CN101515374A - Individualized realistic virtual character modeling method based on images - Google Patents

Abstract

The invention provides an individualized realistic virtual character modeling method based on images. Four human body images on the facade, lateral surfaces and back quickly generate an individualized three-dimensional human body model with realistic appearance. The method comprises the following steps: using a human body model based on a swept surface as a human body template model; extracting human body characteristic information from the images, establishing a corresponding relation between the human body template model and the human body images, and deforming the human body template model to generate an individualized human body geometric model; and on the two-dimensional parameterized results of the human body model, generating vivid human body textures by the human body images, and seamlessly mapping the human body textures to the three-dimensional individualized human body model to obtain the individualized human body model with realistic visual effect. The human body images are from a camera, a video camera and the like. The method provides a new low-cost and efficient human body modeling path for the field of virtual reality and digital entertainment.

Description

Individualized realistic virtual character modeling method based on image

Technical field

The invention belongs to computer graphic image treatment technology and virtual reality and field of digital entertainment, relate to a kind of utilization from the front, four width of cloth human body images of side, back side picked-up, generate the method for the personalized manikin of realistic visual appearance.

Background technology

3 D human body moulding at present realizes by following three kinds of methods mostly: the method for the method of Geometric Modeling, three-dimensional reconstruction and based on the method for sample alternately.

The method of mutual Geometric Modeling mostly according to human anatomic structure and profile nature, by means of 3D sculpting softwares such as Maya, is finished by the mode of manual interaction.Usually need the art designing personnel of specialty could be competent at modeling work, Production Time is longer, and cost is higher.

The method of rebuilding, by the mode of spatial digitizer or structured light, can obtain meticulous geometric model, but among the result who rebuilds, usually there is a large amount of noise datas and, need does a large amount of subsequent treatment work owing to the model cavity that model blocks certainly or the light shade causes.

Based on the method for sample,, obtain meeting the manikin of personal characteristics by interpolation or distortion sample manikin.This method has been utilized the constant structure of existing model, can avoid the model cavitation in the method for reconstructing, reduces the model surface The noise, has demonstrated numerous superiority in people's volume modeling.Method based on sample interpolation, based on big-sample data, suppose that the individual character model can obtain by the linear combination to sample, but when personality exceeds the sample set span, with no longer valid, thereby these class methods are not suitable for making up the manikin that is installed with loose dress ornament.Based on the validity of sample deformation method, then be not subjected to the influence of personality value.By importing the personalized existing manikin of transformation of data such as human dimension, obtain personalized results model.This method based on sample distortion in recent years, obtains more application in the three-dimensional personalized modelling of human body based on human body image, but existent method obtains by the summit of distorted pattern one by one mostly, realizes texture by cylindrical surface projecting.This based on the method for vertex deformation one by one, lack the semantic information of human body usually, cause the deformation process efficient of model not high, deformation process is easy to generate model selfing phenomenon etc. simultaneously; And, human body need be divided into trunk and 5 zones of four limbs by cylindrical surface projecting mapping texture, and distinguish texture, this is easy to cause the distortion and the texture crack of texture on the border of adjacent area, influenced visual effect.

Summary of the invention

The objective of the invention is to: based on human body image, provide a kind of operation simple and easy, with low cost, can generate the personalized visual human of the sense of reality of band dress ornament texture apace.The distortion of manikin has the advantage of efficient height, algorithm robust based on the feature semanteme of human body; By generating complete human body texture, realized seamless texture.Method provided by the invention may further comprise the steps:

Step 1: structure is the basic geometric element of sweeping surface based on the manikin behaviour phantom template die type of sweeping surface with the planar cross-sectional ring, and each point on the manikin all belongs to corresponding cross section ring;

Step 2: in input picture, be partitioned into human body image, the human body characteristic information;

Step 3: obtain human geometry's information from image, distortion 3 D human body template model generates personalized human geometry's model;

Step 4: by two-dimensional parameterization, obtain the texture coordinate of personalized manikin, on manikin two-dimensional parameter result, generate human body texture true to nature by four width of cloth quadrature human body images to manikin;

Step 5: described texture seamlessly is mapped on the personalized manikin, obtains the visual effect of the sense of reality.

Further, the input picture in the described step 2 comprises front, side and the back side four width of cloth images.

