KR102001944B1 - Method for drawing map of color interaction network - Google Patents

Abstract

A method of making a color interactive network map is disclosed. The present invention relates to a method of producing a color interactive network map for analyzing divided results by applying a bisectional algorithm to an image, the method comprising the steps of: (a) horizontally dividing horizontal and vertical Searching for an optimal partitioning position; (b) continuously dividing the image to be analyzed until a predetermined size is reached; (c) finding and pairing adjacent pieces for each piece; And (d) finding the most frequent color between the pair of pieces and connecting the two colors to a line.

Description

METHOD FOR DRAWING MAP OF COLOR INTERACTION NETWORK [0002]

The present invention relates to a method of producing a color interactive network map, and more particularly, to a method of creating a color interactive network map that allows a user to observe color interactions of a local region in units of blocks recognized by a person .

To create a harmonious image, use the same series of colors as a whole, or use colors that are complementary to surrounding colors to emphasize a point. The way colors are used is a major discourse in terms of design. So people have been studying what 'color' is and how to use colors harmoniously from the past. Early color theory began in the 18th century, studying light, and developed by Newton, Geothe, Young, and Maxwell. However, research on dispersion and blending of light and color wheel using basic color have been presented. Then, in the modern color theory of the 20th century, systematic color space and color space for classifying colors were presented, and the harmonious color composition was examined. A representative example is J. Itten's 12-color circle. He introduced a new kind of color wheel that describes color harmony based on the color location on the color wheel.

Since then, Y. Matsuda has based on Itten's color wheel, and has introduced 80 defined color schemes using psychological studies to combine several types of color distribution. Using this, M. Tokumaru presented eight harmonic templates for harmonic evaluation and color design. This pattern was rearranged by D. Cohen-Or, defining the angle and color harmonic scheme.

Conventionally, a color matching pattern for an entire image has been calculated. In the present invention, however, a method for drawing an interaction network between colors in a local area using a configuration information amount is presented. This is to analyze the interactions between adjacent colors using the pieces that are divided according to the amount of constituent information.

Existing color interactions were analyzed only in whole image or pixel unit. Because it differs from the cognitive range of the human, it sometimes does not reflect the characteristics of the image. In addition, the existing image segmentation method is calculated in RGB space, and there is a limitation in that it can not grasp the relation of color such as complementary color relation in RGB color space.

1. Korean Patent Publication No. 10-2005-0062834 (June 28, 2005) 2. Korean Patent No. 10-0408508 (November 24, 2003)

1. Color Harmonization: Daniel Cohen-Or, Olga Sorkine, Ran Gal, Tommer Leyvand, Ying-Qing Xu, Tel Aviv University Microsoft Research Asia 2. Color harmonization for images: Zhen Tang; Zhenjiang Miao; Yanli Wan; Zhifei Wang 3. Informational Aesthetics Measures; Jaume Rigau; Miquel Feixas; Mateu Sbert 4. Dual segmentation algorithm for informational and aesthetic indices; Shin, In - Sup; Han, Seung - Gi; Lee, Geon - Myung; Lee, Seung - Bok; Jung 5. Space division of Western art works using composition information amount; Seo, Min - Kyung; Shin, In - Sup; Kim Dong Hyun; Han Seung-ki

It is an object of the present invention to provide a method of producing a color interaction network map that enables to analyze color interactions of a local region in units of blocks perceivable by a human.

In order to accomplish the above object, a method of producing a color interaction network map according to the present invention is a method of creating a color interaction network map, in which a dual division algorithm is applied to an image so as to observe color interactions of a local region in units of blocks perceived by a person A method of producing a color interactive network map for analyzing a divided result by applying,
(a) an image is transformed into an HSV color space, and horizontal division in the horizontal direction and vertical division in the vertical direction are performed while shifting by one pixel with respect to all the pixels of the image, thereby selecting the point having the greatest amount of configuration information as the division position ; (b) selecting a large piece among the two pieces divided by the selection of the division position, searching for a division position where the amount of configuration information becomes maximum in the same manner as in the step (a), and performing horizontal division or vertical division; (c) continuously dividing the image to be analyzed until a predetermined size is reached; (d) finding and pairing adjacent pieces for each piece; And (e) traversing the paired pieces to find the most frequent color in each piece, and connecting the two colors to a line in a color distribution histogram represented by an angle,
Wherein the configuration information amount can be acquired for at least one parameter of hue, saturation, and brightness.

delete

식(1)
In the step (a), a point at which the configuration information amount I (C, R ₁ ) obtained by the following formula ( ₁ ) is the largest is selected as the first division position,
Wherein the upper piece or the left piece is represented by 1, and the lower piece or the right piece is represented by 2.

