JP3139617B2 - Portrait making method and device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a portrait creating method and apparatus for automatically creating a portrait based on a face image of a target person obtained by, for example, a video camera or an electronic still camera. In particular, the relative positional relationship between the face part images, and the size and shape of the face part images themselves,
BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a caricature creation method and apparatus capable of automatically creating an exaggerated face image which is an essential feature of a caricature by exaggerating a measured value of a face of a target person.

[0002]

2. Description of the Related Art Conventionally, as an apparatus used for creating a face image such as a portrait, for example, Japanese Patent Application Laid-Open No. Hei 6-324671 is disclosed.
A "graphic display device" described in Japanese Patent Application Laid-Open Publication No. H10-209,985 is known.
In this apparatus, in short, an appropriate part image is selected from an image database of face parts such as eyebrows, eyes, nose, and mouth, and the selected part image is displayed on a plain face image without face parts. It is to be arranged in.

That is, in the apparatus, a parameter D indicating the direction of a portrait in which a portrait is inserted in the background, an enlargement ratio M for performing enlargement / reduction processing of the portrait, a portrait in the background. Are respectively selected to indicate the position at which the is inserted. Then, the portrait synthesis result data is synthesized with the position indicated by the position parameter Cn on the background data and displayed on the display device. Thereby, the portrait is input from the front, and then the portrait data of a different angle is automatically generated simply by selecting the direction parameter D corresponding to the display angle data d, and the portrait is easily matched with the content of the background data. In addition, the size of each figure is appropriately adjusted.

[0004]

However, in such a conventional apparatus, the place where the face part image is arranged is the same as the place of the fixed or real part, and therefore, the essential feature of the portrait. There was a problem that a certain exaggerated face image could not be created.

[0005] In addition, in such a conventional device,
There is a problem that many face part images must be prepared in advance so as to be able to handle various human face parts.

[0006] The present invention has been made in view of such conventional problems.
It is an object of the present invention to provide a portrait creating method and apparatus capable of automatically creating an exaggerated face image which is an essential feature of a portrait.

Another object of the present invention is to provide a caricature creating method and apparatus which can eliminate the need to prepare many face parts in advance even when dealing with various human face parts. It is in.

[0008]

According to a first aspect of the present invention, there is provided a face image input means for inputting a face image, and a face image input from the face image input means by analyzing the face image. A face feature extraction means for extracting a face feature which is a feature related to the shape and arrangement of the part, and a face part data storage for storing a plurality of face part data prepared in advance for each type of face part Means for extracting, from the face part data storage means, face part data corresponding to the shape feature amount for each type of face part using a feature amount relating to the shape of the face part; Change parameter input means for inputting a change parameter for controlling a change of a position at which a face part represented by the extracted face part data is arranged; and a change of a face part in the face feature amount based on the change parameter. Arrangement Face part arranging means for correcting the amount of collection and arranging a face part pattern corresponding to the face part data extracted by the face part data extracting means at a position indicated by the corrected arrangement feature amount. The feature is in the portrait creation device.

Here, the "arrangement feature amount" means a feature amount at which an arrangement position or an arrangement position can be generated. In an embodiment described later, X1, Y
1, Eye height and the like correspond to this.

According to the first aspect of the present invention, it is possible to arrange face parts so as to emphasize the features of the face, so that an exaggerated face image, which is an essential feature of a portrait, is automatically generated. Can be created.

In the invention according to claim 2 of the present application, the change parameter includes gender information of a portrait creation target person and / or information on whether the portrait creation target person is an adult. In the portrait creation device.

According to the second aspect of the present invention, a more exaggerated caricature can be automatically created according to gender and adult / children.

According to a third aspect of the present invention, there is provided a face feature amount extracting step of analyzing a given face image to extract a face feature amount which is a feature amount related to the shape and arrangement of a face part. From a plurality of pieces of face part data stored in advance for each type of part, face part data corresponding to the shape feature amount is extracted for each type of face part by using a feature amount related to the shape of the face part. A face part data extraction step; a change parameter input step of inputting a change parameter for controlling a change of a position at which a face part represented by the extracted face part data is to be arranged; and the face based on the input change parameter. The arrangement feature amount of the face part in the feature amount is corrected, and the face part pattern corresponding to the face part data extracted in the face part data extraction step is indicated by the corrected arrangement feature amount. In the portrait creation method characterized by comprising a face component arrangement step of arranging the location, the.

According to the third aspect of the present invention, it is possible to arrange face parts so as to emphasize the features of the face, so that an exaggerated face image which is an essential feature of a portrait is automatically generated. Can be created.

In the invention described in claim 4 of this application, the change parameter includes information on the gender of the portrait creation target person or whether the portrait creation target person is an adult.
In the caricature creation method described in.

According to the present invention, a more exaggerated caricature can be automatically created according to gender or adult children.

According to a fifth aspect of the present invention, there is provided a face image inputting means for inputting a face image, and a face part characteristic amount extracting means for extracting a characteristic amount of a face part from the input face image. A face part data storage unit for storing a plurality of face part data prepared in advance for each type of face part; and a face part data used in caricature creation from the face part data storage unit. A face part extraction unit for extracting each type of face part and a correlation for correcting the size or shape of the face part represented by the extracted face part data based on the feature amount of the face part are generated. Correlation generating means, and a portrait synthesizing means for synthesizing a portrait using a facial part obtained by correcting the size or shape of the facial part data read from the facial part data storage means by the correlation. Characteristic In the portrait creating apparatus for.

According to the fifth aspect of the present invention, the shape and size of the face part can be changed and exaggerated.
An impressive portrait can be created, and the storage amount of the face part image can be reduced.

According to a sixth aspect of the present invention, in the fifth aspect, the correlation generating means generates the correlation based on the gender of the person whose portrait is to be created or whether or not it is an adult. In the portrait creation device.

According to the invention described in claim 6, it is possible to easily create a child-like face, a feminine face, and the like.

According to a seventh aspect of the present invention, there is provided a face image input means for inputting a face image, and a feature quantity relating to the shape and arrangement of the face parts by analyzing the input face image. A face feature amount extraction unit for extracting a face feature amount, a face part data storage unit for storing a plurality of face part data prepared in advance for each type of face part, and a face part data storage unit. Parameter input means for inputting parameters used when selecting face part data to be used in the portrait, and face part data to be used in the portrait, from the face part data storage means for each type of face part based on the parameter; And a face for correcting the size or shape of the face part represented by the extracted face part data on the basis of a feature amount related to the shape of the face part. And goods correcting means, a facial parts obtained by correcting by the facial part modifying means, in the portrait creation device including a portrait synthesis unit, a synthesized portrait arranged based on the feature quantity relating to the arrangement.

According to the seventh aspect of the present invention, since the shape, size and arrangement of the face parts can be changed and exaggerated, a more impressive portrait can be created. Thus, it is not necessary to provide a large number of face parts in advance, and the storage amount of the face part image can be reduced.

The invention described in claim 8 of this application is a face part feature value extracting step of analyzing a given face image and extracting feature values of face parts, and is prepared in advance for each type of face part. A face data product extracting step of extracting face component data used in creating a portrait from each of a plurality of stored face component data for each type of face component; and a face represented by the extracted face component data. A correlation generation step of generating a correlation for correcting the size or shape of the part based on the feature amount of the face part, and the size or shape of the face part represented by the extracted face part data by the correlation. And a portrait synthesizing step of synthesizing a portrait using the face parts obtained by the correction.

According to the eighth aspect of the present invention, since the shape and size of the face part can be changed and exaggerated, a more impressive portrait can be created. Thus, it is not necessary to provide a large number of face parts in advance, and the storage amount of the face part image can be reduced.

The invention according to claim 9 of the present application is the invention according to claim 8, wherein the correlation generating step generates a correlation based on the gender of the person whose portrait is to be created or whether it is an adult. See the caricature creation method described.

According to the ninth aspect, it is possible to easily create a child-like face, a feminine face, and the like.

According to a tenth aspect of the present invention, there is provided a face image input step of inputting a face image for which a portrait is to be created, and analyzing the input face image to determine the shape and arrangement of face parts. A face feature amount extraction step of extracting a face feature amount, which is an amount, and a parameter used when selecting face part data used in a portrait from a plurality of face part data prepared in advance for each type of face part. A parameter inputting step of inputting, a face part extraction step of extracting face part data used in a portrait from the plurality of face part data prepared in advance for each type of face part based on the parameter, A face part correction step of correcting the size or shape of the face part represented by the extracted face part data based on a feature amount related to the shape of the face part; A face part obtained by correcting by the facial part modification step, a portrait synthesis step of synthesizing a portrait arranged based on the feature quantity relating to the arrangement,
A portrait drawing creation method characterized by comprising:

According to the tenth aspect of the present invention, the shape, size, and arrangement of the face parts can be changed and exaggerated, so that an impressive portrait can be created and a face part image for that purpose can be stored. The amount can be saved.

[0029]

Preferred embodiments of the present invention will be described below in detail with reference to the accompanying drawings.

FIG. 1 shows the overall configuration of a portrait creating apparatus according to a preferred embodiment (first embodiment) of the present invention, and FIG. 2 is a flowchart for explaining the operation procedure.

As shown in FIG. 1, the portrait creating apparatus includes a data input unit 1, an image input unit 2, a feature extraction unit 3, a feature storage unit 4, a nonlinear function creation unit 5, Contour image extraction means 6 and component image storage means 7
, Component arrangement means 8 and face image output means 9. These means 1 to 9 are specifically configured as follows.

The data input means 1 is used by a user to input a person parameter and an exaggeration degree parameter, which will be described later, and is constituted by an input device such as a keyboard and a mouse. Here, in this example,
The value of the person parameter (Person_para) is an integer, and specifically, 0 (adult male), 1 (adult female),
(Child). In this example, the distinction between an adult and a child is 15 years or older as an adult and a child under 15 years as a child.

The image input means 2 is for taking in a color face image of a target person into the apparatus. Specifically, the image input means 2 can be constituted by, for example, a video camera, an electronic still camera, or the like. The captured color face image is stored in some storage medium (eg, floppy disk, MO
(A disc, a DVD, or the like), a reproducing apparatus that reproduces a color face image from such a recording medium may be used.

The feature value extracting means 3 is for extracting the feature value of the face of the target person from the color face image input by the image input means 2. Here, the feature amounts include a selection feature amount, an arrangement feature amount, and a scaling ratio (Scale_rati
o). The feature amounts for selection include eye size (Eye_size), eye shape (Eye_shape), nose shape (Nose_shape), mouth size (Mouth_size), mouth shape (Mouth_shape), and eyebrow depth (Brow_thickness) )It is included. The feature amount for arrangement includes the feature amount (X1,
Y1), eye height (Eye_height), eye-nose distance (Eye_no)
se), distance between nose and mouth (Nose_mouth), distance between eyes (Eye_mouth)
space), eye-brow spacing (Eye_brow).
The above-mentioned feature quantity for selection is used for selecting a component image stored in the component image storage means 7, and the feature quantity for arrangement is used for synthesizing the component image on the contour image. .

A detailed description will now be given of a specific method of obtaining the selection feature, the arrangement feature, and the enlargement / reduction ratio performed by the feature extraction means 3 described above.

First, a method for obtaining a feature point of a face contour will be described. Here, the feature points of the face outline (the vertex P of the head)
1, the lower end point P2 of the jaw, the right end point P3 of the jaw, and the left end point P4 of the jaw) are defined as shown in FIG. In order to obtain such feature points, first, the input image (RGB)
Is converted to HSV, and a flesh color region is extracted. Next, the jaw edge is detected by searching the flesh color area from below, and the lowest point of the jaw edge is set to P2. Next, the right end point of the skin color area is set to P3, and the left end point is set to P4. Then
A black area is extracted from the HSV-converted face image, and the highest point is set to P1. After obtaining the above four points (P1 to P4), the face image is converted into a gray shade image.

Next, the feature points (P1
To P4), a method of obtaining the feature points X1 and Y1 of the face contour will be described. xpi is the X coordinate of feature point i, ypi is the feature point i
, The characteristic points X1 and Y1 of the face contour are represented by the following equation (Equation 1).

[0038]

[Number 1] _{X1 = x p4 -x p3 Y1 =} y p1 -y p2 ( Equation 1) Next, inner corner, eye area, nose tip, nose left, nose right end, mouth left, a method for finding the feature points of the mouth rightmost Will be described. here,
These feature points (right eye corner P5, right eye corner P6, left eye corner P6)
7, left eye corner P8, nose head P9, nose right end P10, nose left end P1
1, the right end P12 and the left end P13) are defined as shown in FIG. These feature points are obtained by template matching of the image around the feature points including the feature points. As a template matching method, for example, a normalized density correlation method can be cited. In the normalized density correlation method, a gray shade image is used. That is, the change pattern of the gray density is compared between the target image and the template image, and the more similar the change pattern of the gray density, the greater the degree of matching match.
Then, the point having the highest degree of matching in the search range is extracted as a feature point. At this time, as the template, image data obtained by averaging gray densities in the vicinity of each feature point of the face images of a plurality of persons is used. FIG. 19 shows an example of a specific matching result by template matching.

Next, the feature quantities for eye selection (Eye_size, Eye_s
hape) and a method for obtaining the arrangement feature amount (Eye_height, Eye_space) will be described with reference to FIG. Although only the left eye will be described here, the right eye can be obtained in a similar manner. In FIG. 4, the coordinate value (xp7, yp7) of the inner corner P7 and the coordinate value (xp8, y
p8) has already been given. In the case of the right eye, it is assumed that the inner corner P5 and the outer corner P6 are given.

