Invention content
The purpose of the present invention is to solve existing the problem of being easily tampered using the database of three-dimensional face model, sheets
Invention provides a kind of method for building up of four-dimensional faceform's database.
The method for building up of four-dimensional faceform's database of the present invention, described method includes following steps:
Step 1:Current time is taken to be used as with reference to amount, settling time axis;
Step 2:According to the facial image of input, space two-dimensional coordinate system is established, X-Y scheme is carried out to the facial image
As information extraction and face characteristic are extracted;
Step 3:Structure and the fusion of 3-D view are carried out according to the two-dimensional image information of acquisition and face characteristic;
Step 4:According to the 3-D view after fusion, 3 d space coordinate system is established, forms three-dimensional face model, extraction
The three-dimensional image information of three-dimensional face model;
Step 5:Each pixel of three-dimensional face model with the time of time shaft established is corresponded, forms four
Faceform is tieed up, four-dimensional image information is extracted to four-dimensional faceform, is stored into database, forms four-dimensional faceform's data
Library.
In the step 2, according to the facial image of input, space two-dimensional coordinate system is established, the facial image is carried out
Two-dimensional image information extracts and the method for face characteristic extraction includes:
The facial image of input is pre-processed, chooses the center of facial image two after pretreatment as origin, X
Axis passes through the center of two, and horizontal is to the right positive direction, and Y-axis is straight up positive direction, completes building for space two-dimensional coordinate system
It is vertical;
The scanning of Pixel-level is carried out to the two dimensional image after foundation, obtains the coordinate position and gray value of each pixel,
Two dimensional image textural characteristics are formed, and obtain the color value of each pixel by clustering method, the color value is with RGB come table
Show;
The division of face face structure is carried out to the two dimensional image after foundation;
Characteristics of image edge detection and enhancing are carried out to the two dimensional image after foundation, it, will according to the face structure after division
Image is divided into each separate unit, then carries out feature extraction to each separate unit.
In the step 2, being divided into for face face structure is carried out to the image after foundation:
Forehead, eyebrow, upper eyelid, eyelash, fissura palpebrae, commissura palpebrarum medialis area, association area, endocanthion, outer canthus, nose on the outside of eyelid
Root, nose, bridge of the nose, nose bottom, nostril, nose pillar, the wing of nose, nasal surface ditch, labial surface ditch, nasolabial groove, infraorbital region zygomatic region, schistostoma, bicker,
Lip is red, prolabium, lip bow, people midpoint, lip peak, lip pearl, philtrum ridge, buccal region, chin labial groove, gnathion, zygomatic region, Nie Mian areas, ear,
Parotideomasseteric region, hair and beard.
The facial image of the input includes face direct picture, face orthogonal image and face side image.
In the step 4, according to the 3-D view after fusion, 3 d space coordinate system is established, forms three-dimensional face mould
Type, the method for extracting the three-dimensional image information of three-dimensional face model include:
To the 3-D view after fusion, the center for choosing face two is origin, and X-axis passes through the center of two, from
Front sees to be positive direction from left to right;Y-axis is positive direction straight up;Z axis passes through head center outside, is from front to back pros
To, the foundation of completion 3 d space coordinate system, then based on CANDIDE-3 models, establish three-dimensional face model;
The scanning of Pixel-level is carried out to the three-dimensional face model of foundation, obtains the coordinate position and gray scale of each pixel
Value forms 3-D view textural characteristics, and obtain the color value of each pixel, the color value RGB by clustering method
To represent.
In the step 5, extraction four-dimensional image information includes coordinate position, textural characteristics, color value and time.
The beneficial effects of the present invention are, four-dimensional faceform's database of the invention include using current settling time as
The time shaft of reference time, when someone changes the four-dimensional faceform's database established, because of modification time and settling time pair
Should not on, and forbid being changed, so four-dimensional faceform's database security higher there is provided time shaft.
Specific embodiment
The present embodiment will be described with reference to Fig. 1, the method for building up of four-dimensional faceform's database described in present embodiment,
Step 1:Current time is taken to be used as with reference to amount, settling time axis;
Step 2:According to the facial image of input, space two-dimensional coordinate system is established, X-Y scheme is carried out to the facial image
As information extraction and face characteristic are extracted;
In the step 2, according to the facial image of input, space two-dimensional coordinate system is established, the facial image is carried out
Two-dimensional image information extracts and the method for face characteristic extraction includes:
The facial image of input is pre-processed, chooses the center of facial image two after pretreatment as origin, X
Axis passes through the center of two, and horizontal is to the right positive direction, and Y-axis is straight up positive direction, completes building for space two-dimensional coordinate system
It is vertical;
The scanning of Pixel-level is carried out to the two dimensional image after foundation, obtains the coordinate position and gray value of each pixel,
Two dimensional image textural characteristics are formed, and obtain the color value of each pixel by clustering method, the color value is with RGB come table
Show;
The division of face face structure is carried out to the two dimensional image after foundation;
Characteristics of image edge detection and enhancing are carried out to the two dimensional image after foundation, it, will according to the face structure after division
Image is divided into each separate unit, then carries out feature extraction to each separate unit.
