If sub is designated as yellow and the first non-target color Cnsub is designated as skin color, the green leaf portion is also included in the mask area. That is, as shown in FIG. 1B, a binary mask having the yellow and green portions as mask regions can be obtained.

As described above, steps S1a to S1 in FIG.
9, the first cycle of the target color Csu
b and the first non-target color Cnsub,
A mask that includes a color portion close to the first target color Csub and does not include a color close to the first non-target color Cnsub can be easily obtained.

When the first mask has been created in this way, in step S7, the image data of the color document and the binary mask data are combined to display the cutout target area on the color CRT 38. FIG. 18A is an explanatory diagram showing a color image in which a cutout target area and a non-cutout target area are displayed separately. In FIG. 18A, the area hatched by a broken line is displayed in gray. In this example, only the area of the first mask (the area to be cut out) in the color original shown in FIG. , And other areas are displayed in multi-tone gray.
Note that the first mask region is a yellow and green region as shown in FIG. Various methods other than those described above can be considered as a method for distinguishing the cutout target area (mask area) from the cutout target area. For example, an area that is not to be clipped may be displayed in black and white binary.

The user observes the cutout target area displayed on the color CRT 38 and determines whether or not a desired mask has been obtained. If the desired mask has not been obtained, the process returns from step S8 in FIG. 3 to step S1a,
The routine of steps S1a to S7 is repeated. At this time, the determination of the representative color in step S1a is performed on the color image portion in the cutout target area, the designation of the second target color and the second non-target color in steps S2 and S3, and the processing in steps S4 to S7 Is also executed for the color image portion in the cutout target area. For example, as shown in FIG. 6, the same yellow as the first target color is designated as the second target color, and green is designated as the second non-target color. Since the first mask is a yellow and green area as shown in FIG. 18A, by specifying green as a non-target color in this mask area, a yellow color is obtained as shown in FIG. A second mask containing only the region is obtained. If the second mask is not the desired mask, steps S1a to S1a are performed.
By repeatedly executing S8, a mask including only colors close to the desired target color can be generated.

When step S1a is executed again, it is not necessary to scan the color original again, and only the image data of the color image portion in the cutout target area among the image data obtained in advance is used. To determine the representative color.

When a desired mask is obtained in this manner, the process proceeds from step S8 to step S9, and a mask (at least one of a binary mask and a multi-tone mask) is output.

B. Modifications of First Embodiment: (1) In the first embodiment, the RGB coordinate system is used as the color space, but other colors such as the CIE-XYZ coordinate system, the YMC coordinate system, and the YUV coordinate system are used. A space may be used.

(2) The first coefficient k1 or ku1 obtained by the above formula 4 or 5 is used as it is as the multi-tone mask data, and the value of the binary mask is set for the pixel where the multi-tone mask data is positive. May be set to 1, and the value of the binary mask may be set to 0 for a negative pixel. That is, 0 may be used as the threshold. In particular, when creating a binary mask for changing the color of a color image, a good mask can be obtained by such a method.

(3) The first to third color vectors need not be unit vectors. However, if the first to third color vectors are unit vectors, each color vector is a vector representing only the hue. Therefore, if a mask is created according to the first coefficient as described above, it is possible to determine the mask area according to the hue of each pixel of the color image.

(4) In the above embodiment, the same color as the first target color is specified as the second target color. However, the target color is always the same without specifying the second target color. May be assumed. Further, a different color may be designated as a target color in each cycle of steps S1a to S8.