The plotting conventions in FIG. 4 are as follows (a) to (nu). (A) The vertical center line of the eye is parallel to the Y axis, and the horizontal center line of the eye is parallel to the X axis. (B) The X coordinate of the vertical center line of the eye = (X coordinate value of the inner corner of the eye + X coordinate value of the outer corner of the eye) / 2 (c) Y coordinate of horizontal center line of eye = (Y coordinate value of inner corner of eye + Y coordinate value of outer corner of eye) / 2 (d) EyeSearchArea_X = Y1 × 0.02 (e) EyeSearchArea_Y = Y1 × 0.1 (f) ) Additional brightness at Y coordinate = sum of brightness at corresponding Y coordinate in EyeSearchArea_X (g) Difference of added brightness at Y coordinate = additional brightness at Y coordinate− (Y-1) brightness at coordinate (h) D1 = Maximum value of the added brightness existing at a position below the horizontal center line of the eye and closest to the horizontal center line. (I) D2 = Existing at a position above the horizontal center line of the eye and closest to the horizontal center line. Maximum value of added brightness (nu) D3 = 0.7 × min (D1, D2) However, in the above formula, small Truncate below several points (coordinates are integers)

First, an eye reference horizontal line and an eye reference vertical line are obtained. Here, the reference horizontal line of the eye is (yp7 + yp8)
/ 2 is obtained as a straight line parallel to the X axis, and the reference vertical line of the eye is obtained as a straight line passing (xp7 + xp8) / 2 and parallel to the Y axis. Next, the search range in the X and Y directions of the eyes is determined according to the following equation (Equation 2).

[0042]

[Equation 2] EyeSearchArea_X = Y1 × 0.02 EyeSearchArea_Y = Y1 × 0.1 (Equation 2) Next, FIG. 4 shows (EyeSearchArea_X × EyeSearch)
The luminance values are summed at each Y coordinate within the range of (Area_Y). However, the luminance value is 255-density value. In this example, the density of black pixels is 255, and the density of white pixels is 0. Then, at each Y coordinate, ｛(xp7 + xp8) /
2- (EyeSearchArea_X / 2)｝ ～ ｛(xp7 + xp8) / 2 +
(EyeSearchArea_X / 2) The brightness value in the range of｝ is defined as “additional brightness,” and the defined addition brightness is defined as ｛(yp7 +
yp8) / 2- (EyeSearchArea_Y / 2)｝-｛(yp7 + yp
8) Determine for each Y coordinate within the range of / 2 + (EyeSearchArea_Y / 2)｝. Here, the reason for obtaining the added brightness without using the brightness of one pixel is to prevent the influence of noise. Next, FIG. 4 shows (EyeSearchArea_X × EyeSe
For each Y coordinate within the range of (archArea_Y), the above difference in the added brightness is obtained. The difference obtained at this time is the difference between the added brightnesses of adjacent Y coordinates, as shown in the following equation (Equation 3).

[0043]

## EQU00003 ## Difference = addition brightness of Y coordinate-addition brightness of (Y-1) coordinate (Expression 3) Next, values D1, D2, and D3 shown in FIG. 4 are obtained.
Here, D1 is the value of the maximum point of the added brightness existing below the reference horizontal line and closest to the reference horizontal line, and D2 is the value of the added brightness existing above the reference horizontal line and closest to the reference horizontal line. The value of the maximum point, D3, is a value obtained by multiplying the smaller value of D1 and D2 by 0.7, ie,
D3 = 0.7 × min (D1, D2). Then
A search is made upward from the horizontal center line of the eye to find a point where the added brightness is greater than D3, and a straight line passing through this point and parallel to the X axis is set as EyeLineTop2. Next, a search is made downward from the horizontal center line of the eye to find a point where the added brightness becomes larger than D3, and a straight line passing through this point and parallel to the X axis is defined as EyeLineBot2. Next, a search is made in the vertical direction within a range of Y1 × 0.03 from the straight line EyeLineTop2 to find the maximum point of the difference in brightness. A straight line that passes through the maximum point and is parallel to the X axis is defined as EyeLineTop1. Then, the straight line EyeLineBo
A search is made in the range of Y1 × 0.03 in the vertical direction from t2 to find the maximum point of the difference in brightness. A straight line passing through the maximum point and parallel to the X axis is defined as EyeLineBot1. Next, the distance between two parallel lines, EyeLineTop1 and EyeLineBot1, is set to Left_Eye
Width_Y. Next, Left EyeW
Find idth_X.

[0044]

[Equation 4] Left_EyeWidth_X = X _p7 −X _p8 (Equation 4) Then, for the right eye, Right_EyeWidth_X and Right_EyeWidth_Y are similarly obtained, and the left and right values are averaged according to the following equation (Equation 5).

[0045]

[Equation 5] EyeWidth_X = (Left_EyeWidth_X + Right_EyeWidth_X) / 2 EyeWidth_Y = (Left_EyeWidth_Y + Right_EyeWidth_Y) / 2 (Equation 5) Then, the feature quantity for selecting an eye is calculated by the following equation (Equation 6).

[0046]

[Equation 6] Eye_size = (EyeWidth_X × EyeWidth_Y) / (X1 × Y1) Eye_shape = EyeWidth_Y / EyeWidth_X (Equation 6) Next, the feature amount for arranging eyes is calculated by the following equation (Equation 7).

[0047]

[Equation 7] _{Right_Eye_height = (y p5 + y p6} ) / 2-y p2 Left_Eye_height = (y p7 + y p8) / 2-y p2 Y2 = (Right_Eye_height + Left_Eye_height) / 2 Eye_height = Y2 / Y1 Eye_space = (x p7 -x p5 ) / X1 (Equation 7) Next, the nose selection feature amount Nose_shape and the arrangement feature amount
A method for obtaining Eye_nose will be described. In this case, as shown in the following equation (Equation 8), the left and right inner corners P5,
Right Eye_nose, Lef using Y coordinate value of P7 and nasal head P9
t Find Eye_nose. Next, the calculated Right Eye_no
The se and Left Eye_nose are averaged to obtain Y4. Then
Eye_nose is obtained from the left and right nose end points P10 and P11 and the previously obtained Y4 using Nose_shape and Y2.

[0048]

Equation 8] _{Right_eye_nose = y p5 -y p9 Left_eye_nose =} y p7 -y p9 Y4 = (Right_eye_nose + Left_eye_nose) / 2 Nose_shape = (x p11 -x p10) / Y4 Eye_nose = Y4 / Y2 ( 8) Next, mouth A method for obtaining the selection feature amount Mouth_size, Mouth_shape and the placement feature amount Nose_mouth will be described with reference to FIG. First, the eye features EyeLineTop1, EyeLin
MouthLineTop1, MouthLine
Ask for Bot1. Next, straight lines MouthLineTop parallel to each other
The distance between 1 and MouthLineBot1 is MouthWidth_Y. Next, MouthWidth_X is calculated using the following equation (Equation 9).

[0049]

MouthWidth_X = X _p13 −X _p12 (Equation 9) Then, using the following equation (Equation 10), the mouth selection feature amount Mo is obtained.
Find uth_size and Mouth_shape.

[0050]

Mouth_size = MouthWidth_X / X1 Mouth_shape = MouthWidth_Y / MouthWidth_X (Equation 10) Then, using the following expression (Equation 11), the mouth arrangement feature amount No.
Ask for se_mouth. In Equation 11, ymc is MouthLi
This is the Y coordinate of a straight line that is parallel to neTop1 and MouthLineBot1 and passes through the middle of these straight lines.

[0051]

Equation 11] _{Y5 = y p9 -y mc Nose_mouth =} Y5 / Y2 ( number 11) Next, a method of obtaining the selected feature quantity Brow_thickness and placed feature quantity Eye_brow eyebrow reference to FIG. 6 will be described.
Here, only the case of the left eyebrow will be described, but the same can be obtained for the right eyebrow. First, a search is made on the reference vertical line of the eye upward from the upper line EyeLineTop1 of the eye to find a point at which the luminance becomes minimum. This point is the darkest part and becomes the center of the eyebrows. Next, a straight line that passes through the above minimum point and is parallel to the X axis is obtained, and this is set as a reference horizontal line of the eyebrows. The reference vertical line of the eyebrows is a line obtained by extending the reference vertical line of the eye. Next, BrowLineTop1 and BrowLineBot1 are obtained in the same manner as when eye characteristic amounts EyeLineTop1 and EyeLineBot1 are obtained. Then
The distance between the straight lines BrowLineTop1 and BrowLineBot1 that are parallel to each other is determined, and this is defined as Left_BrowWidth_Y. Similarly, Right_Browwidth_Y is obtained for the right eyebrow. Next, according to the following equation (Equation 12), the selection feature amount Brow_thi
ckness is required.

[0052]

Brow_thickness = (Left_BrowWidth_Y + Right_BrowWidth_Y) / 2 / Y1 (Formula 12) Next, the layout feature amount is obtained according to the following formula (Formula 13). In the following equation, ybL is BrowL of the left eyebrow
Y coordinate value of ineBot1, ybR is Y coordinate value of BrowLineBot1 of right eyebrow, yeL is Y coordinate of EyelineTop1 of left eye, yeR is Eye of right eye
This is the Y coordinate of lineTop1.

[0053]

Equation 13] _{Y3 = (y bL + y bR} ) / 2- (y eL + y eR) / 2 Eye_brow = Y3 / Y2 ( number 13) then determines a scaling factor Scale_ratio according the following equation (equation 14). In the following equation, Scale_base is an integer value set in advance.

[0054]

## EQU14 ## Scale_ratio = (X1 × Y1) / Scale_base (Expression 14) Data output from the feature amount extracting means 3 described above to the feature amount storing means 4, the component image storing means 7, and the component arranging means 8. Is shown in FIG.

As is clear from the drawing, the feature amount extraction means 3 sends the feature amount storage means 4 Eye_height, Eye_s
Pace, Eye_nose, Eye_mouth, Eye_brow are output. In addition, from the feature amount extraction means 3 to the part image storage means 7, Eye_size, Eye_shape, Nose_shape, Mouth_size, Mouth
_shape, Brow_thickness are output. Further, the feature amount extraction means 3 sends the Scale_rati
o, Eye_height, Eye_space, Eye_nose, Nose_mouth, Eye_bro
w, X1, and Y1 are output.

Next, the configuration of the feature amount storage means 4 will be described. The input of this means is a person parameter Person_para and a feature amount for arrangement, and the output is a data string of (person parameter, feature amount for arrangement). The feature amount storage unit 4 stores a pair of a person parameter Person_para and an arrangement feature amount. FIG. 8 shows the structure of each data stored in the feature amount storage unit 4. As shown in FIG. 3, the above-mentioned Person_para, Eye_height, Eye
_space, Eye_nose, Nose_mouth, Eye_brow are stored.

Next, the configuration of the nonlinear function creating means 5 is shown in FIG.
This will be described with reference to the flowchart of FIG. The input of the non-linear function creating means 5 is a data string of the person parameters and the arrangement feature values, and the output is a non-linear function.
Here, the non-linear function is used by the component arranging means 8 to exaggerate the positional relationship between the face contour and the component and the positional relationship between the component and the component. The input of this non-linear function is an arrangement feature obtained from a real object, and the output is a distance parameter between parts of a face image to be created. As the non-linear function, for example, a function such as a curve shown in FIG. 10 in which the value of the distance parameter between components tends to rapidly increase when the arrangement feature value exceeds a certain value is used. As the nonlinear function used in this example, a fuzzy inference rule is adopted. FIG. 11 shows an example of the fuzzy inference rule. In this example, the value of the consequent part of the fuzzy inference rule is set to be properly used depending on the value of the person parameter, and in the same manner as in the feature amount database, for the adult male, for the adult female, and for the child. Are prepared. Fig. 1 shows an example of the value of the consequent part
1 is shown at the right end. As for fuzzy labels, SMALL, MEDIUM,
Three types of BIG are prepared. Each fuzzy inference rule has one input and one output. The input values of the fuzzy inference are the feature amounts (Eye_height, Eye_space, Eye_nose, Nose_mouth, Eye_b) obtained by the feature amount extraction means 3.
row). In the mathematical expressions described later, X used as an input value refers to these feature amounts (Eye_
height, Eye_space, Eye_nose, Nose_mouth, Eye_brow). The output value of the fuzzy inference is a parameter between components.

A method of determining a membership function for each fuzzy label will be described below with reference to the flowchart of FIG. First, an adult man, an adult woman, and a child are stored in the feature amount storage means 4 (step 9).
02 to 905), a data group of the arrangement feature amount is read (step 901), and then each of them is read (step 9031,
9035, 9036), the minimum value MIN, the average value AVE,
The maximum value MAX is obtained (step 9032). Then
Minimum value MIN, average value AVE, maximum value MA obtained above
Using X, each numerical value LEFT,
RIGHT is obtained by calculation (step 9033).

[0059]

[Expression 15] LEFT = (MIN + AVE) / 2 RIGHT = (AVE + MAX) / 2 (Expression 15) Then, using the values of the minimum value MIN, average value AVE, maximum value MAX, LEFT, and RIGHT determined above, SM
All, MEDIUM, and BIG membership functions are obtained (step 9034). In the flowchart of FIG. 9, I is a person parameter, and data of an adult male when I = 0, an adult female when I = 1, and a child when I = 2 is processed. Also,
J indicates the type of feature amount, and when J = 0, Eye_height,
Eye_space when J = 1, Eye_nose when J = 2,
When J = 3, processing is performed on the feature amount of Nose_mouth, and when J = 4, processing is performed on the feature amount of Eye_brow.

That is, the membership function of SMALL is obtained by the following equation (Equation 16) when the input value is X and the output value is μs (X).

[0061]

(Equation 16) The MEDIUM membership function is obtained by the following equation (Equation 17) when the input value is X and the output value is μm (X).

[0062]

[Equation 17] In addition, the BIG membership function uses input values X,
When the output value is μb (X), it is obtained by the following equation (Equation 18).