Further, the characteristics of human body's information in the described step 2 comprises: human body feature point, human skeleton etc.

Further, the quadrature human body image in the described step 4 is four width of cloth quadrature human body images.

Further, described step 1 comprises:

Step 11: according to the physiological structure information of human body, extract human body feature point, generate human skeleton;

Step 12: the orthogonal frame UVW that arbitrfary point on the human skeleton is set;

Step 13: the sampling human skeleton, do the cross section according to the local frame on the sampled point to manikin, the cross section ring that obtains is the basic geometric element in the sweeping surface;

Step 14: the relative coordinate relation of setting up ring center, cross section and skeleton;

Step 15: sampling cross section ring generates the grid vertex of sweeping surface, and the trigonometric ratio grid vertex, obtains sweeping surface; Set up cross section ring and grid vertex subordinate relation; Human skeleton, cross section ring sweeping surface constitute the body templates model.

Further, described step 2 comprises:

Step 21: on input picture, alternatively be partitioned into human body image; According to human body physiological characteristics information, on the human body image profile, extract human joint points;

Step 22: connect adjacent articulation point, generate the human skeleton on the image.

Further, described step 3 comprises:

Step 31: set up body templates model and the corresponding relation of two-dimension human body image on human skeleton;

Step 32: set up the corresponding relation between body templates model upper section ring and the human body image;

Step 33: obtain the geometry deformation parameter of cross section ring, obtain the individual character manikin by the pair cross-section annular strain.

Further, described step 4 comprises:

Step 41: the personalized manikin that will seal automatically cuts into the curved surface that is equal to plate-like on the topology, and parametrization obtains the texture coordinate of personalized manikin to two dimensional surface;

Step 42: on different direction of visual lines, blocking certainly between automatic human body image and the three-dimensional (3 D) manikin upper limb body;

Step 43: by human body image, on the two-dimensional parameter result of manikin, generate complete human body texture, comprise processing to shield portions texture on the three-dimensional model.

The invention has the beneficial effects as follows: adopt above-mentioned steps, can realize by easy operation, generating the personalized manikin of realistic visual appearance apace by front, side and the back side four width of cloth images of ordinary camera or video camera picked-up.In the present invention, by the relativeness between cross section ring and the human skeleton, set up the inner link between human body curved surface and the skeleton, make each point on the human body curved surface all have the feature semantic information, be the foundation of corresponding relation between manikin and the human body image, and from image, obtain individual character human bodily form parameter and provide a kind of quick, effective method by image generation human body texture true to nature.By control cross section annular shape, distortion body templates model generates personalized manikin, and it is higher on efficient to compare traditional deformation method based on grid vertex, and more insensitive to the human body image noise.The present invention provides a kind of modeling method fast and effectively for personalized manikin embeds in reality environment, gamespace, movie and video programs.

Description of drawings

Fig. 1 is based on the personalized modelling of human body flow process of the three dimension realistic of image;

Fig. 2 is the personalized modelling of human body legend of a three dimension realistic process;

Fig. 3 (a)～Fig. 3 (e) is the body templates model generative process based on sweeping surface;

Fig. 4 (a) and Fig. 4 (b) are cut apart for the human body image of cutting based on figure;

Fig. 5 (a) and Fig. 5 (b) are unique point and the skeleton on the human body image;

Fig. 6 (a)～Fig. 6 (c) is the individual character manikin generation based on the cross section annular strain;

Fig. 7 (a) and Fig. 7 (b) are the three-dimensional (3 D) manikin two-dimensional parameterization;

Fig. 8 blocks judgement certainly for manikin under each direction of visual lines;

Fig. 9 (a) and Fig. 9 (b) generate the human body texture on the parametrization result;

Figure 10 is the texture result on personalized manikin.