(5) As a method of dividing a color space into subspaces when obtaining a representative color, various methods can be considered in addition to the above method. FIG. 19 is an explanatory diagram showing another division method of the subspace. The graph HR on the R axis in FIG. 19 is a histogram of the R component of the color original. Similarly, a graph HG on the G axis and a graph HB on the B axis are histograms of the G component and the B component, respectively. When determining the representative color, the color space is divided at the valley (minimum value) of the histogram HR for the R component. As a result, in the example of FIG. 19, the R axis is divided into three regions RR1, RR2, and RR3. Similarly, the G component and the B component are also divided by the valleys of the histograms HG and HB. For example, when a histogram of the G component is created after the histogram of the R component is created, a histogram of the G component is created in a color space divided by each region on the R axis, and a histogram of the B component is further created. When a histogram of components is created, a histogram of B components is created in a color space partitioned by regions on the R axis and the G axis. If there is only one peak in the histogram, there is no need to classify the coordinate axis of the color component. Thus, by dividing the coordinate axes of each color component, the three-dimensional color space is divided into a plurality of subspaces. The appearance frequency of each subspace color and its average color are calculated in the same manner as in the first embodiment. Then, the average color of the partial space whose appearance frequency is equal to or higher than the predetermined threshold value is determined as the representative color.

As described above, by dividing the color space near the valley (minimum value) of the histogram of each color component to determine the subspace for obtaining the representative color, the color space can be set to an appropriate boundary. There is an advantage that it can be divided by.

When the color space is divided as shown in FIG. 19, any three linearly independent coordinate axes other than the RGB axes can be used. For example, it is also possible to determine the main components of the color of the color image and divide the color space along three main component axes in the same manner as in FIG.

C. Second embodiment: Step S1a in FIG.
In each cycle of S8 to S8, it is also possible to specify n target colors and m non-target colors. In this case as well, the procedure for creating the mask is the same as the procedure shown in FIG.
The procedure is the same as that of the embodiment.

In steps S 2 and S 3, the user selects n from a plurality of representative colors displayed on the color CRT 38.
The target color of the cutout and the m non-target colors are specified. Here, n and m are integers, and either one is at least 2 or more. For example, yellow as the background color of the clothes in FIG. 1A is designated as a target color, and green as a rose leaf color and a skin color of a model are designated as non-target colors. In this case, n = 1 and m = 2. The designation of the color is performed by selecting a representative color patch (FIG. 5) displayed on the color CRT 38 using a pointing device such as the mouse 34. FIG. 20 is an explanatory diagram showing the positions of the target color and the non-target color specified in the second embodiment.

In step S4, the matrix operation unit 50
(FIG. 2) calculates the color vector of the target color and the color vector of the non-target color, and performs a process of obtaining a matrix (matrix) described later.

The detailed procedure of step S4 is the same as the procedure in the first embodiment shown in FIG. Step S11
Then, n first color vectors are obtained by normalizing the color vectors of the n target colors. FIG. 21 (A)
FIG. 21B is a graph showing a first color vector and a second color vector, respectively. RGB of target color Csub1
Let (RA1, GA1, BA1) be each component of the target color Csu
The vector representing b1 is represented as Vsub1 (RA1, GA1, BA1). In FIGS. 21A and 21B, for convenience of illustration, the G axis in the RGB space is omitted. Step S11
, The matrix operation unit 50 calculates the color vector V
The unit vector VA1 (rA1, gA1, bA1) of sub1 is obtained,
The components (rA1, gA1, bA1) are stored in a vector value memory 24.
To memorize. Here, the unit vector is a vector having a length of one.

Similarly, in step S12, the color vectors Vnsub1 (RB1, GB1, BB1) of the two non-target colors,
Vnsub2 (RB2, GB2, BB2) unit vector VB1 (r
B1, gB1, bB1) and VB2 (rB2, gB2, bB2),
The component is stored in the vector value memory 24.

The thus obtained n first color vectors VAi (i = 1 to n) and m second color vectors VBj
The n × m combinations of (j = 1 to m) form n × m sets of bases for expressing colors in the RGB space.
As the third color vector VCij of the basis of each set, a unit vector independent of the first and second color vectors VAi and VBj is selected. That is, the third color vector is a unit vector that cannot be expressed by a linear combination of the first and second color vectors VAi and VBj. For example, a unit vector VWH representing white is selected as the third color vector VCij. The component of the white unit vector VWH is (1 / {3,1 / √)
3,1 / √3).