[0063]

(Equation 18) Next, the configuration of the contour image extracting means 6 will be described. The input of the contour image extracting means 6 is a color face image, and the output is a binary face contour image. First, the contour image extracting means 6 extracts a skin color area of a color face image. Next, the contour image extracting means 6 binarizes the color face image. Next, the contour image extracting means 6 sets the density value inside the area corresponding to the skin color area to 0 in the binarized face image. Some examples of face contours extracted by the contour image extracting means 6 are shown in FIG. As can be seen from the figure, the contour image extracting means 6 can extract the face contour as a figure surrounded by the hairstyle and the contour of the chin for each of the target persons.

Next, the configuration of the component image storage means 7 will be described. The component image storage means 7 stores a plurality of types of right eyebrows, left eyebrows,
Images of facial parts for the right eye, left eye, nose, and mouth are stored as an image database, and these part images are appropriately selected according to the value of the selecting feature amount input from the feature amount extracting unit 3, It is output to the component placement means 8. here,
The part image is a binary image. In addition, the component image storage unit 7 evaluates the value of the selection feature amount input from the feature amount extraction unit 3 using the boundary shown in FIG. 12, and according to the evaluation result, selects the selection feature amount as follows. Change the value of a parameter.

If the value of the feature amount for selection ≦ the boundary 1, the parameter value for selection = 0. If the boundary 1 <the value of the feature amount for the selection ≦ boundary 2, the parameter value for selection = 1 the boundary 2 <the feature value for the selection. If the value is a selection parameter value = 2, a component corresponding to the selection parameter is prepared as the component image, and the component corresponding to the selection parameter is output to the component placement unit 8. That is, as eyebrow parts, three parts are prepared for three selection parameters of Brow_thickness. However, a total of six parts are used for the left eyebrow and the right eyebrow, and the left and right part shapes are targeted for the Y axis. In addition, Eye_size 3
Nine parts are prepared for a combination of three selection parameters and three selection parameters of Eye_shape. However, a total of 18 parts are used for the left eye and the right eye, and the shapes of the left and right parts are targeted with respect to the Y axis. Also,
As nose parts, three parts are prepared for three selection parameters of Nose_shape. In addition, the mouth part has three selection parameters of Mouth_size and
Nine parts are prepared for combinations of three selection parameters of Mouth_shape.

Next, the configuration of the component placement means 8 will be described with reference to the flowchart of FIG. The input of the component arranging means 8 is an arrangement feature amount, an enlargement / reduction ratio, a component image, a contour image, a nonlinear function, a person parameter, and an exaggeration degree parameter. The output of the component arranging means 8 is a face image obtained by synthesizing the component image and the contour image. That is, the component placement unit 8 reads the person parameter and the exaggeration degree parameter from the data input unit 1 (step 1301), and reads the contour image from the contour image extraction unit 6 (step 13).
02), the component image is read from the component image storage means 7 (step 1303), and finally the arrangement feature quantity and the enlargement / reduction ratio are read from the feature quantity extraction means 3 (step 1).
304). Then, based on the read data, the component placement parameters are obtained by inference using fuzzy inference rules (step 1305).

The fuzzy inference method will be described below. When the component placement rules shown in FIG. 11 are generalized, each location is represented by three rules expressed by the following equation (Equation 19).

[0068]

IF X is SMALL THEN FZ is W1 IF X is MEDIUM THEN FZ is W2 IF X is BIG THEN FZ is W3 (Equation 19) Here, the output values of the SMALL, MEDIUM, and BIG membership functions are respectively μs (X), μm (X),
When μb (X) is used, the inter-component distance parameter FZ, which is an output value of fuzzy inference, is represented by a function using W1, W2, and W3 expressed by the following equation (Equation 20).

[0069]

(Equation 20) However, FZ, which is a component placement parameter, is FA for eye height, FB for eyebrow and eye spacing, FC for eye and nose spacing, FD for nose and mouth spacing, FD, left and right. Is the FE, the real value (YY2, YY3, YY4, XX2) for component placement is obtained using these values.

Next, the form of the nonlinear function is changed according to the value of the exaggeration degree parameter (step 1306). When the value of the exaggeration degree parameter is Exaggerate_para, the values of W1 and W3 in the component placement rule are changed. When the values of W1 and W3 before the change are W1_old and W3_old, and the values after the change are W1_new and W3_new, the relationship between the two is expressed by the following equation (Equation 21).

[0071]

W1_new = W1_old × (1−Exaggerate_para) W3_new = W3_old × (1 + Exaggerate_para) where 0 ≦ Exaggerate_para <1 (Formula 21) It should be noted here that Exaggerate_para
Is different between the two formulas. So, Exaggera
The greater the value of te_para, the greater the degree of exaggeration.

Next, each component image is enlarged or reduced by using the value Scale_ratio of the enlargement / reduction parameter among the feature amounts (step 1307). A point (x,
If y) is scaled to change its position to point (X, Y), a relationship represented by the following equation (Equation 22) is established between the two. It should be noted that a person skilled in the art can easily understand a specific conversion method by referring to the description in the reference document “Practical image processing learned in C language” (Ohmsha).

[0073]

Y = Scale_ratio × y, X = Scale_ratio × x (Equation 22) Then, based on the component distance parameter and X1, Y1,
Real values (YY2, YY2) for component placement shown in FIG.
3, YY4, XX2) is obtained by the following equation (Equation 23) (step 1306). However, FA, FB, FC, F
D and FE are inter-component distance parameters which are output values of fuzzy inference.

[0074]

[Equation 23] Eye height YY2 = FA × Y1 Eyebrow and eye distance YY3 = FB × YY2 Eye and nose distance YY4 = FC × YY2 Nose and mouth distance YY5 = FD × YY2 Left and right eye distance XX2 = FE × X1 (Equation 23) Next, based on the obtained real values for component placement, component images are arranged as shown in FIG. 14 (step 1308). However, the components are arranged such that the representative point of each component shown in FIG. That is, the coordinates at which the representative point of each part is arranged are expressed by the following equation (Equation 2).
4). The image created here is 2
It is a value image.

[0075]

[Number 24] _{x center = (x p3 + x} p4) / 2 right _{eyebrow: (x center -XX2 / 2,} y p2 + YY2 + Eye_half_Y + YY3) left _{eyebrow: (x center + XX2 / 2} , y p2 + YY2 + Eye_half_Y + YY3) right eye: (x _center - _{XX2 / 2, y p2 + YY2} ) left _{eye: (x center + XX2 / 2} , y p2 + YY2) _{nose: (x center, y p2 +} YY2-YY4) _{mouth: (x center, y p2 +} YY2-YY4-YY5) ( number 24 Next, the configuration of the face image output means 9 will be described. The face image output means 9 visually outputs the face image synthesized as described above, and specifically, a printing device for outputting a hard copy of a binary image as a face image or a display device for displaying an image. And the like. FIGS. 17 and 18 show some examples of face images output in this manner.
Shown in As is apparent from these figures, by exaggerating the relative positional relationship between the face part images using the measured values of the face of the target person, the exaggerated face image, which is an essential feature of the portrait, is automatically generated. It will be understood that it is made in.

According to the first embodiment,
(1) By exaggerating the positional relationship of parts, an exaggerated face image, which is an essential feature of a portrait, can be automatically created. (2) The face image does not resemble the face of the subject. It has an extraordinary effect such as the ability to prevent excessive exaggeration.

Next, FIG. 20 shows the overall configuration of a portrait creating apparatus according to another preferred embodiment (second embodiment) of the present invention, and FIG. 21 is a flowchart for explaining the operation procedure. Shown in

As shown in FIG. 20, this caricature creation device includes a data input unit 101 and an image input unit 102
, Feature amount extraction means 103, feature amount storage means 104
A component exaggeration nonlinear function creating unit 105, an arrangement exaggeration nonlinear function creating unit 106, and a contour image extracting unit 10.
7, component image storage means 108, component exaggeration means 109
And a part arranging unit 110 and a face image outputting unit 111. These means 101 to 111 are specifically configured as follows.

The data input means 101 is used by a user to input a person parameter, an exaggerated portion selection parameter, an exaggeration degree parameter, etc., which will be described later, and is constituted by an input device such as a keyboard and a mouse. ing.

Here, in this example, the person parameter (Pe
rson_para) is an integer, specifically, 0
(Adult male), 1 (Adult female), 2 (Child). In this example, the distinction between an adult and a child is 15 years or older as an adult and a child under 15 years as a child.

The exaggerated part selection parameter (Parts_se
lect_para) is also an integer value. Specifically,
1 (eye size), 2 (eye shape), 3 (nose shape),
4 (size of mouth), 5 (shape of mouth), 6 (thickness of eyebrows),
7 (eye height), 8 (eye and nose interval), 9 (nose and mouth interval), 10 (left and right eye interval), 11 (eye and eyebrow interval)
It has been decided.

Further, the value of the exaggeration degree parameter is a real value. Specifically, the value of Exaggerate_para, the value of R1, R2
And the case exists.

The image input means 102 is for taking in a color face image of a target person into the apparatus. Specifically, the image input means 102 can be constituted by, for example, a video camera, an electronic still camera, or the like. The captured color face image is stored on some storage medium (for example, floppy disk,
(MO disk, DVD, etc.), a reproducing apparatus for reproducing a color face image from such a recording medium may be used.

The feature amount extracting means 103 includes the image input means 2
This is for extracting the feature amount of the face of the target person from the color face image input at. The feature amount here is
It is composed of a feature amount for component exaggeration, a feature amount for arrangement exaggeration, and a scaling ratio (Scale_ratio).

The feature amount for component exaggeration includes an eye size (Eye_
size), eye shape (Eye_shape), nose shape (Nose_shap
e), mouth size (Mouth_size), mouth shape (Mouth_sha
pe), and the depth of the eyebrows (Brow_thickness).

The layout exaggeration features include face outline features (X1, Y1), eye height (Eye_height), eye-nose distance (Eye_nose), nose-mouth distance (Nose_mouth), eyes and eyes. (Eye_space) and the distance between eyes and eyebrows (Eye_brow) are included.

The above-described feature amount for component exaggeration and the feature amount for placement exaggeration are stored in the feature amount storage unit 104. Among them, the feature amount for component exaggeration is used to exaggerate the size and shape of the component image, and the feature amount for placement exaggeration determines the positional relationship between the component images when the component image is synthesized on the outline image. Used to exaggerate.

Hereinafter, a specific method for obtaining the feature amount for component exaggeration, the feature amount for layout exaggeration, and the enlargement / reduction ratio performed by the feature amount extraction unit 103 will be described in detail. Note that this description will partially overlap with the first embodiment described above, but will be repeated dare to facilitate understanding of the second embodiment.

First, a method for obtaining a feature point of a face contour will be described. Here, the feature points of the face outline (the vertex P of the head)
1, the lower end point P2 of the jaw, the right end point P3 of the jaw, and the left end point P4 of the jaw) are defined as shown in FIG. In order to obtain such feature points, first, the input image (RG
B) is converted to HSV, and a flesh color region is extracted. Then
The jaw edge is detected by searching the skin color region from below, and the lowest point of the jaw edge is set to P2. Next, the right end point of the skin color area is set to P3, and the left end point is set to P4. Next, a black area is extracted from the HSV-converted face image, and the highest point is set to P1. After obtaining the above four points (P1 to P4), the face image is converted into a gray shade image.

Next, the characteristic points (P1
To P4), a method of obtaining the feature points X1 and Y1 of the face contour will be described. xpi is the X coordinate of feature point i, ypi is the feature point i
, The characteristic points X1 and Y1 of the face contour are represented by the following equation (Equation 25).

[0091]

X1 = x _p4 −x _p3 Y1 = y _p1 −y _p2 (Equation 25) Next, a method for obtaining characteristic points of the inner and outer eyes, the nose head, the left nose, the right nose, the left end of the mouth, and the right end of the mouth Will be described. here,
These feature points (right eye corner P5, right eye corner P6, left eye corner P6)
7, left eye corner P8, nose head P9, nose right end P10, nose left end P1
1, the right end P12 and the left end P13) are defined as shown in FIG. These feature points are obtained by template matching of the image around the feature points including the feature points. As a template matching method, for example, a normalized density correlation method can be cited. In the normalized density correlation method, a gray shade image is used. That is,
The change pattern of the gray density is compared between the target image and the template image, and the more similar the change pattern of the gray density, the greater the degree of matching match. Then, the point having the highest degree of matching in the search range is extracted as a feature point. At this time, as the template, image data obtained by averaging gray densities in the vicinity of each feature point of the face images of a plurality of persons is used. FIG. 1 shows an example of a specific matching result by template matching.
It is shown in FIG.

Next, the feature amount (Eye_size, E
ye_shape) and layout exaggeration features (Eye_height, Eye_s
The method for determining pace) will be described with reference to FIG. Although only the left eye will be described here, the right eye can be obtained in a similar manner. FIG.
In 3, the coordinate value (xp7, yp7) of the inner corner P7 and the outer corner P8
It is assumed that the coordinate values (xp8, yp8) have already been given. In the case of the right eye, it is assumed that the inner corner P5 and the outer corner P6 are given.

