CN100456326C - Image processing method for face ageing - Google Patents

Abstract

The disclosed image processing method for aged face comprises: re-sampling to rebuild face image for the elderly; converting the image from RGB space into YUV space, taking Laplace modeling to face wrinkle and applying result on the earlier face, and converting the result from YUV space into RGB space. This invention is fast and low cost.

Description

A kind of image processing method that is used for face ageing

Technical field

The present invention relates to image processing method, particularly relate to a kind of image processing method that is used for face ageing.

Background technology

One of 20th century mankind's outstanding achievement computer technology has been brought human society into the information age.Computer technology has all brought revolutionary variation for every field.Along with going deep into of computer utility, computer graphics important effect of play more and more in recreation and making video, handle by stunt, people can produce many role or the scenes that can't imitate by actor simply, make picture more true to nature and natural.These stunts at many recreation and film with having obtained application, such as " The Lord of the Rings ", " Troy " etc.These stunts are advanced to the digital entertainment epoch with human society.And make the flourish power and source of digital entertainment is exactly to be the Digital Image Processing/computer graphics and the multimedia technology of carrier with the computing machine.

The ageing of people's face is the problem that often will face in image or the game making, and under the situation, people reach such effect by making up usually, the length that expends time in, and the effect instability usually makes the people feel true inadequately.The facial image that obtains ageing by digital image processing method is a kind of low cost, method fast, old man's face generally has more wrinkle and spot, at this characteristics, face ageing is carried out modeling, by the old human face photo in the sample, the ageing of more convenient realization target people face is handled.The method that face ageing is carried out in traditional Digital Image Processing generally need a people by youth to old process photo, by the subtracting each other or relatively obtain the ageing operating value of front and back photo, and this class photo is difficult to obtain in practice.

Summary of the invention

The object of the present invention is to provide a kind of Laplce of employing to operate the image processing method that to color of image gradient modeling is used for face ageing.

The technical scheme that the present invention solves its technical matters employing is as follows:

1. image processing method that is used for face ageing, the step of this method is as follows:

1) at first, facial image is resampled, realize reconstruction, and old man's image that the young man is rebuild is transformed into yuv space by rgb space old everybody face;

2) secondly, people's face wrinkle is carried out Laplce's modeling;

3) last, modeling result is applied to young man's face, and the result is converted to rgb space by yuv space.

2. described facial image is resampled, realizes reconstruction, and old man's image that the young man rebuilds is transformed into yuv space by rgb space is old everybody face:

1) by young man's face old man's face is resampled, realize reconstruction the elderly's face.Resampling comprises two parts:

A) the elderly and youngster's face are carried out the unique point mark

B) label alignment the elderly face and the young man's face that provides by unique point, the elderly's face that the elderly's face pixel is sampled and to be rebuild

2) at first calculate the brightness value (Y) of each pixel in the YUV color space, by formula carry out (1).

Y＝0.299R+0.587G+0.114B

U＝-0.147R-0.289G+0.436B (1)

V＝0.615R-0.515G-0.100B

Wherein, Y is the Y component in the yuv space, and R, G, B are respectively the numerical value of three components of pixel in 16 RGB images.

3. describedly people's face wrinkle carried out Laplce's modeling be:

Extract the local laplace model of each pixel in the elderly's face and the young man's image respectively at level and vertical both direction.During modeling, carry out the calculating of horizontal direction, carry out the calculating of vertical direction according to formula (3) according to formula (2).

$g_x=e-\frac{a+c}{2}-(e_1-\frac{a_1+c_1}{2})-(e_2-\frac{a_2+c_2}{2})---\left(2\right)$

Wherein: g _xRepresent the local laplace model of horizontal direction, the e representative waits to ask the brightness value of pixel, and a represents the pixel brightness value on the e left side, and c represents the pixel brightness value on e the right, similarly, and e ₁, e ₂, a ₁, a ₂, c ₁And c ₂Expression and front e, a, the elderly's face of c correspondence and the pixel in young man's face.

$g_y=e-\frac{b+d}{2}-(e_1-\frac{b_1+d_1}{2})-(e_2-\frac{b_2+d_2}{2})---\left(3\right)$

Wherein, g _yRepresent the local laplace model of horizontal direction, b is the brightness value of the pixel above the e, and d is the pixel brightness value below the e, similarly, and b ₁, b ₂, d ₁And d ₂Expression and front b, the elderly's face of d correspondence and the pixel in young man's face.