FIG. 22 shows the first to third color vectors V
FIG. 4 is an explanatory diagram showing a relationship between Ai, VBj, VCij and an RGB space. In the second embodiment, as shown in the following Expression 6, using each of n × m sets of bases composed of three linearly independent color vectors VAi, VBj, and VCij, an arbitrary color space in the RGB space is used. The color E (Re, Ge, Be) is represented by a linear combination (linear combination) of the three color vectors VAi, VBj, VCij of each set.

[0091]

(Equation 6)

As will be described later, mask data representing a mask is created based on the coefficient k1 (i, j) for the first color vector VAi (that is, the color vector for the target color). By the way, the first and second color vectors VA
If any one of i and VBj matches the third color vector VCij, the three colors are no longer linearly independent, so any color can be assigned to the three vectors VAi, VBj,
It cannot be represented by VCij. Further, even when one of the first and second color vectors VAi and VBj is sufficiently close to the third color vector VCij, the coefficient k1 (i, j) for the first color vector VAi is not changed. May be too large or too small. Therefore, in step S13 of FIG. 7, the first and second color vectors VA
i or VBj is a white unit vector VWH (1
/ √3,1 / √3,1 / √3). And if they are close enough,
In step S14, a vector other than the white unit vector VWH is set as the third color vector VCij.

First, in step S13, the first and second color vectors VAi and VBj are converted to HS
It is converted into a V coordinate system (hue / saturation / lightness coordinate system).

According to Equation 3, the first color vector VAi
(RAi, gAi, bAi) and the second color vector VBj (rBj,
gBj, bBj), the hue value H, lightness value V, and saturation value S are obtained.

If at least one of the first and second color vectors VAi and VBj is sufficiently close to the white unit vector VWH, in step S14, the white unit vector is set as the third unit color vector VCij. Vectors other than VWH are determined.

The third color vector VCij obtained by the above-described procedure is independent of the first and second color vectors VAi and VBj, so that the three color vectors VAi, VBj and VCij can be used as a basis. In other words, an arbitrary color in the RGB color space is represented by a linear combination of the three color vectors VAi, VBj, and VCij, as in Equation (2). Further, the third color vector VCij is a vector indicating a color that actually exists, since the third color vector VCij is obtained by rotating the hue of the first or second color vector. Three color vectors VAi, V
Since Bj and VCij are both vectors indicating the actual color,
Coefficient k1 (i, j), k2 for each color vector
(The values of ((i, j), k3 (i, j) will have a realistic meaning representing the mixture ratio of the three colors.)

As the third color vector, a vector independent of the first and second color vectors VAi and VBj may be selected. Therefore, the third color vector may be set by various methods other than those described above. Can be. For example, the cross product VAi × VBj of the first and second color vectors VAi, VBj is converted to the third color vector VCi.
It can be used as j. However, the outer product VAi × VBj may not represent an actual color, and in this case, the values of the coefficients k1, k2, and k3 for each color vector have no practical meaning. Therefore, as described above, it is preferable to determine the third color vector by rotating the hues of the first and second color vectors.

In the following, three color vectors VAi, V
A base composed of Bj and VCij is referred to as a “base BB (i,
j) ".

When the third color vector VCij is determined in this way, in step S15 (FIG. 7), each base BB
A 3 × 3 matrix M (i, j) (Equation 6) composed of components of three color vectors VAi, VBj, VCij of (i, j)
Is created and stored in the matrix memory 26.

The coefficient K (i, j) for an arbitrary color E (Re, Ge, Be) is given by the following equation (7).

[0101]

(Equation 7)

Accordingly, if the inverse matrix M (i, j) ^-1 of the matrix M (i, j) of each base BB (i, j) is obtained, an arbitrary color E (Re, Ge, Be) can be obtained. This inverse matrix M
By multiplying by (i, j) ⁻¹ , the coefficient K (i, j)
Can be requested. Therefore, in step S16,
The matrix calculation unit 50 calculates n × m sets of bases BB (i, j)
^Is computed, and the result is stored in the matrix memory 26.