Equation (1)

Where π _a is the width of the pixel of the piece a, H (a) is the color information amount of the piece a, C is the color variable, and R ₁ is the first space division.

delete

delete

In the step (a), before proceeding to the second division, the largest fragment among the remaining fragments is selected, and a point having the greatest amount of configuration information is found and divided in the same manner as the first division, 3. ≪ / RTI >

delete

식(2)
Characterized in that when the entire image to be analyzed is divided into {R ₁ , R ₂ , ...., R _n }, the amount of constituent information is obtained by the following equation (2).

Equation (2)

Here, π _i is i, and pixels (pixel) width of the second piece, and the color information amount of H (r _i) is a piece of r _i, and C is the color variable, R ₁ is the first space division, R ₂ is a second tessellation And R _n is the n-th spatial division.

delete

delete

When a person recognizes an image, the unit is not an entire image nor a pixel unit. It is recognized as a block unit of a size that is not too large or too small to be distinguished by a not too small eye. However, in the past, we observed only the distribution of color for the entire image or the color distribution for each pixel. In other words, we observed interactions between colors in units that are irrelevant to human cognitive units. However, according to the present invention, there is an effect that the color interactions of the local region can be observed in block units perceived by a person as image division information.

Also, if we observe the entire image at once, we may not be able to observe the effect of the color we want. For example, suppose you wear a red shirt and a blue tie or necklace. The handbag or tie is very small compared to the size of the clothes, so when you analyze the whole image, you may not observe the desired contrast effect.

However, according to the present invention, when a method of dividing an image is applied to the above-described example and observed in a block unit, an intense contrast effect can be observed. In other words, we can better observe the characteristics of the image through interaction between colors in the local area, and it will help to design costume, magazine or homepage.

FIG. 1 is a diagram showing that the distribution of colors before and after division in an RGB color space is changed when an image is divided, and the amount of color information is changed accordingly.
Fig. 2 (a) shows positions for horizontal and vertical division, Fig. 2 (b) shows the amount of constituent information obtained from each horizontal and vertical division, and FIG.
FIG. 3 shows a model of an HSV color space, where H denotes a color Hue, S denotes a saturation, and V denotes a value.
FIG. 4 (a) shows eight color matching patterns preferred by people, and FIGS. 4 (b) and 4 (c) are histograms of color distribution histograms of a hat figure of a gogh.
Fig. 5 is a graph of the configuration information obtained from all the horizontal and vertical divisions in order to find the optimum division position when the first division of the gogh sunflower picture is performed. 1 < / RTI > piece, and the underside is a double piece.
FIG. 6 is a view showing an example of dividing the sunflower figure of gogh 30 times and a piece adjacent to the 1-piece.
Figure 7 depicts a histogram of the color distribution of the 5th piece and the 30th piece. 219 ° for the 5th piece and 33 ° for the 30th piece. The right side of the line shows the interaction of these two angles with a line.
FIG. 8 shows V. Gogh's <Six Sunflowers> divided into 30 times and represented by a color interaction network map.
FIG. 9 shows the division of J. Brueghel's <Vase of Flowers with Jewellery, Coins and Shells> by 30 times, and a color-interaction network map.

Just as there is a pleasant chord when we listen to music, there is a set of harmonious colors that we prefer to look aesthetically. To find a set of more harmonious colors used more in art, let us explain the information-theoretic method of analyzing color interactions between pieces in a picture. To do this, first, the image is divided into nearly uniform blocks through the spatial division process. Next, a color interaction network map is drawn by obtaining the form of color interaction between all neighboring parts. In the present invention, this analysis method will be described for any artwork and the results will be compared.

In the present invention, in order to examine the interrelationship between colors in a local region as well as an analysis of a color matching pattern for an overall color distribution, a divided result is analyzed by applying a bisection algorithm to the image.

Before describing the present invention in detail, color information palette information, image segmentation and compositional information, HSV color space, and color harmonic pattern will be described.

<Color information amount (palette information)>

For a given two-dimensional image, the color diversity can be defined using 'shannon entropy'. Shannon entropy is simply the amount of 'how rare events are defined'. The color information amount H (C) can be obtained by the following equation (3). Where P (c) is the probability of using the color c in the overall color variable C.