Embodiment

Fig. 1 has provided based on the personalized modelling of human body flow process of the three dimension realistic of image.Whole flow process is divided three parts, and promptly the template manikin generates (corresponding to following step 1); The individual character human body information extracts (corresponding to following step 2) in the image; (corresponding to following step 3), and the texture that obtains the sense of reality (corresponding to following step 4), realizes that seamless texture effect is (corresponding to following step 5) by the individual character human body information template manikin to be deformed into personalized manikin.Fig. 2 then further uses marginal data based on the personalized modelling of human body flow process of the three dimension realistic of image.In Fig. 2, first row's picture represents that the input data of personalized modelling of human body are human body image and body templates model based on sweeping surface of four width of cloth quadratures; The second row figure expression can generate the personalized human geometry's model and the human body texture of the sense of reality by these input data; The 3rd then expression of row figure to human geometry's model, is exported the personalized manikin of realistic visual effect with the human body texture.Concrete by following 5 steps enforcement:

Step 1: structure is based on the manikin behaviour phantom template die type of sweeping surface.

Shown in Fig. 3 (a), import common three-dimensional (3 D) manikin, require the four limbs of model to part a little.Model is adjusted into positive towards computer screen.By the coordinate extreme point of manikin, obtain human body feature points such as the crown, finger tip, toe point.These unique points also be human skeleton overhead, the articulation point of finger tip, toe point.From the beginning the summit begins, and is parallel to the plane of human body short transverse with normal vector, from the crown down, and the scanning manikin.When the closed cross-section ring when 1 gradually becomes 3, the cut-point between the ring of cross section is the armpit point.Cross the plane that the armpit point is parallel to the human body short transverse and be the shoulder articulation point that is connected arm and shoulder with the mid point that manikin intersects the cross section ring that obtains; This plane is divided into arm and torso portion with human body.Downward from armpit point, continue the scanning trunk, when the closed cross-section ring when 1 becomes 2, the cut-point between the ring of cross section is the perineum point.Crossing perineum point separates both legs and trunk perpendicular to the plane of human body short transverse.

Along human limb and trunk, scan manikin respectively.Physiological characteristic according near the human body curved surface human synovial position has bigger local geometric to change obtains articulation points such as neck, ancon, knee on the mid point of corresponding cross section ring.By mutual adjustment, optimize the articulation point position.Connect the adjacent segment point, obtain the human skeleton model, shown in Fig. 3 (b).

Shown in Fig. 3 (c), set up partial points frame UVW on the human skeleton, wherein the V axle is manikin front direction of visual lines (being generally the Z direction); The U direction is by V * W decision.And the W axle is chosen different directions because of the difference of skeleton region:

(1) for the point on head and the following torso portion human skeleton of armpit, its W axle is parallel to human body short transverse (being generally the Y direction);

(2) for point on the bone that connects two shoulder articulation points, its W axle is parallel to this bone;

(3), calculate by step (1) in the local coordinate at the articulation point place that is connected with trunk for the skeleton on the four limbs; The W axle of the local frame on the skeleton end (being finger tip, toe point) is parallel to the bone that connects this point; The W axle that connects the articulation point place of two adjacent bones is parallel to the normal vector of these two adjacent bone equidistant points.Calculate the W axle of the local frame of the arbitrfary point on the bones by linear interpolation.

Every section bone on the sampling human skeleton on sampled point, is done the cross section to three-dimensional (3 D) manikin, and the normal vector in cross section is parallel to the local frame W direction of principal axis on this sampled point.The cross section ring that cross section and three-dimensional (3 D) manikin intersect is the basic geometric element in the human body sweeping surface, and its result is shown in Fig. 3 (d).

Relativeness between record cross section ring and the human skeleton: shown in Fig. 3 (c), if the intersection point that the cross section is encircled between the bone Q1Q2 corresponding with Si of the plane, place of Si is Pi, the coordinate of the center O i of Si under the local frame of Pi is (ui, vi, wi), then with the position of relative coordinate record Oi with respect to human skeleton:

<|Q1Pi/|Q1Q2|，ui/|Q1Q2|，vi/|Q1Q2|，wi/|Q1Q2|> (1)

Thus can be by the counter center point coordinate of obtaining its corresponding cross section ring of skeleton.

Shown in Fig. 3 (e), to every cross section ring, U axle and the V axle of crossing its center are divided into 4 sections with this cross section ring, to each section, the length of curve sampling such as press, and obtain the net point of sweeping surface.Grid vertex on the Denaulay trigonometric ratio adjacent sections ring obtains sweeping surface.

Step 2: in front, side and the back side four width of cloth images of input, be partitioned into human body image, the human body characteristic information extracts human skeleton.