When the processing in step S4 in FIG. 3 is completed, a multi-tone mask is created in step S5. FIG. 23 is a flowchart showing a procedure for creating a multi-tone mask in the second embodiment. Step S41
Then, the parameters i and j are initialized to 0, and in steps S42 and S43, j and i are each incremented by one.

In step S44, the multi-tone mask calculating section 52 reads out the pixel value E (Re, Ge, Be) of each pixel of the color image in the original image memory 22, substitutes it into the right side of Equation 7 to obtain i × Coefficient K for the first base BB (i, 1)
(I, 1) is obtained. If the parameter i is smaller than n in step S45, steps S43 and S44
repeat. Thus, n sets of coefficients K (i, 1) (i = 1 to n) for n sets of bases BB (i, 1) for the same second color vector VB1 are obtained. FIG.
FIG. 9 is an explanatory diagram showing a procedure for obtaining multi-tone mask data for one pixel value. Steps S43 to j = 1
By the processing of S45, the n coefficients k1 (1,1) to k
1 (n, 1) is obtained.

In step S46, the coefficients k1 (1,1) to k1 (n, 1 of the first color vector VAi are selected from the obtained n sets of coefficients K (i, 1) (i = 1 to n). )
The maximum value km (i) = max {k1 (i, 1)}
Ask for.

FIG. 25 shows that the maximum value km (i) = max ｛k
FIG. 4 is an explanatory diagram showing a reason for obtaining 1 (i, 1)}. First
The coefficient k1 (i, j) of the color
The intensity of the first color vector VAi when represented by (i, j) is shown. In other words, as the color E of the pixel includes more components of the first color vector VAi, the first coefficient k1 (i, i,
The value of 1) is large. In FIG. 25, two coefficients k1 (1,1) and k1 (2,1) for the same second color vector VB1
1) is shown. Two first color vectors VA1, VA2
Of the first with the larger coefficient k1 (1,1)
Is closer to the pixel color E than the other first color vector VA2. Therefore, when a plurality of target colors are specified, the maximum value k among n coefficients k1 (i, 1) (i = 1 to n) for the same second color vector VB1
By calculating m (i) = max {k1 (i, 1)}, the coefficient k when the color E is represented by the basis including the color vector representing the target color closest to the color E of the pixel.
1 can be obtained.

As described above, when the maximum value km (1) of the first coefficient for j = 1 is obtained, step S4 in FIG.
At 7, it is determined whether the parameter j is less than m.
If j is less than m, the parameter i is reset to 0 in step S50, and the above-described steps S42 to S42 are performed.
S47 is repeated.

Thus, for each value of the parameter j (j = 1 to m), the maximum value km (j) (j = 1 to 1) of the first coefficient
When m) is obtained, in step S48, these m
Value min ｛km of the maximum values km (j)
(J)｝ is obtained, and this minimum value is stored in the multi-tone mask memory 28 as multi-tone mask data kmask (see FIG. 24).

FIG. 26 is an explanatory diagram showing the reason why the minimum value min {km (j)} is adopted as the multi-tone mask value kmask. In the example of FIG. 26, the same first color vector VA1
And two coefficients k1 (1,1) and k1 (1,2) when two second color vectors VB1 and VB2 are combined, respectively.
2) is shown. These coefficients k1 (1,1), k1
(1, 2) is the maximum value km obtained in step S46.
(1), km (2). As can be seen from FIG. 26, the color E of the pixel is closer to the color indicated by the second color vector VB1 on the right side than the second color vector VB2 on the left side.
In such a case, the coefficient k1 (1,1) when the right second color vector VB1 indicating a closer color is used becomes smaller. If there is a second color vector that is closer to the color E than the second color vector on the right, its coefficient k
1 becomes smaller. As described above, the value of the coefficient k1 decreases as the second color vector approaches the color E of the pixel.
By the way, a color close to the second color vector is a color that is desired to be recognized as being out of the target of the mask area. Therefore, a plurality of second
By using the minimum value among a plurality of first coefficients k1 (1,1) and k1 (1,2) obtained when the color vectors VB1 and VB2 of the above are used as the multi-tone mask value kmask, A color close to any of the plurality of non-target colors may not be included in the mask area.