4. described modeling result is applied to young man's face, and the result is converted to rgb space by yuv space is:

1) value of e is calculated according to formula (4)

$e=e_1+e_2+\frac{a+b+c+d}{4}-\frac{a_1+b_1+c_1+d_1}{4}-\frac{a_2+b_2+c_2+d_2}{4}---\left(4\right)$

2) e is carried out iteration and upgrade computing, up to ε _i＜0.01, ε wherein _iBe iteration error, calculate according to formula (5)

${\mathrm{\ϵ}}_i=\underset{n=1}{\overset{N}{\mathrm{\Σ}}}\mathrm{intensity}{\left(n\right)}_i-\mathrm{intensity}{\left(n\right)}_{i-1}---\left(5\right)$

Wherein, N is the number of pixel, and n is the sequence number of pixel, and i is meant number of iterations, the brightness value of n pixel of intensity (n) expression.

3) according to formula (6) image is converted to rgb space by yuv space.

R＝Y+1.140V

G＝Y-0.395U-0.581V (6)

B＝Y+2.032U。

In Digital Image Processing and computer vision, Laplace operator is widely used in rim detection in digital picture for people's wrinkle and spot, and the part that graded in the image is violent is carried out modeling.Laplce's operation can be the color gradient field model with image transitions, and wrinkle and spot that this can handle the elderly face just carry out modeling to it.In digital picture, people's face wrinkle and spot show as the variation of brightness in the image, therefore, realize the ageing operation by the modeling to brightness.

The beneficial effect that the present invention has is: it is applicable to image or game making, old man's face generally has more wrinkle and spot, at this characteristics, face ageing is carried out modeling, by the old human face photo in the sample, the ageing of more convenient realization target people face is handled.In addition, the present invention do not need a people by youth to old variation photo, data are obtained conveniently.Therefore, the facial image that obtains ageing by the present invention is a kind of low cost, method fast.

Description of drawings

Fig. 1 is human face characteristic point (a) and the topological structure (b) thereof that the present invention adopts;

Fig. 2 is the synoptic diagram that a pixel is carried out the local modeling of Laplce;

Embodiment

When implementing the image processing method of face ageing, the specific implementation flow process is as follows:

The first step: prepare data.At first prepare an old human face photo (source human face photo) and a human face photo (target human face photo) that needs ageing according to the ageing needs.

Second step: human face characteristic point is demarcated.According to shown in Figure 1, on source human face photo and target human face photo, demarcate corresponding human face characteristic point (Fig. 1 (a)) respectively, and set up corresponding topological structure (Fig. 1 (b)).

The 3rd step: utilize unique point and topological structure thereof, calculate the relative position of pixel in the target people face, realize index, people from source face is sampled, realize the foundation of people from source face and target people face corresponding relation with this index to pixel in the target people face.

The 4th step: color space is transformed into YUV by RGB.Mode according to definition in the formula (1) is carried out the conversion of color space.

Y＝0.299R+0.587G+0.114B

U＝-0.147R-0.289G+0.436B(1)

V＝0.615R-0.515G-0.100B

Wherein, Y is the Y component in the yuv space, and R, G, B are respectively the numerical value of three components of pixel in 16 RGB images.

The 5th step: to the local Laplce's modeling of wrinkle variation carrying out.Carry out local Laplce's modeling of wrinkle mapping according to formula (2)-(4), carry out local message that modeling adopts as shown in Figure 2.

Extract the local laplace model of each pixel in the elderly's face (Fig. 2 (a)) and the young man's image (Fig. 2 (a)) respectively at level and vertical both direction.During modeling, carry out the calculating of horizontal direction, carry out the calculating of vertical direction according to formula (3) according to formula (2).

$g_x=e-\frac{a+c}{2}-(e_1-\frac{a_1+c_1}{2})-(e_2-\frac{a_2+c_2}{2})---\left(2\right)$

Wherein: g _xRepresent the local laplace model of horizontal direction, the e representative waits to ask the brightness value of pixel, and a represents the pixel brightness value on the e left side, and c represents the pixel brightness value on e the right, similarly, and e ₁, e ₂, a ₁, a ₂, c ₁And c ₂Expression and front e, a, the elderly's face of c correspondence and the pixel in young man's face.

$g_y=e-\frac{b+d}{2}-(e_1-\frac{b_1+d_1}{2})-(e_2-\frac{b_2+d_2}{2})---\left(3\right)$

Wherein, g _yRepresent the local laplace model of horizontal direction, b is the brightness value of the pixel above the e, and d is the pixel brightness value below the e, similarly, and b ₁, b ₂, d ₁And d ₂Expression and front b, the elderly's face of d correspondence and the pixel in young man's face.

The 6th step: carry out interative computation until convergence.

1) value of e (Fig. 2 (c)) is calculated according to formula (4).