On input picture, utilize the figure incision technology to be partitioned into human body image.The figure incision technology is the algorithms most in use during computer picture is cut apart, and is a kind of cutting techniques of optimizing based on global energy, by the binaryzation to image, at grey scale change violent prospect and background edge local protectionism image.Shown in Fig. 4 (a), alternatively identify prospect and background image by white and black free curve respectively, be partitioned into human body image shown in Fig. 4 (b) by the figure incision technology.Can be partitioned into the human body image on other direction of visual lines photo similarly.Further in the human body image that is partitioned into, extract human geometry's information, specific as follows:

For the front face human body image, store the coordinate of human body image profile counterclockwise.Shown in Fig. 5 (a),, obtain the crown, finger tip, tiptoe unique point by the extreme point of coordinate; According near the human body curved surface unique point position human body physiological characteristics of bigger variation is arranged usually, extract all the other unique points, as: to the profile line segment of right crus of diaphragm tiptoe, right armpit point has maximum coordinate figure on short transverse at right finger tip; And the right crus of diaphragm point is to the profile line segment of left foot point, and perineum point has maximum coordinate figure on short transverse.For the unconspicuous point of geometric properties, then determine initial position by Human Physiology statistical nature data, probably be positioned at perineum as the waist unique point and put 1/4 place on summit to the end.Further obtain rational unique point by mutual adjustment.

Cross the armpit point and do the sweep trace that is parallel to the human body short transverse, the mid point of the line segment that is intercepted by human body contour outline is the articulation point of arm and shoulder junction in the human skeleton; This sweep trace is divided into arm and trunk part with human body image simultaneously.In like manner, the horizontal scanning line of crossing perineum point can be divided into human body image shank and trunk, and can obtain connecting on the skeleton articulation point of thigh and trunk.The horizontal scanning line of crossing armpit further is subdivided into human body image characteristic areas such as shoulder, head with the horizontal scanning line of crossing neck feature point.

Crossing the human body feature point on the image, do sweep trace to the human body image outline line, is the articulation point of human body with the line segment mid point that is intercepted by human body contour outline.The direction of sweep trace is determined in the following manner: the sweep trace of head and the following torso portion of armpit is perpendicular to the human body short transverse; The sweep trace of shoulder part is parallel to the human body short transverse; The scan-line direction of four limbs is then determined by the articulation point at its two ends, as for a characteristics P on the left hand arm, the vertical direction vector of its sweep trace is (1-α) d1+ α d2, wherein d1=(1,0) is the vertical direction vector that connects the articulation point K1 upper tracer of arm and shoulder, d2=(K2-K1), K2 are left finger tip point, α=| PK1|/| K1K2|.Can define the scan-line direction on right arm and two legs similarly.

For the image of other direction of visual lines photographs, make its profile on short transverse, be aligned in the front face human body image outline.And on the height value identical, obtaining unique point and articulation point on these human body images with the direct picture unique point, its result is shown in Fig. 5 (a).

To every width of cloth human body image, connect adjacent articulation point, generate the human skeleton on the image, shown in Fig. 5 (b).

Step 3: obtain human geometry's information from image, distortion 3 D human body template model generates personalized human geometry's model;

According to the skeleton on positive and the side human body image, recover individual character human body three-dimensional skeleton, promptly the human body image of front image can be considered the projection of three-dimensional framework on the direction of visual lines of front.According to the corresponding relation between the skeleton shangguan node, with the framework deformation in the body templates model to three-dimensional individual character human skeleton, thereby set up skeleton on the 3 D human body template model and the corresponding relation between the two dimensional image upper skeleton indirectly.Upgrade the local frame on the template model human skeleton, and upgrade body templates model upper section direction of loop by formula (1).