Further, the conventions for drawing in FIG. 23 are as follows (a) to (nu). (A) The vertical center line of the eye is parallel to the Y axis, and the horizontal center line of the eye is parallel to the X axis. (B) The X coordinate of the vertical center line of the eye = (X coordinate value of the inner corner of the eye + X coordinate value of the outer corner of the eye) / 2 (c) Y coordinate of horizontal center line of eye = (Y coordinate value of inner corner of eye + Y coordinate value of outer corner of eye) / 2 (d) EyeSearchArea_X = Y1 × 0.02 (e) EyeSearchArea_Y = Y1 × 0.1 (f ) Addition brightness at Y coordinate = sum of brightness at corresponding Y coordinate in EyeSearchArea_X (g) Difference of addition brightness at Y coordinate = addition brightness at Y coordinate− (Y-1) brightness at coordinate (h) D1 = Maximum value of the added brightness existing at a position below the horizontal center line of the eye and closest to the horizontal center line. Maximum value of added brightness (D) D3 = 0.7 × min (D1, D2) However, in the above formula, the decimal point Truncate below (coordinates are integers)

First, a reference horizontal line of the eye and a reference vertical line of the eye are obtained. Here, the reference horizontal line of the eye is (yp7 + yp8)
/ 2 is obtained as a straight line parallel to the X axis, and the reference vertical line of the eye is obtained as a straight line passing (xp7 + xp8) / 2 and parallel to the Y axis. Next, the search range in the X and Y directions of the eyes is determined according to the following equation (Equation 26).

[0095]

[Equation 26] EyeSearchArea_X = Y1 × 0.02 EyeSearchArea_Y = Y1 × 0.1 (Equation 26) Next, (EyeSearchArea_X × EyeSear) shown in FIG.
The luminance values are summed at each Y coordinate within the range of (chArea_Y). However, the luminance value is 255-density value. In this example, the density value of a black pixel is 255, and the density value of a white pixel is 0.
And Next, at each Y coordinate, ｛(xp7 + xp
8) / 2- (EyeSearchArea_X / 2)｝-｛(xp7 + xp8)
/ 2 + (EyeSearchArea_X / 2)} is defined as “additional brightness”, and the defined additional brightness is {(yp7 + yp8) / 2− (EyeSearchArea_Y / 2)} ~
It is determined for each Y coordinate within the range of {(yp7 + yp8) / 2 + (EyeSearchArea_Y / 2)}. Here, the reason for obtaining the added brightness without using the brightness of one pixel is to prevent the influence of noise.

Next, FIG. 23 shows (EyeSearchAre
a_X × EyeSearchArea_Y)
A difference between the above added brightnesses is obtained. The difference obtained at this time is, as shown in the following equation (Equation 27), the difference between adjacent Y
It is the difference in the added brightness of the coordinates.

[0097]

## EQU27 ## Difference = additional brightness of Y coordinate−additional brightness of (Y-1) coordinate (Expression 27) Next, values D1, D2, and D3 shown in FIG. 23 are obtained. Here, D1 is the value of the maximum point of the added brightness existing at a position below the reference horizontal line and closest to the reference horizontal line, D2
Is the value of the maximum point of the added brightness existing at a position above the reference horizontal line and closest to the reference horizontal line, and D3 is a value obtained by multiplying the smaller value of D1 and D2 by 0.7, that is, , D3 = 0.7 × min (D1, D2).

Next, a search is made upward from the horizontal center line of the eye to find a point where the added brightness is greater than D3, and a straight line passing through this point and parallel to the X axis is defined as EyeLineTop2.

Next, a search is made downward from the horizontal center line of the eye to find a point where the added brightness becomes larger than D3, and a straight line passing through this point and parallel to the X axis is defined as EyeLineBot2.

Next, a search is made in the vertical direction within a range of Y1 × 0.03 from the straight line EyeLineTop2 to find the maximum point of the difference in brightness. A straight line passing through this maximum point and parallel to the X axis is Ey
eLineTop1.

Next, a search is made in the vertical direction within a range of Y1 × 0.03 from the straight line EyeLineBot2 to find the maximum point of the difference in brightness. A straight line passing through this maximum point and parallel to the X axis is Ey
eLineBot1.

Next, two mutually parallel straight lines EyeLineTop1
And the distance between EyeLineBot1 and Left_EyeWidth_Y.

Next, Left_EyeWi is obtained by the following equation (Equation 28).
Find dth X.

[0104]

[Equation 28] Left_EyeWidth_X = X _p7 −X _p8 (Equation 28) Next, Right_EyeWidth_X and Right_EyeWidth_Y are similarly obtained for the right eye, and the left and right values are averaged according to the following equation (Equation 29).

[0105]

[Equation 29] EyeWidth_X = (Left_EyeWidth_X + Right_EyeWidth_X) / 2 EyeWidth_Y = (Left_EyeWidth_Y + Right_EyeWidth_Y) / 2 (Equation 29) Next, the feature amount for eye part exaggeration (Eye_size, Eye_shape)
Is calculated by the following equation (Equation 30).

[0106]

[Equation 30] Eye_size = (EyeWidth_X × EyeWidth_Y) / (X1 × Y1) Eye_shape = EyeWidth_Y / EyeWidth_X (Equation 30)
t_Eye_height, Y2, Eye_height, Eye_space)
It is determined by calculation according to 1).

[0107]

[Number 31] _{Right_Eye_height = (y p5 + y p6} ) / 2-y p2 Left_Eye_height = (y p7 + y p8) / 2-y p2 Y2 = (Right_Eye_height + Left_Eye_height) / 2 Eye_height = Y2 / Y1 Eye_space = (x p7 -x p5 ) / X1 (Equation 31) Next, a method for obtaining the nose part exaggeration feature amount Nose_shape and the arrangement exaggeration feature amount Eye_nose will be described. In this case, as shown in the following equation (Equation 32), Right_Eye_Right_Eye_Right is obtained using the Y coordinate values of the left and right inner corners P5 and P7 and the nose head P9.
Find nose and Left_Eye_nose.

Next, the calculated Right_Eye_nose, Left_
Y4 is obtained by averaging Eye_nose.

Next, using the Nose_shape and Y2, the left and right nose end points P10 and P11 and the previously obtained Y4 are used.
Ask for Eye_nose.

[0110]

Equation 32] _{Right_eye_nose = y p5 -y p9 Left_eye_nose =} y p7 -y p9 Y4 = (Right_eye_nose + Left_eye_nose) / 2 Nose_shape = (x p11 -x p10) / Y4 Eye_nose = Y4 / Y2 ( number 32) Next, mouth A method for obtaining the feature amount Mouth_size and Mouth_shape for component exaggeration and the feature amount Nose_mouth for arrangement exaggeration will be described with reference to FIG. First, EyeLineT
MouthLineTop in the same way as op1 and EyeLineBot1
1. Find MouthLineBot1.

Next, straight lines MouthLineTop1 parallel to each other
And the distance between MouthLineBot1 and MouthWidth_Y. Next, MouthWidth_X is obtained using the following equation (Equation 33).

[0112]

MouthWidth_X = X _p13 −X _p12 (Equation 33) Then, the mouth part exaggeration feature amounts Mouth_size and Mouth_shape are obtained using the following equation (Equation 34).

[0113]

Mouth_size = MouthWidth_X / X1 Mouth_shape = MouthWidth_Y / MouthWidth_X (Expression 34) Next, the mouth arrangement exaggeration feature amount Nose_mouth is obtained using the following expression (Expression 35). In Equation 35, ymc is Mou
This is the Y coordinate of a straight line that is parallel to thLineTop1 and MouthLineBot1 and that passes through the middle of these straight lines.

[0114]

Equation 35] _{Y5 = y p9 -y mc Nose_mouth =} Y5 / Y2 ( number 35) Next, a method for obtaining the components exaggerated for feature quantity Brow_thickness and arranged exaggerated feature quantity Eye_brow eyebrow reference to FIG. 25 . Here, only the case of the left eyebrow will be described, but the same can be obtained for the right eyebrow. First, a search is made on the reference vertical line of the eye upward from the upper line EyeLineTop1 of the eye to find a point at which the luminance becomes minimum. This point is the darkest part and is the center of the eyebrows.

Next, a straight line that passes through the minimum point and is parallel to the X axis is determined, and this is set as a reference horizontal line of the eyebrows. The reference vertical line of the eyebrows is a line obtained by extending the reference vertical line of the eye.

Next, the eye features EyeLineTop1, EyeLineB
BrowLineTop1, BrowLine
Ask for Bot1.

Next, a straight line BrowLineTop1 parallel to each other
Find the distance to BrowLineBot1 and call this Left_BrowWidth_
Let it be Y. Similarly for the right eyebrow, Right_Browwi
dth_Y is determined.

Next, the feature amount Brow_thickness for component exaggeration is calculated according to the following equation (Equation 36).

[0119]

Brow_thickness = (Left_BrowWidth_Y + Right_BrowWidth_Y) / 2 / Y1 (Formula 36) Next, the feature amount for layout exaggeration is expressed by the following formula (Formula 37).
Is required in accordance with In the following equation, ybL is the left eyebrow B
Y coordinate value of rowLineBot1, ybR is Y of BrowLineBot1 of right eyebrow
The coordinate value, yeL, is the Y coordinate of EyelineTop1 of the left eye, and yeR is the Y coordinate of EyelineTop1 of the right eye.

[0120]

Equation 37] _{Y3 = (y bL + y bR} ) / 2- (y eL + y eR) / 2 Eye_brow = Y3 / Y2 ( number 37) then determines a scaling factor Scale_ratio according the following equation (Equation 38). In the following equation, Scale_base is an integer value set in advance.

[0121]

[Equation 38] Scale_ratio = (X1 × Y1) / Scale_base (Equation 38) From the feature amount extraction means 103 described above, the feature amount storage means 104, the component exaggeration means 109, and the component arrangement means 1
FIG. 26 shows the structure of the data output to 10.

As is clear from the drawing, the Eye_heig is transmitted from the feature extracting means 103 to the feature storing means 104.
ht, Eye_space, Eye_nose, Nose_mouth, Eye_brow, Eye_siz
e, Eye_shape, Nose_shape, Mouth_size, Mouth_shape, Brow
_thickness is output.

Further, the feature amount extraction means 103 sends the Eye_size, Eye_shape, Nose_shap
e, Mouth_size, Mouth_shape, Brow_thickness are output.

Further, the feature amount extraction means 103 sends the Scale_ratio, Eye_height, Eye
_space, Eye_nose, Nose_mouth, Eye_brow, X1, Y1 are output.

Next, the configuration of the feature amount storage means 104 will be described. The input of this means is a person parameter Person_para, a component exaggeration feature amount, and an arrangement exaggeration feature amount.

The output is output to the component exaggeration nonlinear function creating means 105 and the arrangement exaggeration nonlinear function creating means 1.
06. Here, the output to the component exaggeration nonlinear function creating unit 105 is a person parameter and a component exaggeration feature amount. The output to the arrangement exaggeration nonlinear function creating means 106 is a person parameter and an arrangement exaggeration feature amount.

The feature amount storage means 104 stores a person parameter Person_para, a feature amount for component exaggeration, and a feature amount for layout exaggeration. FIG. 27 shows the structure of each data stored in the feature amount storage unit 104. As shown in FIG.
_para, Eye_height, Eye_space, Eye_nose, Nose_mouth, Eye
_brow, Eye_size, Eye_shape, Nose_shape, Mouth_size, Mou
th_shape and Brow_thickness are stored.

Next, the part exaggeration nonlinear function creating means 10
The configuration of FIG. 5 will be described with reference to the flowcharts of FIG. 28 (when the non-linear function is a fuzzy rule expression) and FIG. 29 (when the non-linear function is a mathematical expression).

This part exaggeration nonlinear function creating means 105
Is a data string of a person parameter and a feature amount for part exaggeration, and its output is a non-linear function and a feature value intermediate value (when the non-linear function is a mathematical expression). here,
The non-linear function is used by the part exaggerating means 109 to exaggerate the shape and size of the face part.
The input of this non-linear function is the feature amount for part exaggeration obtained from the real thing, and the output thereof is the part exaggeration parameter (enlargement / reduction ratio of the image) of the face part to be created. As the nonlinear function, for example, as shown in the curves in FIGS. 32 and 33, when the input component exaggeration feature amount x exceeds a certain value, the output component exaggeration parameter f (x) rapidly increases. Are used which have an S-shaped tendency to increase.

The non-linear functions used in this example include:
There are a case where a fuzzy inference rule is used and a case where a mathematical expression is used. FIG. 34 shows an example of a fuzzy inference rule for parts exaggeration. In this example, the value of the consequent part of the fuzzy inference rule is set to be properly used depending on the value of the person parameter, and in the same manner as in the feature amount database, for the adult male, for the adult female, and for the child. Are prepared. An example of the value of the consequent part is shown at the right end of FIG. As for fuzzy labels, SMALL, MEDIUM, BI
G types are prepared. Each fuzzy inference rule has one input and one output. The input value of the fuzzy inference is the feature amount for component exaggeration (Eye_size, Eye_shape, Nose_shape, Mouth_
size, Mouth_shape, Brow_thickness). In the mathematical expressions described later, X used as an input value refers to these feature amounts (Eye_size, Eye_shape, Nose_sh
ape, Mouth_size, Mouth_shape, Brow_thickness). The output value of the fuzzy inference is a component exaggeration parameter.

Next, the configuration of the component exaggeration non-linear function creating means 105 in the case where the non-linear function is a fuzzy rule expression will be described with reference to the flowchart of FIG. 28, especially the method of determining the membership function for each fuzzy label. It will be explained while doing.

First, for each of an adult male, an adult female, and a child (step 28)
02 to 2805), the data group of the feature amount for component exaggeration is read (step 2801), and then each of them is read (step 28).
031, 28035, and 28036), the minimum value MIN, the average value AVE, and the maximum value MAX are obtained (step 2803).
2).

Next, using the minimum value MIN, average value AVE, and maximum value MAX obtained above, numerical values LEFT and RIGHT are obtained by calculation in accordance with the following equation (Equation 39) (step 28033).

[0134]

LEFT = (MIN + AVE) / 2 RIGHT = (AVE + MAX) / 2 (Expression 39) Next, SM is calculated using the values of the minimum value MIN, average value AVE, maximum value MAX, LEFT, and RIGHT determined above.
The membership functions of ALL, MEDIUM, and BIG are obtained (step 28034). That is, SMAL
The membership function of L is defined as X for input value and μs for output value.
In the case of (X), it is obtained by the following equation (Equation 40).