As described above, the maximum value km of the n first coefficients k1 for the same second color vector is obtained, and the minimum value of the m maximum values km obtained in this manner is obtained. With the multi-tone mask value kmask, it is possible to set a portion of a color close to n target colors as a mask region and exclude a portion of a color close to m non-target colors from the mask region.

By executing the procedure shown in FIG. 23 for each pixel of the color image, multi-tone mask data kmask for each pixel can be obtained. This multi-tone mask data kmask does not approximate to m non-target colors and n
It is bitmap data indicating an area of a color similar to the number of target colors.

When the multi-tone mask value kmask is negative, the pixel has a color directly opposite to the color to be clipped, so that the multi-tone mask value kmask may be replaced with zero.

The multi-tone mask value kmask of each pixel obtained as described above is divided by the multi-tone mask value kmaska (= R1 / r1 = G1 / g1 = B1 / b1) for the target color. The obtained value kka (= k1 / kmaska) can be used as a final multi-tone mask value. At this time, if the value of kka is 1 or more, it may be replaced with 1, and a negative value may be replaced with 0. This allows the value of the bitmap data representing the multi-tone mask to be within the range of 0 to 1, thereby facilitating the setting of the threshold value in the binarization processing described later.

Alternatively, three coefficients k for each pixel
The values of 1, k2, and k3 may be normalized according to the above-described Expression 5, and the bitmap data of the normalized coefficient ku1 may be used as a multi-tone mask.

Incidentally, also for the coefficient ku1, a negative value is set to 0.
Is preferable. The coefficient ku1 obtained in this way can also be set within the range of 0 to 1, so that
Setting of the threshold value in the binarization process is facilitated.

The normalization according to Equation 5 is performed in step S44 in FIG. 23 by using the coefficients k1, k2 for the base BB (i, j).
It may be executed each time k2 and k3 are obtained.
It may be executed after the multi-tone mask value kmask of each pixel is determined in 48.

When the multi-tone mask is created in this way, a binary mask is created in step S6 in FIG. 3, and the cutout target area is displayed in color in step S7.

If the desired mask has not been obtained, the process returns from step S8 to step S1a, and the routine is repeated in steps S1a to S8 described above.
This is the same as in the first embodiment described above.

In the first embodiment, since only the skin color is designated as the first non-target color Cnsub and yellow is designated as the first target color Cnsub, the green leaf portion is also designated by the first mask (FIG. 1).
(A)). On the other hand, in the second embodiment, since skin color and green can be designated as the first non-target color, only the yellow portion can be obtained as the first mask. Further, if steps S1a to S8 are executed a plurality of times, it is possible to create a mask including only a color very close to the target color even for a color image including an extremely large number of colors.

As described above, in the second embodiment, by specifying n target colors Csub and m non-target colors Cnsub, a color portion close to the n target colors Csub is included, and , M non-target colors Cnsub can be easily obtained.

The processing of the second embodiment can be considered to include the following steps (a) to (g). (A) determining a plurality of representative colors based on a statistical distribution of colors in a color image and displaying the plurality of representative colors on a display device; and (b) the mask from among the plurality of representative colors. A step of designating n target colors and m non-target colors (at least one of n and m is 2 or more). (C) obtaining a first color vector of each of the n target colors in a predetermined color space; (D) obtaining a second color vector of each of the m non-target colors in the predetermined color space; (E) For each of the n × m combinations of the first and second color vectors, a third color vector independent of the first and second color vectors is obtained, and the first to third color vectors are obtained. Creating n × m sets of bases by combining the color vectors. (F) For each of the n × m sets of bases, the color of each pixel in the color image is represented by a linear combination of the first to third color vectors with respect to the first to third color vectors. 1st to 3rd
A step of obtaining the coefficients of (G) creating, for each pixel in the color image, the mask according to the value of the first coefficient obtained for each of the n × m sets of bases;

Note that the above steps (a) and (b) are executed again, and a set of n ′ target colors and m ′ non-target colors in the color image portion in the mask area covered by the mask Can be newly specified, and the steps (c) to (g) can be executed for the color image portion.