식(3)

Equation (3)

The higher the value, the more uniform the color. The smaller the value, the more the specific color is used. If only one color is used, the amount of color information is zero. Also, the total amount of information that can be obtained from an image is exactly the same as the amount of color information.

<Image segmentation and compositional information amount>

The configuration information amount I (C, R ₁ ) obtained when the first spatial division R ₁ is introduced into the two-dimensional image using the color C as shown in FIG. 1 is given by the above-mentioned equation (1). Let us obtain the configuration information amount I (C, R ₁ ) at this time. (C 2, R ₁ ) = 17.27 - (0.53 X 17.03 + 0.47 X 15.80) = 0.82. That is, the amount of information obtained by the first spatial division is 0.82. In FIG. 1, when the RGB space of the square shape is viewed, it can be seen that the colors are divided after the division. A maximum of 1 bit can be obtained by dividing once, and the larger the amount of information obtained by the space division, the more clearly the color is divided.

However, since the amount of information greatly varies depending on the position to be divided, the following is performed to find an optimal space division position. First, calculate the amount of information for all the horizontal and vertical divisions by moving one pixel from the top. Then, the position with the largest amount of information is selected as the split position.

For example, as shown in Fig. 2 (a), the configuration information amount is calculated by performing space division at one pixel position from above. Then, calculate the amount of information when the pixel is shifted downward by 1 pixel and finally divided horizontally from the top (vertical length - 1) pixels. Vertical division is likewise shifted by 1 pixel and divided for all cases. Figure 2 (b) shows the graph obtained by calculating the amount of information for all cases . Horizontally at 456 pixels, the largest amount of information.

From the second split, select the largest piece that is not monochromatic. Then, we find and divide the best segmentation position that maximizes the amount of information as we did in the first segmentation. Proceed in the same way for the third and fourth. And when all the pieces are completely solid, the division ends. At this time, the general configuration information amount is obtained by the above-mentioned expression (2).

<HSV color space>

A color space is a method of displaying a color as coordinates like RGB. RGB means x (red), x (green), y (green) and b (blue), and these three coordinates are used to represent a specific color. The RGB color space borrows an orthogonal coordinate system, that is, it looks like a cube. The HSV model we use is also a kind of color space, H stands for Hue, S stands for Saturation, and V stands for Value. In other words, the coordinate system is drawn by the axis of color, saturation, and brightness. HSV borrows a cylindrical coordinate system, and can be expressed in the shape shown in FIG. The main feature of this color space is that it displays the color at an angle between 0 ° and 360 °. 0 ° is red, and the complementary blue is 180 °. And 360 ° becomes red again.

&Lt; Harmonic pattern >

The color matching pattern preferred by many people is proposed to have the eight patterns {i, V, L, I, T, Y, X, N} shown in FIG. In this pattern, i, L, I, and Y have a width of 18 °, which is 5% of the total disk, and V, Y, and X are 93.6 °, 26% of the disk. L is 79.2 °, 22%, and T is 180 °, which is half. The Van Gogh sunflower work shown in Fig. 4 (b) is close to the L pattern, and (c) is expected to be close to the I pattern.

Hereinafter, a method of producing a color interactive network map according to the present invention will be described in detail with reference to the accompanying drawings.

First, in the HSV image, the hue is made an integer from 0 to 360 ° in 1-degree intervals. In the present invention, the interval is 1 degree for fast calculation, but it may be any interval such as 2 degrees or 3 degrees or a prime number. In addition, the division will proceed with only H. H must be included, but S or V may or may not be included.

Referring to FIG. 5, horizontal division in the horizontal direction and vertical division in the vertical direction are executed for all the pixels. And the point where the constituent information amount I (C, R ₁ ) is largest is selected as the dividing position. At this time, the amount of the configuration information in the first division can be obtained as shown in the above-mentioned equation (1). The left or right piece is 1 and the remaining piece is 2.

And, before proceeding to the second division, find the largest piece of the remaining pieces first. In the same way as the first partition, the point with the greatest amount of configuration information is found and divided. At this time, the newly formed piece is divided into three pieces. In general, the amount of configuration information that can be obtained by segmentation can be obtained by the above-mentioned equation (2).

Thereafter, the image is divided until it becomes a predetermined size as shown in Fig. We split it only 30 times for explanation. And numbered from 1 to 31 in the order of division.