[0135]

(Equation 40) The MEDIUM membership function is obtained by the following equation (Equation 41) when the input value is X and the output value is μm (X).

[0136]

[Equation 41] In addition, the BIG membership function uses input values X,
When the output value is μb (X), it can be obtained by the following equation (Equation 42).

[0137]

(Equation 42) Then, the non-linear function of the fuzzy rule expression obtained in this manner is sent to the component exaggeration means 109, and is subjected to component exaggeration processing described later (step 280).
6).

Next, the configuration of the component exaggeration non-linear function creating means 105 when the non-linear function is a mathematical expression will be described with reference to FIG.
This will be described with reference to the flowchart of FIG.

First, for each of an adult male, an adult female, and a child (step 29)
02 to 2905), a data group of the feature amount for component exaggeration is read (step 2901), and then each of them is read (step 29).
031, 29035, and 29036) are obtained (step 29032).

Next, using the minimum value MIN and the maximum value MAX obtained above, a feature amount intermediate value MID is obtained by calculation according to the following equation (Equation 43) (step 2903).
3).

[0141]

MID ← (MIN + MAX) / 2 (Equation 43) Next, using the values of the minimum feature value MIN, maximum feature value MAX, intermediate feature value MID, and constants R1 and R2 obtained above, In accordance with the equation (Equation 44), the coefficient a of the nonlinear function is obtained by calculation for the feature amount corresponding to the current values of I and J (step 29034). Here, I is a person parameter, and it is assumed that data is processed for adult male when I = 0, adult female when I = 1, and child when I = 2. J is the type of the feature amount.
If Eye = 0, Eye_size, If J = 1, Eye_shape, J
= 2 for Nose_shape, J = 3 for Mouth_size,
Mouth_shape when J = 4, Brow_thic when J = 5
It is assumed that the kness feature amount is processed.

[0142]

[Equation 44] When the intermediate value of the feature data group is input,
This means that a coefficient a such that the output of the nonlinear function f (x) becomes 1 is obtained. That is, when the intermediate value MID is input, the size and shape of the component image are not exaggerated.

Here, a method of realizing a non-linear function using a mathematical expression will be described theoretically. For example, the non-linear function having the S-shape shown in FIGS. 32 and 33 is generally represented by the following equation (Equation 45).

[0144]

[Equation 45] Here, the values of R1, R2, and a are different between the case of the feature amount for component exaggeration described here and the case of the feature amount for arrangement exaggeration described later. The initial values of R1 and R2 are default values, but the degree of exaggeration can be adjusted by changing these values. For example, when a default value is used, when a = 10, the maximum value of the nonlinear function f (x) is 1.2 and the minimum value is 0.8. Also,
As described above, the value of the coefficient a is obtained by calculation according to the equation (Equation 44).

The non-linear function of the mathematical expression obtained in this way is sent to the component exaggeration means 109, and is subjected to component exaggeration processing described later (step 2906).

Next, the arrangement exaggeration nonlinear function creating means 10
The configuration of FIG. 6 will be described with reference to the flowcharts of FIG. 30 (when the nonlinear function is a fuzzy rule expression) and FIG. 31 (when the nonlinear function is a mathematical expression).

The arrangement exaggeration nonlinear function creating means 106
Is a data string of a person parameter and a feature amount for layout exaggeration, and its output is a non-linear function and a feature value intermediate value (when the non-linear function is a mathematical expression). here,
The non-linear function is used by the component arranging means 110 to exaggerate the positional relationship between the face contour and the component, the positional relationship between the components, and the like. The input of this non-linear function is the layout exaggeration feature quantity obtained from the real thing, and its output is the layout exaggeration parameter between the parts of the face image to be created. As the nonlinear function, for example, as shown in the curves in FIGS. 32 and 33, when the input layout exaggeration feature quantity x exceeds a certain value, the output layout exaggeration parameter f
Those having an S-shaped tendency in which the value of (x) rapidly increases are used.

The non-linear functions used in this example include:
There are a case where a fuzzy inference rule is used and a case where a mathematical expression is used. An example of a fuzzy inference rule for layout exaggeration is shown in FIG. In this example, the value of the consequent part of the fuzzy inference rule is set to be properly used depending on the value of the person parameter, and in the same manner as in the feature amount database, for the adult male, for the adult female, and for the child. Are prepared. An example of the value of the consequent is shown at the right end of FIG. As for fuzzy labels, SMALL, MEDIUM, BI
G types are prepared. Each fuzzy inference rule has one input and one output. The input values of the fuzzy inference are the layout exaggeration feature amounts (Eye_height, Eye_brow, Eye_nose, Nose_mo
uth, Eye_space). In the mathematical expressions described later, X used as an input value refers to these feature amounts (Eye_height, Eye_brow, Eye_nose, Nose_mouth, Eye_mouth).
space). The output value of the fuzzy inference is an arrangement exaggeration parameter.

Next, the configuration of the arrangement exaggeration non-linear function creating means 106 when the non-linear function is a fuzzy rule expression will be described with reference to the flow chart of FIG. 30, especially the method of determining the membership function for each fuzzy label. It will be explained while doing.

First, for each of an adult man, an adult woman, and a child (step 30)
02 to 3005), a data group of the layout exaggeration feature amount is read (Step 3001), and then each of them is read (Step 30).
031, 30035, and 30036), the minimum value MIN, the average value AVE, and the maximum value MAX are obtained (step 3003).
2).

Next, using the minimum value MIN, the average value AVE, and the maximum value MAX obtained above, the respective numerical values LEFT, RIGHT are obtained according to the equation (Equation 39) cited earlier in the description of the means for creating a nonlinear function for component exaggeration. Is obtained by calculation (step 28033).

Next, using the values of the minimum value MIN, average value AVE, maximum value MAX, LEFT, and RIGHT determined above, each membership function of SMALL, MEDIUM, and BIG is determined (step 28034).

That is, when the input value is X and the output value is μs (X), the SMALL membership function is obtained by the equation (Equation 40) cited earlier in the description of the component exaggeration nonlinear function creating means. .

When the input value is X and the output value is μm (X), the membership function of MEDIUM can be obtained by the equation (Equation 41) cited earlier in the description of the component exaggeration nonlinear function creating means. .

Further, the membership function of BIG is:
Assuming that the input value is X and the output value is μb (X), the value is obtained by the equation (Equation 42) cited earlier in the description of the component exaggeration nonlinear function creating means.

Then, the non-linear function of the fuzzy rule expression thus obtained is sent to the component arranging means 110, where it is subjected to a component exaggeration process described later (step 3006).

Next, the configuration of the arrangement exaggeration nonlinear function creating means 106 when the nonlinear function is a mathematical expression will be described with reference to FIG.
1 will be described with appropriate reference to the flowchart.

First, an adult man, an adult woman, and a child are stored in the feature amount storage means 104 (step 31).
02 to 3105), the data group of the layout exaggeration feature amount is read (step 3101), and then each of them is read (step 31).
031, 31035, 31036) is obtained (step 31032).

Then, using the minimum value MIN and the maximum value MAX obtained above, a feature amount intermediate value MID is obtained by calculation in accordance with the equation (Equation 43) cited earlier in the description of the component exaggeration nonlinear function creating means. (Step 3103
3).

Next, the minimum feature value MIN obtained above,
Feature amount maximum value MAX, feature amount intermediate value MID, constant R1,
Using the value of R2, according to the equation (Equation 44) cited earlier in the description of the component exaggeration nonlinear function creating means, the current I
The coefficient a of the non-linear function is obtained by calculation for the feature amount corresponding to the values of and J (step 31034). Here, I is a person parameter, and it is assumed that data is processed for adult male when I = 0, adult female when I = 1, and child when I = 2. J is the type of the feature amount. Eye_height when J = 0, Eye_space when J = 1, Eye_nose when J = 2, Nose_mouth when J = 3, Eye_brow when J = 4. It is assumed that the feature amount is processed.

The above equation (Equation 44) indicates that when the intermediate value of the feature amount data group is input, the output of the nonlinear function f (x) is 1
This means that a coefficient a such that That is, when the intermediate value MID is input, the component arrangement is not exaggerated.

Here, the theoretical description of the method of realizing the nonlinear function using the mathematical formula is omitted because it is the same as the case of the component exaggeration feature amount described above.

Then, the non-linear function of the mathematical expression obtained in this way is sent to the component exaggeration means 109, and is subjected to a component exaggeration process described later (step 3106).

Next, the configuration of the contour image extracting means 107 will be described. The input of the contour image extracting means 107 is a color face image, and the output is a binary face contour image. First, the contour image extracting unit 107 extracts a skin color area of a color face image. Next, the contour image extracting means 107
Converts a color face image into a binary image. Next, the contour image extracting unit 107 sets the density value inside the area corresponding to the skin color area to 0 in the binarized face image.

As described in the first embodiment, several examples of the face outline extracted by the outline image extracting means 6 are shown in FIG. As can be seen from the drawing, the contour image extracting means 107 can extract the contour of the face as a figure surrounded by the hairstyle and the contour line of the chin for each of the target persons.

The contour image thus extracted by the contour image extracting means 107 is output to the component arranging means 110.

Next, the configuration of the component image storage means 108 will be described. In the part image storage unit 108, images of the face parts of the right eyebrow, the left eyebrow, the right eye, the left eye, the nose, and the mouth are stored as an image database for each value of the person parameter (I = 0, 1, 2). These component images are appropriately selected according to the value of the person parameter I input from the data input unit 101, and output to the component exaggeration unit 109. Here, the component image is a binary image.

Next, the structure of the component exaggeration means 109 is shown in FIG.
This will be described with reference to the flowchart of FIG. The input of the component exaggeration means 109 includes the feature quantity for component exaggeration coming from the feature quantity extraction means 103 and the non-linear function creation means 10 for component exaggeration.
5, a non-linear function and a feature amount intermediate value, a component image from the component image storage unit 108, a person parameter, an exaggerated portion selection parameter, and an exaggeration degree parameter from the data input unit 101. The output of the component exaggeration means 109 is a component image whose size or shape has been changed.

That is, the part exaggeration means 109 reads the person parameter, the exaggerated part selection parameter, and the exaggeration degree parameter from the data input means 101 (step 3601), and reads the part image from the part image storage means 108 (step 3602). Finally, the feature amount for component exaggeration is read from the feature amount extraction means 103 (step 3603).

Thereafter, a component exaggeration parameter is obtained by using a non-linear function based on the read data (step 3604). At this time,
The value of the consequent part of the part exaggeration rule shown in FIG. 34 is properly used depending on the value of the person parameter.

A fuzzy inference method will be described below. When the parts exaggeration rules shown in FIG. 34 are generalized, each arrangement location is expressed by three rules expressed by the following equation (Equation 46).

[0172]

IF X is SMALL THEN FZ is W1 IF X is MEDIUM THEN FZ is W2 IF X is BIG THEN FZ is W3 (Expression 46) Here, the output values of the SMALL, MEDIUM, and BIG membership functions are respectively μs (X), μm (X),
When μb (X) is used, the component exaggeration parameter FZ which is an output value of fuzzy inference is expressed by W expressed by the following equation (Equation 47).
1, W2, and W3.

[0173]

[Equation 47] However, FZ, which is a part exaggeration parameter, is FA for eye size, FB for eye shape, FC for nose shape, FD for mouth size, and FD for mouth shape. F
In the case of E and the thickness of the eyebrows, the value is FF, and the size and shape of the part are changed using these values.

When the exaggerated portion selection parameter and the exaggeration degree parameter are input to the data input means 101, the form of the nonlinear function is changed in accordance with the values (step 3604). That is, the non-linear function of the feature specified by the exaggerated portion selection parameter is modified.

The exaggeration degree parameter and the deformation method are different between the case where the nonlinear function is expressed by fuzzy rules and the case where it is expressed by mathematical expressions.

When the non-linear function is represented by fuzzy rules, the processing proceeds as follows. When the value of the exaggeration degree parameter input by the user is Exaggerate_para, the values of W1 and W3 in the component placement rule are changed. The values of W1 and W3 before the change are W1_old and W, respectively.
If 3_old and the changed values are W1_new and W3_new,
The relationship between the two is represented by the following equation (Equation 48).

[0177]

W1_new = W1_old × (1−Exaggerate_para) W3_new = W3_old × (1 + Exaggerate_para) where 0 ≦ Exaggerate_para <1 (Equation 48) It should be noted here that in the previous equation (Equation 48), Exagge
The sign of rate_para is different between the two formulas. Therefore, the greater the value of Exaggerate_para, the greater the degree of exaggeration.

When the non-linear function is represented by an equation, the processing proceeds as follows. In this case, as described above, the nonlinear function is expressed by the following equation (Equation 49).

[0179]

[Equation 49] In the above equation, the values of the exaggeration degree parameter input by the user are R1 and R2. Depending on these values, the function is directly transformed to change the degree of exaggeration. However, it is necessary to change the value of the coefficient a using the feature amount intermediate value MID. The change of the value of the coefficient a is performed according to the following equation (Equation 50).

[0180]

[Equation 50] Next, a method of changing the size and shape of the component using the output value of the nonlinear function will be described below. Assuming that the point (x, y) on the component image is scaled to change its position to the point (X, Y), the change in size and shape is expressed by the following equation (Equation 51).

[0181]

[Equation 51]-For eye size Y = FA x y, X = FA x x-For eye shape Y = FB x y, X = x-For nose shape Y = y, X = FC × x In the case of mouth size Y = FD × y, X = FD × x (Equation 51) Here, if the scaling ratios in the x direction and the y direction are equal, the component images should be scaled similarly. And the x direction and y
If the enlargement / reduction ratio in the direction is different, the component image is deformed. It should be noted that a person skilled in the art can easily understand a specific method of converting the size and shape of an image by referring to the description in the reference document “Practical image processing learned in C language” (Ohm). Should be.