In the above steps (a) to (f),
Each of the number n of target colors and the number m of non-target colors can be an integer of 1 or more. At this time, in the first embodiment, n =
This corresponds to a special case where m = 1.

The present invention is not limited to the above-described embodiment, but can be implemented in various modes without departing from the gist of the invention.

[0125]

As described above, according to the first aspect of the present invention, even when the target color and the non-target color have similar hues, a color portion close to the target color is used as a mask area, and Can be excluded from the mask area. In addition, since a plurality of representative colors are obtained based on the statistical distribution of the colors of the color image, and the target color and the non-target color are designated from among them, the target color and the non-target color are easily specified. be able to.

According to the second aspect of the present invention, colors having a high appearance frequency in a color image can be easily designated as target colors or non-target colors.

According to the third aspect of the present invention, since the subspace is divided according to the frequency of appearance of the color in the color image, a plurality of representative colors representing the main colors existing in the color image are obtained. Can be requested.

According to the fourth aspect of the present invention, it is possible to easily generate binary mask data indicating only a mask area.

According to the invention described in claim 5, the threshold value for binarization can be automatically determined according to the histogram.

According to the invention described in claim 6, the third color vector can be easily determined.

According to the invention described in claim 7, the third color vector independent of the first and second color vectors can be easily determined. In addition, since the first to third color vectors each represent an actual color, the first coefficient for the color of each pixel can be used as an index that actually represents the component of the target color.

According to the invention described in claim 8, since the first to third color vectors are vectors representing only hues, a mask can be created according to the hues of each color image.

[Brief description of the drawings]

FIG. 1 is an explanatory view showing a color photographic document and its mask.

FIG. 2 is a block diagram showing a configuration of a mask creating system that creates a cutout mask by applying an embodiment of the present invention.

FIG. 3 is a flowchart showing an overall procedure for creating a cutout mask in the embodiment.

FIG. 4 is an explanatory diagram showing a color space division method used for creating a histogram.

FIG. 5 is an explanatory diagram showing a display example of a representative color.

FIG. 6 is an explanatory diagram showing positions of a target color (yellow) and a non-target color (skin color) on a Munsell hue circle.

FIG. 7 is a flowchart showing a detailed procedure of step S4.

FIG. 8 is a graph showing first and second color vectors V1 and V2.

FIG. 9 shows first to third color vectors V1, V2, and V3.
FIG. 4 is an explanatory diagram showing the relationship between the RGB space.

FIG. 10 is an explanatory diagram showing an HSV space.

FIG. 11 is an explanatory diagram showing a method of setting a vector other than a white unit vector VWH as a third color vector V3.

FIG. 12 is a flowchart showing a procedure for creating a multi-tone mask.

FIG. 13 is a flowchart showing a procedure for creating a binary mask.

FIG. 14 is a histogram of multi-tone mask data.

FIG. 15 is a histogram of multi-tone mask data.

FIG. 16 is a graph showing a color region cut out by a binary mask.

FIG. 17 is an explanatory diagram showing the effect of the mask created by the example.

FIG. 18 is an explanatory diagram showing a first mask region display and a second mask.

FIG. 19 is an explanatory diagram showing another division method of a subspace.

FIG. 20 is an explanatory diagram showing positions of a target color (yellow) and a non-target color (skin color) in the second embodiment.

FIG. 21 is a graph showing first and second color vectors VAi and VBj in the second embodiment.

FIG. 22 is an explanatory diagram showing a relationship between first to third color vectors VAi, VBj, VCij and an RGB space in the second embodiment.

FIG. 23 is a flowchart showing a procedure for creating a multi-tone mask in the second embodiment.

FIG. 24 is an explanatory diagram showing a procedure for obtaining multi-tone mask data for one pixel value in the second embodiment.