 First, each piece is searched for and adjoined to adjacent pieces. For example, a piece adjacent to a piece 1 is a piece 3, 21, 22, 31. You can create a total of four pairs of pieces in a pair of adjacent pieces that you find, pair 1 and 3, pair 1 and 21, pair 1 and 22, pair 1 and 31. This is done up to 31 times in this way. There are a total of 158 fragment pairs here.

Now, we draw a color interaction network map using the method of finding and connecting the most frequent colors among the pair of pieces. For example, if you refer to pairs 5 and 30, then 219 ° for 5 pieces and 33 ° for 30 pieces. Because HSV space, colors were expressed in degrees. FIG. 7 is a histogram of how the color distribution is in the 5th and 30th pieces, and the line between the two angles is continued.

In the same way as above, find the most frequent color in every pair of sculptures, and line between the two colors. There are a total of 158 pairs of sculptures, so I drawn 158 times. Then, as shown in FIG. 8, a color interaction network map for this picture can be obtained.

FIG. 9 shows a color interaction network map created by dividing J. Brueghel's work Vase of Flowers with Jewelery, Coins and Shells in the same manner as above.

The artwork of FIG. 9 showed 154 pairs of pieces when dividing 30 times, and it appeared similar to the T pattern when the network map was drawn in the same way. However, in each case, it can be called L pattern. Here, 44 pairs of i patterns with 0 ° ± 18 angle between two pieces, 42 pairs of L patterns with 90 ° ± 40, 180 ° ± 18 I pattern is 0 pair. That is, in this figure, the same or similar colors were placed in similar places, and a color of about 90 ° away was used for adjacent pieces.

Using the result of the division of art works, we analyzed the interaction of colors between adjacent pieces. First, the relationship between colors between adjacent pieces is shown by a network map. As a result, in the case of the V. Gogh still life in FIG. 6, i-pattern using a similar color and I-pattern using a complementary color were mainly appeared. In the case of J. Brueghel's still life in Fig. 9, i-patterns also appeared, but many L-patterns also appeared. In addition, the result of calculating the similarity for the I and L patterns is the same as that of the color network map. V. Gogh's figure is predominantly I-patterned, and J. Brueghel's figure is predominantly L-patterned. In this way, we were able to analyze and compare the relationships between colors in different pictures.

Using this analysis method, we can analyze the changes of color interactions by analyzing art works by period. Or cover designs such as magazines, and analysis of fashion photographs.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments, Of the right.

Claims (4)

There is provided a method of producing a color interactive network map for analyzing a divided result by applying a dual division algorithm to an image so as to observe color interactions of a local region in a block unit perceived by a person,
(a) an image is transformed into an HSV color space, and horizontal division in the horizontal direction and vertical division in the vertical direction are performed while shifting by one pixel with respect to all the pixels of the image, thereby selecting the point having the greatest amount of configuration information as the division position ;
(b) selecting a large piece among the two pieces divided by the selection of the division position, searching for a division position where the amount of configuration information becomes maximum in the same manner as in the step (a), and performing horizontal division or vertical division;
(c) continuously dividing the image to be analyzed until a predetermined size is reached;
(d) finding and pairing adjacent pieces for each piece; And
(e) finding the most frequent color in each piece with respect to the paired pieces, and connecting the two colors to a line in a color distribution histogram represented by an angle,
Wherein the configuration information amount can be obtained for at least one parameter of hue, saturation, and brightness. The method according to claim 1,
In the step (a)
The point at which the constituent information amount I (C, R ₁ ) obtained by the following equation (1) is the largest is selected as the first division position,
Wherein the upper piece or the left piece is represented by 1, and the lower piece or the right piece is represented by 2.

Equation (1)
Where π _a is the width of the pixel of the piece a, H (a) is the color information amount of the piece a, C is the color variable, and R ₁ is the first space division. 3. The method of claim 2,
In the step (a)
Before proceeding to the second division, the largest fragment among the remaining fragments is firstly selected, and a point having the greatest amount of configuration information is searched and divided in the same manner as the first division, and the newly generated fragment is displayed as 3 How to make color interactive network map. The method of claim 3,
Characterized in that when the entire image to be analyzed is divided into {R ₁ , R ₂ , ...., R _n }, the amount of constituent information is obtained by the following equation (2).

Equation (2)
Here, π _i is i, and pixels (pixel) width of the second piece, and the color information amount of H (r _i) is a piece of r _i, and C is the color variable, R ₁ is the first space division, R ₂ is a second tessellation And R _n is the n-th spatial division.

Images