For each the cross section ring on the body templates model, according to the corresponding relation between the human skeleton on 3 D human body template model skeleton and the two dimensional image, on positive and side human body image, respectively by the U of its frame, the contour curve of V axle intercepting human body image, shown in Fig. 6 (a), the let others have a look at U of body section ring, the line segment that the V axle intercepts on image of white wire segment table among the figure.For normal vector a certain axle Aj (being U axle or V axle) on the cross section ring Si of ni, if its original length is Lj0, projected length on positive (or side) image is Lj ', the unit direction vector of Aj is dj, vector of unit length to the projecting direction of this image is vj, then its target length Lj1 is

$L_j^1=\left\{\begin{array}{cc}{L}_j^{\&prime;}/\sin (d_j^v_j),& d_j\&CenterDot;v_j\&NotEqual;0\\ L_j^0{L}_j^{\&prime;}/\max (d_j\&CenterDot;\stackrel{\&RightArrow;}{O_i{P}_k})& d_j\&CenterDot;v_j=0\end{array}\right.$

Wherein, O _iBe S _iThe center, P _k(0≤k＜N) is S _iOn point, (d _j^v _j) represent vectorial d _jAnd v _jBetween angle.L _j ¹-L _j ⁰Be the body templates model deformation when the individual character manikin, the cross section is encircled at A _jRequired displacement on the direction of principal axis.The displacement that cross section ring is gone up the arbitrfary point obtains by the linear interpolation of the displacement of cross section ring on U, V axle, with this deformed sections ring, thereby with the body templates model deformation shown in Fig. 6 (b) to the individual character manikin shown in Fig. 6 (c).

The projection of grid vertex on the individual character manikin on human body image is the texture coordinate of this grid vertex on respective image.

With the human body height value of mutual input and the ratio of the human body height in the direct picture is scale factor, and the individual character manikin that linear scale is initial obtains final personalized human geometry's model.Step 4:, obtain the texture coordinate of personalized manikin by the two-dimensional parameter manikin.According to four width of cloth images of input, on manikin two-dimensional parameter result, generate human body texture true to nature.

Before the human body curved surface is done two-dimensional parameterization, do necessary pre-service: the curved surface that the human body curved surface of sealing is cut into the plate-like topology.Shown in Fig. 7 (a), cutting path passes through the inboard of four limbs and the back side of trunk and head respectively, is similar to human skeleton on shape and structure.In Fig. 7 (a), cutting path is represented by the adjacent continuous line segment of black; White triangles shape is then represented the adjacent triangle of cutting path both sides.The concrete generation method of cutting path is as follows:

(1) for each the cross section ring on the four limbs, finds out the grid vertex on the U axle in its corresponding UVW frame.The adjacent mesh limit that connects these grid vertexes constitutes the cutting path on the four limbs.

(2) for the cross section of head and the following torso portion of armpit ring, to find out in its corresponding UVW frame-grid vertex on the V axle, the adjacent mesh limit that connects these grid vertexes constitutes the cutting path of these parts.

(3) with the Dijkdtra shortest path cutting path that obtains in (1), (2), (3) is connected, constitutes complete cutting path.

Revise the annexation between the triangle of cutting path both sides, make along cutting path with model be divided into go up for how much adjacent, non-conterminous curved surface on the network, thus the human body curved surface of sealing is cut into the curved surface that is equal to plate-like on the topology.The cutting result does not change leg-of-mutton number, has just duplicated the grid vertex on the cutting path, thereby has kept the corresponding relation between the triangle on the curved surface of cutting front and back.

Manikin parametrization after the cutting to two dimensional surface, shown in Fig. 7 (b), thereby is set up point on the 3 D human body curved surface and the one-to-one relationship between the point on the two-dimensional parameter result.The respective coordinates of point on the two-dimensional parameter result on the three-dimensional (3 D) manikin is the texture coordinate of this point.The three-dimension curved surface parametrization adopts the method for launching based on angle, by minimizing with minor function

$E\left(\mathrm{\&alpha;}\right)=\underset{t\&Element;T}{\mathrm{\&Sigma;}}\underset{k=1}{\overset{3}{\mathrm{\&Sigma;}}}\frac{1}{w_k^t}{({\mathrm{\&alpha;}}_k^t-{\mathrm{\&beta;}}_k^t)}^2$

And make it satisfy following constraint condition to obtain:

$\mathrm{\&pi;}-\underset{k=1}{\overset{3}{\mathrm{\&Sigma;}}}{\mathrm{\&alpha;}}_k^t=0,\&ForAll;t\&Element;T$

$2\mathrm{\&pi;}-\underset{(t,k)\&Element;v^{*}}{\mathrm{\&Sigma;}}{\mathrm{\&alpha;}}_k^t=0,\&ForAll;v\&Element;V_{\mathrm{int}}$