FIG. 25 shows the maximum value km (i) = m in the second embodiment.
FIG. 4 is an explanatory diagram showing the reason for obtaining ax {k1 (i, 1)}.

FIG. 26 shows a minimum value min ｛km in the second embodiment.
(J) Explanatory diagram showing the reason why｝ is multi-tone mask data.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 20 ... CPU 22 ... Original image memory 24 ... Vector value memory 26 ... Matrix memory 28 ... Multi-tone mask memory 32 ... Keyboard 34 ... Mouse 36 ... Reading scanner 38 ... Color CRT 40 ... Hard disk drive 50 ... Matrix computing unit 52 ... Many Tone mask calculation unit Cnsub: Non-target color Csub: Target color

────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-4-61558 (JP, A) JP-A-8-153201 (JP, A) JP-A-5-197804 (JP, A) JP-A-8-183 7106 (JP, A) JP-A-8-7107 (JP, A) JP-A 8-96114 (JP, A) JP-A 8-115431 (JP, A) JP-A 8-272942 (JP, A) Tominaga, G., "Color Classification of Color Images and Its Applications," Journal of the Japan Society of Color Science, 1991, Vol. 15, No. 1, p. 55-56 (58) Field surveyed (Int. Cl. ⁶ , DB name) G06T 7/00

Claims (8)

(57) [Claims] 1. A method for creating a mask indicating a desired color portion in a color image, comprising: (a) determining a plurality of representative colors based on a statistical distribution of colors in the color image; (B) displaying a representative color on a display device;
Specifying a target color of the mask from the plurality of representative colors; (c) specifying a non-target color of the mask from the plurality of representative colors; Obtaining a first color vector of the target color;
(E) obtaining a second color vector of the non-target color in the predetermined color space; (f) obtaining a third color vector independent of the first and second color vectors;
(G) the first pixel when the color of each pixel in the color image is represented by a linear combination of the first to third color vectors;
And (h) creating the mask according to the value of the first coefficient at each pixel in the color image. A method for producing a mask for a color image, comprising: 2. The method according to claim 1, wherein said step (a) comprises: dividing a color space into a plurality of subspaces; and determining a color in the color image in each of the plurality of subspaces. A mask for a color image, comprising: a step of calculating an appearance frequency; and a step of displaying, as the plurality of representative colors, a color representing at least a part of the plurality of partial spaces according to the occurrence frequency. How to make. 3. The method according to claim 2, wherein the step of dividing the color space into a plurality of subspaces comprises: obtaining a histogram for each component of a three-dimensional color system with respect to a color in the color image. And a step of setting a boundary near a local minimum value of the histogram of each component to divide the color space into the plurality of subspaces divided by the boundary. 4. The method according to claim 1, wherein said step (h) comprises: binarizing said first coefficient with a threshold value;
A method for creating a mask for a color image, comprising creating mask data representing the mask. 5. The color image mask creating method according to claim 4, wherein the step of binarizing the first coefficient includes the step of creating a histogram indicating an appearance frequency of the first coefficient; Setting the first coefficient at the minimum value of the histogram as the threshold value. 6. The color image mask creating method according to claim 1, wherein said third color vector is a vector representing white. 7. The method for producing a mask for a color image according to claim 1, wherein said step (f) comprises: converting said first and second color vectors into hue / saturation / brightness. Converting the color components into fourth and fifth color vectors in a space, respectively, if at least one of the saturation components of the fourth and fifth color vectors exceeds a predetermined value, the white color in the predetermined color space Is selected as the third color vector, and when the saturation component of at least one of the fourth and fifth color vectors is less than the predetermined value, the fourth color vector is selected.
And a sixth color vector obtained by rotating the hue by a predetermined angle in the hue / saturation / brightness space for a color vector having a larger saturation component among the fifth color vector and the sixth color vector. Generating the third color vector by converting to a vector in the predetermined color space. 8. The color image mask creating method according to claim 1, wherein the first to third color vectors are unit vectors.