1 two-dimensional image information of table
Attribute |
Coordinate |
Texture |
Color |
Value |
(X,Y) |
(U,V) |
(R,G,B) |
In the step 2, being divided into for face face structure is carried out to the image after foundation:
Forehead, eyebrow, upper eyelid, eyelash, fissura palpebrae, commissura palpebrarum medialis area, association area, endocanthion, outer canthus, nose on the outside of eyelid
Root, nose, bridge of the nose, nose bottom, nostril, nose pillar, the wing of nose, nasal surface ditch, labial surface ditch, nasolabial groove, infraorbital region zygomatic region, schistostoma, bicker,
Lip is red, prolabium, lip bow, people midpoint, lip peak, lip pearl, philtrum ridge, buccal region, chin labial groove, gnathion, zygomatic region, Nie Mian areas, ear,
Parotideomasseteric region, hair and beard.
The facial image of the input includes face direct picture, face orthogonal image and face side image.
The division of above-mentioned face face structure is by division methods medically, using drawing for this face face structure
Point, then carry out feature extraction respectively, resolution higher, and then in recognition of face, no matter identified face which angle, all very
It is easily identified, present embodiment sensitivity higher.
Step 3:Structure and the fusion of 3-D view are carried out according to the two-dimensional image information of acquisition and face characteristic;
The structure of 3-D view is Viola Face datections, base based on Adaboost algorithm with merging in present embodiment
It is realized in the interpolation algorithm of CANDIDE-3 models three-dimensional modeling, the image filtering based on Sobel operators and non-characteristic point.It is described
Method is all technical means commonly used by such a person skilled in the art;
Because the variation of frame can cause the picture of every frame to change, therefore in dynamic picture in the 3-D view of structure
It can cause the background of picture or the spatial position of personage and characteristic value that a variation occurs, so needing to melt this series of images
It closes, present embodiment is mainly using current integration technology, such as pyramid fusion or Wavelet Transform Fusion technology progress image
Mixing operation.
Step 4:According to the 3-D view after fusion, 3 d space coordinate system is established, forms three-dimensional face model, extraction
The three-dimensional image information of three-dimensional face model;
In the step 4, according to the 3-D view after fusion, 3 d space coordinate system is established, forms three-dimensional face mould
Type, the method for extracting the three-dimensional image information of three-dimensional face model include:
To the 3-D view after fusion, the center for choosing face two is origin, and X-axis passes through the center of two, from
Front sees to be positive direction from left to right;Y-axis is positive direction straight up;Z axis passes through head center outside, is from front to back pros
To, the foundation of completion 3 d space coordinate system, then based on CANDIDE-3 models, establish three-dimensional face model;
The scanning of Pixel-level is carried out to the three-dimensional face model of foundation, obtains the coordinate position and gray scale of each pixel
Value forms 3-D view textural characteristics, and obtain the color value of each pixel, the color value RGB by clustering method
To represent.
2 three-dimensional image information of table
Attribute |
Coordinate |
Texture |
Color |
Value |
(X,Y,Z) |
(U,V) |
(R,G,B) |
Step 5:Each pixel of three-dimensional face model with the time of time shaft established is corresponded, forms four
Faceform is tieed up, four-dimensional image information is extracted to four-dimensional faceform, is stored into database, forms four-dimensional faceform's data
Library.
In the step 5, extraction four-dimensional image information includes coordinate position, textural characteristics, color value and time.
The relation schema of four-dimension faceform's database includes face space three-dimensional relation schema, time dimension relationship mould
Formula and other characteristic relation patterns;
In the acquisition of information of threedimensional model, need to add in time dimension in each pixel, formed one it is four-dimensional
Faceform, four-dimensional image information are as shown in table 3:
Table 3:Four-dimensional image information
Attribute |
Coordinate |
Texture |
Color |
Time |
Value |
(X,Y,Z) |
(U,V) |
(R,G,B) |
T |
It is that textural characteristics are described according to the matrix of the grey level histogram of pixel in present embodiment.