Wherein, t represents a triangle among the human body curved surface triangle set T, β _k ^tK interior angle value on the expression triangle t, α _k ^tExpression β _k ^tRespective value on the two-dimensional parameter result; $w_k^t=\frac{1}{{\mathrm{\&beta;}}_k^{\mathrm{<figure-callout\; id="2"\; label="t"\; filenames="A20081005788000171.png"\; state="\{\{state\}\}">t</figure-callout>}2}};$ V represents a summit on the grid surface, V _IntInternal vertex on the expression curved surface; v ^*Represent triangle the interior angle on v adjacent with v;

K 1 represents the triangle interior angle on the last k+1 of triangle t and k-1 the summit respectively.

With the two-dimensional parameter of three-dimensional (3 D) manikin rasterizing as a result, represent each point on the parametrization result with pixel coordinate.Concrete grammar is as follows:

(1) coordinate with the grid vertex on the parametrization result is converted into integer type from floating type.Ask for parametrization result's bounding box, with the length of this bounding box and the length and the width of width means human body texture; Each point with shaping coordinate representation in this bounding box is the pixel on the human body texture.

(2) according to the Bresenham Line generating algorithm, by every grid limit of adjacent pixels point expression.The Bresenham algorithm is the most frequently used Line generating algorithm of field of Computer Graphics, and it generates a line segment of being represented by adjacent image point by the pixel coordinates of two end points of given line segment.

(3) for each triangle on the parametrization result, from its center of gravity, fill by the zone, obtain to be positioned at all pixels of this triangle.Set up corresponding relation between pixel and the triangle, promptly which pixel each triangle comprises; Which triangle each pixel belongs to, and the barycentric coordinates on this triangle.Then for a triangle Δ ABC and inner some families thereof on the parametrization result, the family can be expressed as family [uA+vB+wC], wherein A, B, C are respectively the pixel coordinate on Δ ABC three summits, u, v, w are the barycentric coordinates of family on Δ ABC, and u+v+w=1, [x] expression rounds the coordinate figure of x.

As shown in Figure 8, on different direction of visual lines, detect blocking certainly between the three-dimensional personalized manikin limbs automatically.Method is as follows:

(1) intersection point of the V axle of the UVW frame that the cross section ring is corresponding with it has constituted maximal value and the minimum value of cross section ring on depth direction.Connect the side profile that these points constitute human body.Obtain the broadside lines that partes corporis humanis such as arm, leg and trunk divide with this.

(2) for a triangle on the three-dimensional (3 D) manikin, if the angle of its normal vector and direction of visual lines is greater than 90 degree, then on this direction of visual lines, this triangle is blocked, the VisualFlag=false (initial setting up VisualFlag=true) of this triangle of mark on this direction of visual lines.

(3) when direction of visual lines during over against the human body left surface, whether the projection of triangle on this direction of visual lines that detects VisualFlag=true on trunk and the leg is contained in or intersects at the projection of left hand arm broadside lines at this direction of visual lines, if then put this triangle and be blocked; Detect limbs blocking manikin on the different sight direction similarly.

For human body image, identified partes corporis humani's outline line before us, thereby the partes corporis humani that has been clear and definite also blocking between dividing.

By following method,, on the two-dimensional parameter result of personalized manikin, generate complete human body texture by the human body image of input:

For 1 P on the parametrization result, on three-dimensional (3 D) manikin and i direction of visual lines image, there are its corresponding point P ' and p ' respectively, i=0 wherein, 1,2,3 represent that respectively direction of visual lines is perpendicular to human body front, left side, right side and the back side.Give P texture color C=∑ λ ici/ ∑ λ i, ci represents the color value of p ', λ i=NpVi, Np is the normal vector of P ' on three-dimensional model, and the direction of visual lines when Vi takes for image I i is if P is blocked on the Vi direction of visual lines, then get λ i=0, if ∑ λ i=0, texture color value of putting of P not being ordered then, its result is shown in Fig. 9 (a).P corresponding point coordinate on the image I i of correspondence obtains by following method:

If there is the pixel P (promptly the point on the three-dimensional (3 D) manikin that this point is corresponding all is blocked) that is not set up the texture color value on the parametrization result under each direction of visual lines, then with the weighted mean value of its value of putting for texture value on the pixel of its 8 neighborhood value of putting, weighted value is inversely proportional to the distance of neighbor to P, and its result is shown in Fig. 9 (b).