Next, the structure of the component arranging means 110 is shown in FIG.
This will be described with reference to the flowchart of FIG. The input of the component arranging means 10 includes an arrangement exaggeration feature quantity and an enlargement / reduction rate coming from the feature quantity extraction means 103, a nonlinear function and a feature quantity intermediate value coming from the arrangement exaggeration nonlinear function creating means 106, and a component exaggeration means. A component image arriving from 109, a contour image arriving from the contour image extracting means 107,
Person parameters input from the data input means 101,
An exaggerated portion selection parameter and an exaggeration degree parameter. The output of the component arranging means 110 is a face image obtained by synthesizing the component image and the contour image.

That is, the component arranging means 110 reads the person parameter, the exaggerated portion selection parameter, and the exaggeration degree parameter from the data input means 1 (step 3701), and reads the contour image from the contour image extracting means 107 (step 3702). The component image is read from the component exaggeration means 109 (step 3703). Finally, the feature amount for arrangement exaggeration and the enlargement / reduction ratio are read from the feature amount extraction means 103 (step 3704). Thereafter, based on the read data, a placement exaggeration parameter is obtained using a nonlinear function (step 370).
5). At this time, the value of the consequent part of the layout exaggeration rule shown in FIG. 35 is properly used depending on the value of the person parameter.

The fuzzy inference method will be described below. When the layout exaggeration rules shown in FIG. 35 are generalized, each layout location is expressed by three rules expressed by the following formula (Equation 52).

[0185]

IF X is SMALL THEN FZ is W1 IF X is MEDIUM THEN FZ is W2 IF X is BIG THEN FZ is W3 (Expression 52) Here, the output values of the membership functions SMALL, MEDIUM, and BIG are respectively μs (X), μm (X),
When μb (X) is set, the arrangement exaggeration parameter FZ which is the output value of the fuzzy inference is expressed by W
1, W2, and W3.

[0186]

(Equation 53) However, the layout exaggeration parameter FZ is FG for eye height, FH for eyebrow and eye spacing, FI for eye to nose spacing, FJ for nose and mouth spacing, FJ for left and right. In the case of the eye interval, FK is used, and real values (YY2, YY3, YY4, XX2) for component placement are obtained using these values.

When the exaggerated portion selection parameter and the exaggeration degree parameter are input to the data input means 101, the nonlinear function is changed in accordance with the values in the same manner as in the case of the component exaggeration means 109 described above. The shape is changed (step 3705). That is, when the value of the exaggeration degree parameter input by the user is Exaggerate_para, the values of W1 and W3 in the component placement rule are changed. The values of W1 and W3 before the change are W1_old and W3_ol, respectively.
Assuming that d and the changed values are W1_new and W3_new, the relationship between the two is represented by the equation (Equation 48) cited earlier.

It should be noted here that the sign of Exaggerate_para is different between the two equations in the equation (Equation 48). Therefore, the greater the value of Exaggerate_para, the greater the degree of exaggeration.

Next, in order to adjust the size of each component image to the size of the face outline image, the value Sc of the scaling parameter is set.
Each part image is enlarged or reduced using ale_ratio (step 3707). A point (x, y) on the part image
Is scaled to change the position to the point (X, Y), a relationship represented by the following equation (Equation 54) is established between the two.
It should be noted that a person skilled in the art can easily understand a specific conversion method by referring to the description in the reference document “Practical image processing learned in C language” (Ohmsha).

[0190]

Y = Scale_ratio × y, X = Scale_ratio × x (Equation 54) Then, based on the layout exaggeration parameter and X1, Y1,
Real values (YY2, YY2) for component placement shown in FIG.
3, YY4, XX2) is obtained by the following equation (Equation 55) (step 3706). However, FG, FH, FI, F
J and FK are inter-component distance parameters which are output values of fuzzy inference.

[0191]

[Equation 55] Eye height YY2 = FG × Y1 Eyebrow and eye distance YY3 = FH × YY2 Eye and nose distance YY4 = FI × YY2 Nose and mouth distance YY5 = FJ × YY2 Left and right eye distance XX2 = FK × X1 (Equation 55) Next, based on the obtained real values for component placement, each component image is placed as shown in FIG. 38 (step 3708). However, the components are arranged such that the representative point of each component shown in FIG. 39 is at the position shown by the following equation (Equation 56). That is, the coordinates at which the representative point of each part is arranged are expressed by the following equation (Equation 5).
6). The image created here is 2
It is a value image.

[0192]

Equation 56] _{x center = (x p3 + x} p4) / 2 right _{eyebrow: (x center -XX2 / 2,} y p2 + YY2 + Eye_half_Y + YY3) left _{eyebrow: (x center + XX2 / 2} , y p2 + YY2 + Eye_half_Y + YY3) eye: (x _center - _{XX2 / 2, y p2 + YY2} ) left _{eye: (x center + XX2 / 2} , y p2 + YY2) _{nose: (x center, y p2 +} YY2-YY4) _{mouth: (x center, y p2 +} YY2-YY4-YY5) ( number 56 Next, the configuration of the face image output unit 111 will be described.
The face image output unit 111 visually outputs the face image synthesized as described above, and specifically, a printing device that outputs a hard copy of a binary image, which is a face image, or a display device that displays an image. And the like.

According to the second embodiment,
(1) It is possible to automatically create an exaggerated and impressive face image, which is an essential feature of a portrait, by exaggerating not only the positional relationship of the parts but also the shape and size of the parts. (3) that it is possible to prevent excessive exaggeration such that the face image does not resemble the face of the target person, and (4) the degree of exaggeration according to the user's preference. Can be adjusted, and the like.

Next, FIG. 40 shows the overall configuration of a portrait creating apparatus having a disguise simulation function according to another preferred embodiment (third embodiment) of the present invention, and the operation procedure thereof will be described. The flowchart of FIG. 41
42 and FIG.

As shown in FIG. 40, the portrait creating apparatus includes a data input unit 201 and a portrait creating unit 20.
2, a component image storage unit 203, a selection menu storage unit 204, an image synthesis unit 205, and a display unit 206
And printing means 207. These means 201 to 207 are specifically configured as follows.

First, the data input means 201 stores the attribute number and the part number input by the user in the part image storage means 2.
03 or an instruction input from the user such as deformation or position adjustment of parts to the image synthesizing means 205, and is configured by an input device such as a keyboard and a mouse. ing.

When the user selects “P. print” from a selection menu described later using the data input unit 201, the image synthesized by the image synthesizing unit 205 is sent to the printing unit 207. Is output. This selection operation is performed, for example, by inputting the letter “P” from a keyboard,
Alternatively, it is performed by moving the mouse cursor to the “P. print” portion on the screen and clicking the mouse button.

When the user uses the data input unit 201 to select “Q. end” from a selection menu described later, the image synthesized by the image synthesizing unit 205 is sent to the printing unit 207. Is output. This selection operation is performed, for example, by inputting a letter “Q” from a keyboard,
Alternatively, the user can move the mouse cursor to the “Q. end” portion on the screen and click the mouse button.

Next, the basic configuration of the portrait creating means 202 is the same as the first configuration described above with reference to FIG. 1 or FIG.
Alternatively, it is almost the same as that in the second embodiment, and creates a portrait based on the feature amount extracted from the face photograph. However, the output of the portrait creating means 202 is not a synthesized portrait image, but information on the preceding stage, that is, selected and exaggerated portrait parts (face parts), part image coordinate data, and positional relationship data. This is slightly different from that of the first or second embodiment.

Here, the portrait parts (face parts) include:
In this example, left and right eyebrows, left and right eyes, nose, and mouth are used. Further, as the component image coordinate data, the coordinates of the representative points of the components indicated by the crosses in FIG. 15 in the first embodiment and FIG. 39 in the second embodiment are used. In the case of the first embodiment, the positional relationship data is shown in FIG.
Y2 (eye height), Y3 (eyebrow and eye distance), Y4 (eye and nose distance), Y5 (nose and mouth distance), and X2 (left and right eye distance) are used. In the case of the second embodiment, YY2 (eye height), YY3 shown in FIG.
(Interval between eyebrows and eyes), YY4 (interval between eyes and nose), YY5
(Interval between nose and mouth) and XX2 (Interval between left and right eyes)
Is used.

Next, in the part image storage means 203, background parts for synthesizing, for example, a disguise simulation by synthesizing with a face part corresponding to a portrait, that is, a face outline part, a hairstyle part, a body part, a glasses part, a beard, etc. A plurality of patterns of parts and the like are stored. FIG. 43 shows an example of these background component images. In addition, the size of each part is unified, and the face parts (left and right eyebrows, left and right eyes, nose, and
Mouth) and can be combined without discomfort.

Each part is assigned a category symbol and a part number. The category symbol is a symbol for selecting the category of the background component that the user wants to combine, for example,
The face outline part is determined as “A”, the hairstyle part as “B”, and the body part as “C”. The part number is a number (positive integer value) assigned to each part, and is used when the user selects one part from a plurality of patterns of the same kind.

When a category symbol is input by the user, a list of all the component images included in the type indicated by the category symbol is read out from the component image storage means 203 together with the associated component number and displayed. Sent to the means 206. As a result, the user can select and instruct a part image to be combined from within the category using the part number. When a user inputs a part number, the part image indicated by the part number is read from the part image storage unit 203 and sent to the image combining unit 205.

An image synthesizing means 205 is added to each background part image.
The reference points are determined so that they can be aligned with each other and synthesized. FIG. 43 shows an example of reference points assigned to each background component. That is, as shown in FIG.
Four reference points consisting of 1, P2, P3, and P4 are shown (a in the figure), one reference point consisting of a point Q at the upper center of the hairstyle part is shown (b in the figure), and the body part has the same reference point. One reference point consisting of a point Q at the center of the neck (c in the figure), one reference point consisting of the center point Q of the eyeglass part (d in the figure), and the upper part of the beard part One reference point consisting of the central point Q has been determined (FIG. 9E). Also, corresponding to these reference points, the following coordinate value data is attached to each background component. That is, the coordinate values of the four reference points P1 to P4 are set for the face outline part, the relative coordinate values of the reference point Q with respect to the reference point P1 of the face outline part are set for the hairstyle part, and the reference values of the face outline part are set for the body part. The relative coordinate value of the reference point Q with respect to the point P2, the relative coordinate value of the reference point Q with respect to the left-eye reference point E shown in FIG. Relative coordinate values of the reference point Q with respect to the shown beard part reference point M are additionally stored.

Next, the selection menu storage means 204 stores
The category name of the background component that can be selected by the user is stored as a character string. Here, as the background parts, for example, “face contour”, “hairstyle”, “body”, “glasses”,
"Whiskers" and the like.

From this selection menu storage means 204,
Each of the above-mentioned category names is read with a part symbol added to the head thereof, and sent to the display means 206, for example, "A. face outline", "B. hairstyle". Thus, the user can select and instruct the category of the background part desired to be used in the disguise using the category symbol.

In the selection menu, a guide item “P. print” for receiving a print instruction from the user and a guide item “Q. end” for receiving a program end instruction from the user. Is provided.

Next, the image synthesizing means 205 generates a portrait image subjected to background synthesis by synthesizing the portrait image obtained from the portrait creating means 202 and the background part image obtained from the part image storing means 203. Then, this is sent to the display means 206 or the printing means 207.

In the background synthesizing process, a portrait image is generated by arranging a portrait part (face part) obtained from the portrait creation means 202 in the selected face outline image (see FIG. 43A). And a second process of arranging background parts (see FIGS. 43 (b), (c), (d), and (e)) on the generated portrait image to generate a background composite (disguise) portrait image And

Since the first processing for generating a portrait image is substantially the same as the component arrangement processing in the first and second embodiments described above, the first processing is performed by referring to the description of the corresponding part. It will be easily understood by those skilled in the art. However, as shown in FIG. 43 (a), each of the face contour parts has four reference points (P
1, P2, P3, P4)
By using these coordinate values, synthesis of the face outline part and the portrait part (face part) is performed. In addition, since the face outline part is used as a base at the time of synthesis, it is necessary to first set the face outline part selection mode before selecting another category part.

The second process for generating a portrait image subjected to background synthesis (disguise) is performed as follows. That is, as described above, FIG.
Each of the hairstyle, torso, glasses, and beard parts images shown in FIG. 43 (e) is provided with reference points Q, and these reference points are reference points of the face contour parts (P1 in FIG. 43 (a)). , P2
P3, P4) or the reference point of the portrait part (FIG. 44)
Relative coordinate values with respect to E in (a) and M) in FIG. 44 (b) are given. Then, by using these relative coordinate values, the disguise parts are arranged.

That is, for the hairstyle part shown in FIG. 43B, the reference point Q of the hairstyle part and the reference point P of the face contour
Parts arrangement is performed so that 1 matches.

For the body part shown in FIG. 43 (c), the relative coordinate value (dx, dy) of the reference point Q of the body part to the reference point P2 (Xp2, Yp2) of the face contour part is attached to the body. The stored components are arranged such that the coordinate value (Xq, Yq) of the synthesized reference point Q is expressed by the following equation (Equation 57).

[0214]

Xq = Xp2 + dx, Yq = Yp2 + dy (Equation 57) For the eyeglass part shown in FIG. 43D, the relative coordinates of the reference point Q of the eyeglass part with respect to the reference point E (Xe, Ye) of the left inner corner of the eye. The values (dx, dy) are stored in association with the eyeglass parts, and the coordinate values (Xq, Y) of the combined Q
The components are arranged so that q) becomes the following equation (Equation 58).