$e=e_1+e_2+\frac{a+b+c+d}{4}-\frac{a_1+b_1+c_1+d_1}{4}-\frac{a_2+b_2+c_2+d_2}{4}---\left(4\right)$

2) e is carried out iteration and upgrade computing, up to ε _i＜0.01, ε wherein _iBe iteration error, calculate according to formula (5).

${\mathrm{\ϵ}}_i=\underset{n=1}{\overset{N}{\mathrm{\Σ}}}\mathrm{intensity}{\left(n\right)}_i-\mathrm{intensity}{\left(n\right)}_{i-1}---\left(5\right)$

Wherein, N is the number of pixel, and n is the sequence number of pixel, and i is meant number of iterations, the brightness value of n pixel of intensity (n) expression.

The 7th step: result of calculation is changed back the RGB color space according to formula (6).

R＝Y+1.140V

G＝Y-0.395U-0.581V (6)

B＝Y+2.032U

Therefore, the data of this method are obtained conveniently, and computation complexity is also lower, realize simple and fast, can reduce the cost of manufacture of image, recreation etc.

Claims (1)

1. image processing method that is used for face ageing is characterized in that the step of this method is as follows:

1) at first, facial image is resampled, realize reconstruction, and old man's image that the young man is rebuild is transformed into yuv space by rgb space old everybody face;

Described facial image is resampled, realizes reconstruction, and old man's image that the young man rebuilds is transformed into yuv space by rgb space is old everybody face:

A) by young man's face old man's face is resampled, realizes reconstruction the elderly's face,

Resampling comprises two parts:

A) the elderly and youngster's face are carried out the unique point mark

B) label alignment the elderly face and the young man's face that provides by unique point, the elderly's face that the elderly's face pixel is sampled and to be rebuild

B) at first calculate the brightness value (Y) of each pixel in the YUV color space, by formula carry out (1)

Y＝0.299R+0.587G+0.114B

U＝-0.147R-0.289G+0.436B (1)

V＝0.615R-0.515G-0.100B

Wherein, Y is the Y component in the yuv space, and R, G, B are respectively the numerical value of three components of pixel in 16 RGB images;

2) secondly, people's face wrinkle is carried out Laplce's modeling;

Describedly people's face wrinkle carried out Laplce's modeling be:

Extract the local laplace model of each pixel in the elderly's face and the young man's image respectively at level and vertical both direction, during modeling, carry out the calculating of horizontal direction, carry out the calculating of vertical direction according to formula (3) according to formula (2)

$g_x=e-\frac{a+c}{2}-(e_1-\frac{a_1+c_1}{2})-(e_2-\frac{a_2+c_2}{2})---\left(2\right)$

Wherein: g _xRepresent the local laplace model of horizontal direction, the e representative waits to ask the brightness value of pixel, and a represents the pixel brightness value on the e left side, and c represents the pixel brightness value on e the right, similarly, and e ₁, e ₂, a ₁, a ₂, c ₁And c ₂Expression and front e, a, the elderly's face of c correspondence and the pixel in young man's face,

$g_y=e-\frac{b+d}{2}-(e_1-\frac{b_1+d_1}{2})-(e_2-\frac{b_2+d_2}{2})---\left(3\right)$

Wherein, g _yRepresent the local laplace model of horizontal direction, b is the brightness value of the pixel above the e, and d is the pixel brightness value below the e, similarly, and b ₁, b ₂, d ₁And d ₂Expression and front b, the elderly's face of d correspondence and the pixel in young man's face;

3) last, modeling result is applied to young man's face, and the result is converted to rgb space by yuv space;

Described modeling result is applied to young man's face, and the result is converted to rgb space by yuv space is:

A) value of e is calculated according to formula (4)

$e=e_1+e_2+\frac{a+b+c+d}{4}-\frac{a_1+b_1+c_1+d_1}{4}-\frac{a_2+b_2+c_2+d_2}{4}---\left(4\right)$

B) e is carried out iteration and upgrade computing, up to ε _i＜0.01, ε wherein _iBe iteration error, calculate according to formula (5)

${\mathrm{\ϵ}}_i=\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{\mathrm{intensity}\left(n\right)}_i-{\mathrm{intensity}\left(n\right)}_{i-I}---\left(5\right)$

Wherein, N is the number of pixel, and n is the sequence number of pixel, and i is meant number of iterations, the brightness value of n pixel of intensity (n) expression,

C) according to formula (6) image is converted to rgb space by yuv space,

R＝Y+1.140V

G＝Y-0.395U-0.581V (6)

B＝Y+2.032U。

Images