Step 5: the corresponding relation on triangle according to model HO before the manikin cutting and cutting back model H1, for H ⁰On triangle t _i, get H ¹Go up and t _iCorresponding triangle t _iTexture coordinate on the ` is t _jThe texture coordinate on three summits arrives H with texture ⁰On, obtain the visual effect of the sense of reality.Figure 10 has shown after the texture of Fig. 9 (b) is to the personalized manikin of Fig. 6 (c), the result on the different sight direction.By our method, can generate realistic personalized manikin, can be advantageously applied to fields such as digital entertainment, virtual reality.

The above; only be the embodiment among the present invention; but protection scope of the present invention is not limited thereto; anyly be familiar with the people of this technology in the disclosed technical scope of the present invention; can understand conversion or the replacement expected; all should be encompassed in of the present invention comprising within the scope, therefore, protection scope of the present invention should be as the criterion with the protection domain of claims.

Claims (8)

1. the individualized realistic virtual character modeling method based on image is characterized in that, comprises the steps:

Step 1: structure is the basic geometric element of sweeping surface based on the manikin behaviour phantom template die type of sweeping surface with the planar cross-sectional ring, and each point on the manikin all belongs to corresponding cross section ring;

Step 2: in input picture, be partitioned into human body image, the human body characteristic information;

Step 3: obtain human geometry's information from image, distortion 3 D human body template model generates personalized human geometry's model;

Step 4: by two-dimensional parameterization, obtain the texture coordinate of personalized manikin, on manikin two-dimensional parameter result, generate human body texture true to nature by four width of cloth quadrature human body images to manikin;

Step 5: described texture seamlessly is mapped on the personalized manikin, obtains the visual effect of the sense of reality.

2. method according to claim 1 is characterized in that, the input picture in the described step 2 comprises front, side and the back side four width of cloth images.

3. method according to claim 1 is characterized in that, the characteristics of human body's information in the described step 2 comprises: human body feature point, human skeleton etc.

4. method according to claim 1 is characterized in that, the quadrature human body image in the described step 4 is four width of cloth quadrature human body images.

5. method according to claim 1 is characterized in that, described step 1 comprises:

Step 11: according to the physiological structure information of human body, extract human body feature point, generate human skeleton;

Step 12: the orthogonal frame UVW that arbitrfary point on the human skeleton is set;

Step 13: the sampling human skeleton, do the cross section according to the local frame on the sampled point to manikin, the cross section ring that obtains is the basic geometric element in the sweeping surface;

Step 14: the relative coordinate relation of setting up ring center, cross section and skeleton;

Step 15: sampling cross section ring generates the grid vertex of sweeping surface, and the trigonometric ratio grid vertex, obtains sweeping surface; Set up cross section ring and grid vertex subordinate relation; Human skeleton, cross section ring sweeping surface constitute the body templates model.

6. method according to claim 1 is characterized in that, described step 2 comprises:

Step 21: on input picture, alternatively be partitioned into human body image; According to human body physiological characteristics information, on the human body image profile, extract human joint points;

Step 22: connect adjacent articulation point, generate the human skeleton on the image.

7. method according to claim 1 is characterized in that, described step 3 comprises:

Step 31: set up body templates model and the corresponding relation of two-dimension human body image on human skeleton;

Step 32: set up the corresponding relation between body templates model upper section ring and the human body image;

Step 33: obtain the geometry deformation parameter of cross section ring, obtain the individual character manikin by the pair cross-section annular strain.

8. method according to claim 1 is characterized in that, described step 4 comprises:

Step 41: the personalized manikin that will seal automatically cuts into the curved surface that is equal to plate-like on the topology, and parametrization obtains the texture coordinate of personalized manikin to two dimensional surface;

Step 42: on different direction of visual lines, blocking certainly between automatic human body image and the three-dimensional (3 D) manikin upper limb body;

Step 43: by human body image, on the two-dimensional parameter result of manikin, generate complete human body texture, comprise processing to shield portions texture on the three-dimensional model.

Images