[0215]

Xq = Xe + dx, Yq = Ye + dy (Equation 58) For the beard part shown in FIG. 43 (e), the relative coordinate value of the reference point Q of the beard part with respect to the reference point M (Xm, Ym) of the mouth ( dx, dy) are stored along with the beard parts,
The components are arranged such that the coordinates (Xq, Yq) of the combined Q become the following equation (Equation 59).

[0216]

Xq = Xm + dx, Yq = Ym + dy (Equation 59) By such processing, each component is arranged at a position based on the reference point. Part size change and position change can be performed. This feature is available for graphical user
It can be easily realized by using the function of the interface.

Once a part has been synthesized, it may be desirable to replace that part with a different part of the same category.
In this case, it is preferable to provide the following functions. That is, a selected flag is provided for each category.
The initial values of all the selected flags are “0” (off). Once a category is selected, the value of the selected flag of that category is set to "1" (ON). When the same category is selected again while the selected flag is on, the parts of the corresponding category that have already been synthesized are deleted,
Synthesize a new part.

In order to be able to delete a component already synthesized, a layer structure used in a drawing tool such as a personal computer may be used. For example, by arranging parts for each layer, such as a face outline part on Layer 1, a portrait part on Layer 2, a hairstyle part on Layer 3,..., It is possible to delete the synthesized part.

Next, the display means 206 displays a selection menu,
It displays a component image list, a composite image, and the like, and can be configured by a display device such as a liquid crystal display, a CRT, and a plasma display.

Finally, the image printing means 207 is for generating a hard copy of a portrait image subjected to background synthesis (disguise) processing and the like, and is implemented by various color or monochrome printer devices such as a laser beam printer. Can be configured.

Next, the flow of a series of operations of the portrait creating apparatus having the disguise simulation function described above will be systematically described with reference to the flowcharts of FIGS. 41 and 42. When the user activates the apparatus, a face image of the user is read by an electronic camera or the like, and then the first
In addition, through the process described in detail in the second embodiment, a facial feature amount extraction process, a portrait part (face part) selection process, a portrait part exaggeration process, and a positional relationship data creation process are executed. (Step 4101).

Next, on the screen of the display device,
A list of face part images is displayed, and the apparatus enters a state of waiting for a user to select a face contour (step 4102). In this state, when the part number of the face outline part is received from the user (step 4103), the processing of synthesizing the face outline part of the designated number and the portrait part (face part) is performed, and the portrait is completed. (Step 4104).

Thereafter, the apparatus is in a state of waiting for an input operation by the user (step 4105), and according to the input operation (step 4106), an image synthesizing process (step 4107), an image printing process (step 4108), or an end process. Is executed.

As shown in FIG. 42, in the image synthesizing process (step 4107), first, the content of the selected flag of the category is determined (step 4201). If the flag is off (step 4201 off), After the selected flag of the corresponding category is turned on (step 4202), a list of background part images corresponding to the category symbols is displayed on the screen of the display device (step 4204), and the selection operation of the user is performed. It will be in a state of waiting. On the other hand, if the selected flag of the corresponding category is on (step 4201 on),
After the background component of the corresponding category that has already been combined is deleted from the combined image (step 4203), a list of component images corresponding to the category symbols is displayed on the screen of the display device (step 4204), and It is in a state of waiting for a selection operation.

In this state, when a part number is received from the user (step 4203), a process of synthesizing a portrait with a background part corresponding to the part number is performed (step 4204). Displayed on the screen of the device. Thereafter, when the user performs an operation of changing the component size or position with the mouse, the background synthesis (disguise) portrait is corrected accordingly, and a desired disguise portrait is completed (step 4205).

Thereafter, when the user gives a print instruction (step 4106), the completed disguise portrait is printed out in color or monochrome from the printer device (step 4108), whereby a hard copy of the disguise portrait is obtained. Become.

As described above, according to the portrait creating method and apparatus with the disguise simulation function according to the third embodiment, while using the portrait created automatically and exaggerated from the actual face image, By performing a disguise simulation by appropriately attaching background parts to this, a more interesting portrait can be created.

Next, FIG. 45 shows an overall configuration of a portrait creating apparatus having a simple creating function according to another preferred embodiment (fourth embodiment) of the present invention, and the operation procedure thereof will be described. Are shown in FIGS. 46, 47 and 48.

As shown in FIG. 45, the portrait creating apparatus includes a portrait creating section 301 and a background image synthesizing section 3.
02, display means 303, and component code connection means 304
Encoding means 305, image barcode combining means 306, printing means 307, decoding means 309,
And a bar code reading means 308.

The basic configuration of the portrait creating means 301 is as follows.
This is almost the same as that in the first or second embodiment described above with reference to FIG. 1 or FIG. 20, and creates a portrait based on the feature amount extracted from the face photograph. However, the output of the portrait creating means 302 is not only a portrait image obtained by exaggerating the face image according to a certain rule, but also information on the preceding stage, that is, a portrait part code, a placement code, and The output of the component coordinate data is slightly different from that of the first or second embodiment.

Here, the portrait part code in this example is a code relating to face parts such as eyes, nose, mouth, and eyebrows. FIG. 51 shows an example of a portrait part code when the portrait part is an eye. As shown in the figure, the size of the eyes is determined to be small (1), normal (2), and large (3). Regarding the shape of the eyes, it is determined to be thin (1), normal (2), and round (3). Further, regarding the inclination of the eyes, a falling eye (1), a horizontal eye (2), and an rising eye (3) are determined. Therefore, for example, assuming that the features of the eyes are small, ordinary shapes, and falling eyes, the caricature part code relating to the eyes is (121). The facial part codes relating to the nose, mouth, and eyebrows are similarly generated by converting the respective characteristic items into numerical values, for example.

The arrangement code is obtained by encoding the positional relationship among eyes, nose, mouth, and eyebrows. For example, assuming that the distance between the eye and the eyebrow is 325 pixels, the code related to the positional relationship between the eye and the eyebrow is (325). Similarly,
The positional relationship between the face outline and the eyes, the right eye and the left eye, the eyes and the nose, and the nose and the mouth are coded, respectively, and by connecting these individual mutual positional relationship codes, the target arrangement code is finally generated .

Further, as the part coordinate data, for example, the coordinates of the representative point of each part indicated by a cross in FIG. 15 in the first embodiment and FIG. 39 in the second embodiment are used. .

Next, the background image synthesizing means 302 generates a portrait image with the background synthesized by synthesizing the background image with the portrait obtained from the portrait creating means 301.
Although illustration is omitted to avoid complication,
First, the background image combining means 302 shown in FIG.
The data input unit 201, the component image storage unit 203 shown in FIG.
It should be understood that the selection menu storage means 204 is included. Therefore, as described above in the third embodiment, the background image synthesizing process in the background image synthesizing unit 302 includes the portrait information and the part coordinate data obtained from the portrait creation unit 301 and the user This is performed by synthesizing with the background part image selected by the interactive processing while looking at the menu. The portrait image combined with the background is sent to the display unit 303.

Next, the display means 303 is a liquid crystal type, CRT
It is composed of a color or monochrome display device such as a display type, a plasma display type, and the like, and the display means 303 appropriately displays a portrait image combined with a background and various operation menu images.

Next, the part code linking means 304 is a portrait part (face part) obtained from the portrait creation means 301.
Code and arrangement code, and background image synthesizing means 302
By generating a code string having a predetermined format by linking the background part code obtained from the above.
The code string thus obtained is sent to the encoding means 305.

Next, the encoding means 305 converts the code string sent from the component code linking means 304 into a corresponding bar code. The barcode information thus obtained is sent to the image barcode combining means 306.

Next, the image barcode combining means 306 combines the background combined image and the barcode based on the background combined image obtained from the background image combining means 302 and the barcode information obtained from the encoding means 305. Create image data. The image data thus created is sent to the printing unit 307.

Next, the printer constituting the printing means 307 executes a printing process based on the image data obtained from the image barcode synthesizing means 306, so that an example of the printing result is shown in the lower right of FIG. As described above, an image obtained by synthesizing the background composite image and the barcode 310 is printed out on a predetermined or arbitrary sheet in color or monochrome.

Next, the bar code reading means 308
It has a function of reading the barcode 310 attached to the background composite image from the paper on which the background composite image has been printed out, and can be realized by a known barcode reader. The bar code thus read is sent to the decoding means 309.

Next, the decoding means 309 decodes and converts the bar code sent from the bar code reading means 308 to generate and output a portrait part code (face part) code, an arrangement code, and a background part code. . The portrait part code and the arrangement code thus obtained are sent to the portrait creation means 301, and the background part code is sent to the background image synthesis means 302.

Thereafter, the portrait creating means 301 uses a portrait simple method based on the portrait part code and the arrangement code obtained from the decoding means 309 without depending on the face image of the user such as a face photograph. Execute the creation process. The details of the portrait creation process using this simplified method will be described later in detail with reference to FIG. Further, the background image synthesizing unit 302 executes a background image synthesizing process based on the background part code obtained from the decoding unit 309.

Next, the flowcharts of FIGS.
The operation of the portrait creating apparatus with a simple creation function according to the fourth embodiment will be systematically described with reference to the processing step explanatory diagrams of FIGS. 49 and 50.

As shown in the flow chart of FIG. 46, by performing a predetermined mode designating operation (step 4601), a normal mode for creating a portrait from a photograph (step 4602) and a simple mode for creating a portrait from a barcode (step 4601). Step 4603) can be selectively executed.

The details of the processing in the normal mode for creating a portrait from a photograph are shown in the flowchart of FIG. 47 and the processing step explanatory diagram of FIG. As shown in the flowchart of FIG. 47, when the processing is started, first, the portrait creating unit 301 creates a portrait based on the feature amount extracted from the face photograph (step 4701). That is, as shown in the process step explanatory diagram of FIG. 49, a feature extraction process (4
901) is executed to extract various feature amounts related to the face, and a component selection process (490) is performed according to these feature amounts.
2) is executed to obtain the eye / nose / mouth part codes. Further, the arrangement exaggeration processing (4
903) is executed to determine the eye / nose / mouth arrangement code.

Next, the background image synthesizing means 302 creates a background synthesized image obtained by synthesizing the portrait and the background component (step 4702). That is, the background part synthesizing process (490) is performed based on the portrait image and the part arrangement coordinate data obtained from the portrait creating means 301 and the background part image (body, glasses, etc.) separately specified by the user.
4) is executed to generate a background composite image, and this background composite image is displayed on the screen of the display device constituting the display unit 303.

Next, in the part code linking means 304, a portrait part (face part) code, an arrangement code,
In addition, processing for linking each of the background codes is executed (step 4703). That is, for the eye / nose / mouth part code, the part code connection processing (4905)
However, the arrangement code connection processing (4906) is executed for the eye / nose / mouth arrangement code, and the background code connection processing (4907) is executed for the background part code such as the body and glasses. A code connection process is performed for each of the backgrounds. Furthermore, their part codes
Consolidation processing for arrangement code / background code (4908)
Is executed, and finally a portrait code is generated.

Next, the encoding means 305 performs a process for converting the portrait code composed of the linked code sequence into a portrait bar code (step 4704). That is, the conversion process from the portrait code to the portrait barcode (4909) is executed, and the portrait barcode is output.

Next, the image barcode synthesizing means 306 synthesizes image data obtained by synthesizing the background synthesized image and the barcode, and sends this image data to the printing means 307 (step 4705).

Finally, the printing means 307 prints out the image in which the background synthesized image and the barcode are synthesized by executing a printing process based on the image data obtained from the image barcode synthesizing means 306. (Step 4706).

The details of the processing in the simple mode for creating a portrait from a bar code are shown in the flowchart of FIG. 48 and the processing step explanatory diagram of FIG. As shown in the flowchart of FIG. 48, when the process is started,
First, the barcode reading unit 308 reads a barcode from the hard copy 311 of the portrait image printed out earlier (step 4801).

Next, in the decoding means 309,
The read barcode is converted into a portrait part code, an arrangement code, and a background part code (step 4802). That is, as shown in FIG. 50, the bar code read by the bar code reading means 308 is subjected to a code decomposition process (50).
01) is executed to generate eye / nose / mouth part codes, eye / nose / mouth arrangement codes, and body / glasses background codes.

Next, the portrait creating means 301 selects a part corresponding to the portrait part code and creates a portrait image based on the arrangement code (step 480).
3). That is, as shown in the processing step explanatory diagram of FIG. 50, the part selection processing (5
002), the arrangement exaggeration process (5003) is executed by the arrangement codes of the eyes, nose, and mouth, and a portrait image is created.

Next, in the background image synthesizing means 302, as shown in FIG.
Background component synthesis processing (500
4) is executed to create a portrait image with background synthesis (steps 4804, 4805, 4806, 480).
7). At this time, the user is allowed to change the background image. That is, if the user does not select to change the background image (step 4804 is not changed), the background component corresponding to the background component code is selected (step 4805), and the background synthesis processing is executed. If the user selects to change the background image (step 4804) (step 4807), the user waits until the user selects another background (step 4806).
A background synthesis process is performed using the newly selected background component (step 4807).

Next, the image barcode synthesizing means 306 creates image data obtained by synthesizing the background synthesized image and the barcode (step 4808). That is,
Prior to that, in the part code coupling means 304,
Component selection processing (5002), placement exaggeration processing (500
3) and a code linking process (50) for generating a portrait code by linking the component code, the arrangement code, and the background code used in the background component synthesizing process (5004).
05), and the encoding means 305 executes a code conversion process (5006) from the portrait code to the portrait barcode. Then, the portrait bar code is combined with the portrait combined with the background.

As described above, according to the portrait creating method and apparatus with the simple creating function according to the fourth embodiment,
While using a portrait created by automatically exaggerating the actual face image, by appropriately attaching background components to it and performing a disguise simulation, it is possible to create a more interesting portrait, Moreover, by reading the barcode attached to the hard copy of the portrait into the apparatus, a portrait image created from a favorite face photograph can be repeatedly and easily created.

The above first, second, third, and
In the fourth embodiment, each function realizing means shown in FIG. 1, FIG. 20, FIG. 40, and FIG. 45 can be realized by a computer program. floppy disk,
It will be provided stored in a portable storage medium such as a CD or DVD, and will generally be installed and executed in an auxiliary storage device (such as a hard disk) of a computer.

[0258]

As is apparent from the above description, according to the present invention, an exaggerated face image, which is an essential feature of a portrait, can be automatically created. Further, according to the present invention, it is possible to reduce the amount of face parts required for synthesizing a portrait.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating an overall configuration of a face image creation device according to a first embodiment of the present invention.

FIG. 2 is a flowchart schematically showing an operation of the face image creating device according to the first embodiment of the present invention.

FIG. 3 is a view for explaining feature points of the face in the face image creating apparatus according to the first embodiment of the present invention.

FIG. 4 is a diagram for explaining a method of extracting an eye feature amount in the face image creating apparatus according to the first embodiment of the present invention.

FIG. 5 is a diagram for explaining a method of extracting a feature amount of a mouth in the face image creating device according to the first embodiment of the present invention.

FIG. 6 is a diagram illustrating a method for extracting a feature amount of eyebrows in the face image creating device according to the first embodiment of the present invention.

FIG. 7 is a diagram showing a structure of data output from a feature amount extracting unit to another unit in the face image creating apparatus according to the first embodiment of the present invention.

FIG. 8 is a diagram showing a structure of each data of a feature amount database in the face image creating device according to the first embodiment of the present invention.

FIG. 9 is a flowchart for explaining a configuration of a nonlinear function creating unit in the face image creating apparatus according to the first embodiment of the present invention.

FIG. 10 is a graph for explaining an example of input / output characteristics of a nonlinear function in the face image creating device according to the first embodiment of the present invention.

FIG. 11 is a diagram for explaining a fuzzy inference rule in the face image creating device according to the first embodiment of the present invention.

FIG. 12 is a diagram for explaining a boundary with respect to a feature for selection in the face image creating apparatus according to the first embodiment of the present invention.

FIG. 13 is a flowchart illustrating a configuration of a component arranging unit in the face image creating device according to the first embodiment of the present invention.

FIG. 14 is a diagram for explaining a method of arranging face parts in the face image creating apparatus according to the first embodiment of the present invention.

FIG. 15 is a diagram showing a position of a face part representative point in a face part arrangement in the face image creation device according to the first embodiment of the present invention.

FIG. 16 is a diagram for explaining an example of a plain whole face image in the face image creation device according to the first embodiment of the present invention.

FIG. 17 is a diagram showing some examples of output images in the face image creation device according to the first embodiment of the present invention.

FIG. 18 is a diagram illustrating some examples of output images in the face image creation device according to the first embodiment of the present invention.

FIG. 19 is a diagram showing an example of a feature point extraction result by template matching in the face image creation device according to the first embodiment of the present invention.

FIG. 20 is a block diagram showing an overall configuration of a face image creation device according to a second embodiment of the present invention.

FIG. 21 is a flowchart schematically showing an operation of the face image creating device according to the second embodiment of the present invention.

FIG. 22 is a diagram for describing feature points of a face in the face image creating device according to the second embodiment of the present invention.

FIG. 23 is a diagram for explaining a method of extracting an eye feature amount in the face image creating device according to the second embodiment of the present invention.

FIG. 24 is a diagram for explaining a method for extracting a feature amount of a mouth in the face image creating device according to the second embodiment of the present invention.

FIG. 25 is a diagram for explaining a method of extracting a feature amount of eyebrows in the face image creating device according to the second embodiment of the present invention.

FIG. 26 is a diagram showing a structure of data output from a feature amount extracting unit to another unit in the face image creating apparatus according to the second embodiment of the present invention.

FIG. 27 is a diagram showing a structure of each data of a feature amount database in the face image creating device according to the second embodiment of the present invention.

FIG. 28 is a flowchart illustrating a configuration of a component exaggeration nonlinear function creating unit (when the nonlinear function is a fuzzy rule expression) in the face image creating apparatus according to the second embodiment of the present invention.

FIG. 29 is a flowchart illustrating a configuration of a component exaggeration nonlinear function creating unit (when the nonlinear function is expressed by a mathematical expression) in the face image creating apparatus according to the second embodiment of the present invention.

FIG. 30 is a flowchart for explaining a configuration of a layout exaggeration nonlinear function creating unit (when the nonlinear function is a fuzzy rule expression) in the face image creating apparatus according to the second embodiment of the present invention.

FIG. 31 is a flowchart illustrating a configuration of a layout exaggeration nonlinear function creating unit (when the nonlinear function is expressed by a mathematical expression) in the face image creating apparatus according to the second embodiment of the present invention.

FIG. 32 is a diagram illustrating an example of a nonlinear function in the face image creating device according to the second embodiment of the present invention.

FIG. 33 is a diagram for explaining another example of the nonlinear function in the face image creating device according to the second embodiment of the present invention.

FIG. 34 is a diagram illustrating an example of a fuzzy rule for exaggerating parts in the face image creating apparatus according to the second embodiment of the present invention.

FIG. 35 is a diagram for explaining an example of a fuzzy rule for layout exaggeration in the face image creation device according to the second embodiment of the present invention.

FIG. 36 is a flowchart for describing a configuration of a component exaggeration unit in the face image creation device according to the second embodiment of the present invention.

FIG. 37 is a flowchart for explaining a configuration of a component arranging unit in the face image creating apparatus according to the second embodiment of the present invention.

FIG. 38 is a diagram for explaining a method of arranging face parts in the face image creating apparatus according to the second embodiment of the present invention.

FIG. 39 is a diagram showing the position of a face part representative point in the face part arrangement in the face image creation device according to the second embodiment of the present invention.

FIG. 40 is a block diagram showing an overall configuration of a face image creation device according to a third embodiment of the present invention.

FIG. 41 is a flowchart schematically showing an operation of the face image creating device according to the third embodiment of the present invention.

FIG. 42 is a flowchart illustrating details of image synthesis processing according to the third embodiment of the present invention.

FIG. 43 is a diagram for explaining a relationship between a background component and its reference point in the face image creation device according to the third embodiment of the present invention.

FIG. 44 is a diagram for describing reference points of face parts in the face image creation device according to the third embodiment of the present invention.

FIG. 45 is a block diagram showing an overall configuration of a face image creation device according to a fourth embodiment of the present invention.

FIG. 46 is a flowchart schematically showing an operation of the face image creating device according to the fourth embodiment of the present invention.

FIG. 47 is a flowchart illustrating details of a portrait creation process from a photograph according to the fourth embodiment of the present invention.

FIG. 48 is a flowchart illustrating details of a process for creating a portrait from a barcode according to the fourth embodiment of the present invention.

FIG. 49 is a block diagram for explaining the contents of processing steps of a portrait drawing creation process from a photograph according to the fourth embodiment of the present invention.

FIG. 50 is a block diagram for explaining contents of processing steps of a portrait creation process from a portrait barcode according to the fourth embodiment of the present invention.

FIG. 51 is a diagram illustrating the contents of a component code in the case of an eye component according to the fourth embodiment of the present invention.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 data input means 2 image input means 3 feature quantity extraction means 4 feature quantity storage means 5 nonlinear function creation means 6 contour image extraction means 7 part image storage means 8 component placement means 9 face image output means 101 data input means 102 image input means 103 Feature Amount Extracting Means 104 Feature Amount Storage Means 105 Component Exaggeration Nonlinear Function Creation Means 106 Layout Exaggeration Nonlinear Function Creation Means 107 Contour Image Extraction Means 108 Component Image Storage Means 109 Component Exaggeration Means 110 Component Arrangement Means 111 Face Image Output Means 201 Data input means 202 Portrait creation means 203 Part image storage means 204 Selection menu storage means 205 Image synthesis means 206 Display means 207 Printing means 301 Portrait creation means 302 Background image synthesis means 303 Display means 304 Component code connection means 305 Encoding Hardcopy stage 306 image bar code synthesizing means 307 print unit 308 bar code reading means 309 decode unit 310 bar code 311 Background synthesized portrait image

Continuing on the front page (72) Inventor Masato Kawade 10th Hanazono Todocho, Ukyo-ku, Kyoto-shi, Kyoto, Japan (72) Inventor Tsutomu Ishida 10th Hanazono-Todocho, Ukyo-ku, Kyoto-shi, Kyoto Omron Corporation ( 72) Inventor Yoshiro Tasaka 10 Okado Dodocho, Ukyo-ku, Kyoto-shi, Japan OMRON Corporation (56) References JP-A-7-239945 (JP, A) JP-A-7-85240 (JP, A) JP-A-6-324670 (JP, A) JP-A-4-360276 (JP, A) JP-A-7-129787 (JP, A) JP-A 9-161058 (JP, A) JP-A 9-161086 (JP) , A) Caricature generation system using optical illusion P ICASSO Kazuto Murakami Koshi Miki Yamato Akira Nakayama Akio Fukumura Transactions of the Information Processing Society of Japan Vol. 34 Number 10 P.E. 2106-2116 Published October 15, 1993 (58) Fields investigated (Int. Cl. ⁷ , DB name) G06T 1/00 280 G06T 11/80 Patent file (PATOLIS) JICST file (JOIS)

Claims (10)

(57) [Claims] 1. A face image input means for inputting a face image, and a face image input from the face image input means is analyzed to extract a face feature quantity which is a feature quantity related to a shape and an arrangement of a face part. A face feature amount extraction unit, a face part data storage unit for storing a plurality of face part data prepared in advance for each type of face part, and a shape of the face part from the face part data storage unit. Face part data extraction means for extracting face part data corresponding to the shape feature amount for each type of face part by using the feature amount related to the shape feature amount, and a position for arranging the face part represented by the extracted face part data A change parameter input means for inputting a change parameter for controlling the change of the face part; correcting the arrangement feature amount of the face part in the face feature amount based on the change parameter; Portrait creation device, characterized by comprising: a face component placement unit, the a face part patterns corresponding to the face component data the position indicated by the arrangement feature quantities said modified. 2. The portrait drawing apparatus according to claim 1, wherein the change parameter includes gender information of the portrait creation target person and / or information on whether the portrait creation target person is an adult. 3. A face feature amount extraction step of analyzing a given face image to extract a face feature amount that is a feature amount related to the shape and arrangement of the face part, and a plurality of face feature amounts stored in advance for each type of face part. From among the face part data of the above, using a feature amount related to the shape of the face part,
A face part data extraction step of extracting face part data corresponding to the shape feature amount for each type of face part; and a change parameter for controlling a change of a position at which a face part represented by the extracted face part data is arranged. Inputting a change parameter, and correcting the arrangement feature of the face part in the face feature based on the input change parameter to correspond to the face part data extracted in the face part data extraction step. And a face part arranging step of arranging the face part pattern to be arranged at a position indicated by the corrected arrangement feature amount. 4. The portrait drawing method according to claim 3, wherein the change parameter includes information on the sex of the portrait creation target person or information on whether the portrait creation target person is an adult. 5. A face image input means for inputting a face image; a face part feature quantity extraction means for extracting feature quantities of face parts from the input face image; Face part data storage means for storing a plurality of pieces of face part data obtained, and a face part for extracting face part data used in creating a portrait for each type of face part from the face part data storage means Extraction means; correlation generation means for generating a correlation for correcting the size or shape of the face part represented by the extracted face part data based on the feature amount of the face part; and storing the face part data. A portrait synthesizing means for synthesizing a portrait using a facial part obtained by correcting the size or shape of the facial part data read from the means by the correlation. 6. The portrait drawing apparatus according to claim 5, wherein the correlation generation section generates a correlation based on the gender of the portrait creation target person or information on whether or not the person is an adult. 7. A face image input unit for inputting a face image, and a face feature amount extraction unit that analyzes the input face image and extracts a face feature amount that is a feature amount related to the shape and arrangement of a face part. Means, face part data storage means for storing a plurality of face part data prepared in advance for each type of face part, and selecting face part data to be used in the portrait from the face part data storage means. Parameter input means for inputting parameters used in the process, face part data to be used in a portrait, face part extraction means for extracting each face part type from the face part data storage means based on the parameters, A face part correction unit configured to correct the size or shape of the face part represented by the extracted face part data based on a feature amount related to the shape of the face part; A portrait synthesizing means for arranging a facial part obtained by the correction means based on the feature amount related to the arrangement and synthesizing a portrait. 8. A face part feature amount extracting step of analyzing a given face image to extract a feature amount of the face part, and a plurality of face part data prepared and stored in advance for each type of face part. A face part data product extraction step of extracting face part data used in portrait creation for each type of face part; and determining the size or shape of the face part represented by the extracted face part data by the face part Using a face part obtained by correcting the size or shape of the face part expressed by the extracted face part data by the correlation, A portrait creating method, comprising: a portrait synthesizing step of synthesizing a portrait. 9. The portrait drawing method according to claim 8, wherein in the correlation generation step, a correlation is generated based on the sex of the portrait creation target person or information on whether or not the portrait is an adult. 10. A face image inputting step of inputting a face image for which a portrait is to be created, and a face for analyzing the input face image to extract a face feature quantity which is a feature quantity related to the shape and arrangement of the face part. A feature amount extracting step; a parameter inputting step of inputting a parameter used when selecting face part data to be used in a portrait from a plurality of face part data prepared in advance for each type of face part; A face part extraction step of extracting face part data to be used in a portrait from the plurality of pieces of face part data based on the parameters based on the parameters; A face part correction step of correcting the size or shape of the face part based on a feature amount related to the shape of the face part; Composite picture editing method characterized by comprising a face part obtained by modifying, with the portrait synthesis step of synthesizing a portrait arranged based on the feature quantity relating to the arrangement, the.