CN105427364A - Multi-point touch two-dimensional animation production method - Google Patents

Abstract

The invention relates to a multi-point touch two-dimensional animation production method, and relates to the computer graphics. The production method comprises the following steps: 1) handle assigning, wherein a user assigns editing handles of various types on an input image by adopting handle defining gestures; 2) animation production, wherein a triangular grid is constructed by taking the assigned handles as the constraint; the conversion weight of each handle is calculated; the user edits each handle according to the handle operating gestures; the edited position of each peak of the grid is calculated according to the conversion weight; and finally the image embedded into the grid is updated through a barycentric coordinate interpolation; 3) signal decomposition, wherein the animation represented by the dynamic triangular grid sequence is decomposed into three relatively independent signals of shape, posture and time; 4) signal editing, wherein the user independently edits the three relatively independent signals of shape, posture and time separately through well defined multi-point touch gestures; and 5) animation output, wherein the animation sequences are stored as GIF images in a call library for output. The multi-point touch two-dimensional animation production method is simple and visualized, and has powerful functions.

Description

A kind of method for making of multi-point touch 2 D animation

Technical field

The present invention relates to computer graphics, particularly relate to a kind of method for making of multi-point touch 2 D animation.

Background technology

Animation as a kind of comprehensive art merging numerous artistic class, from from it is born just with the form of expression of himself uniqueness and the using value woods independent of world's art.The high speed development of computer graphics and hardware technology makes computer animation start to move towards brilliant.The primo such as " Star War ", " Toy Story ", " Jurassic Park " allows people have one's quantum of the glamour of Computer Animated Graph.Computer animation has penetrated into each corner of people's life, is widely used in the various fields such as commercial advertisement, television film leader, film and television stunt, cartoon, game, education, military affairs.The manufacturing technology of computer animation has a lot, typical in key-frame animation, animation, human body animation and physically-based animation etc.Wherein, key-frame animation is a kind of technology the most classical.The concept source of key frame makes in traditional cartoon.The make-up room of WaltDisney in early days, the key picture (i.e. so-called key frame) during skilled animation teacher is responsible for designing cartoons, then designs intermediate frame by general animation teacher.In graphics field, scholars are devoted to study the various various technology generating two kinds of frames for convenient and high-quality: conveniently artist's design key frame, researchist proposes various deformation technology.Their basic thought comparatively simply acts on behalf of handle by editor exactly, and the transform effect acting on behalf of handle is delivered on whole complex object in a natural manner, thus the entirety editor realized object, key problem here how the conversion acting on behalf of handle to be delivered in a natural manner on whole object.Some researchists adopt the method based on grid, the people such as Igarashi ^[1]devise the analytical algorithm of a substep, first determine each leg-of-mutton suitable rotation on triangle gridding according to handle conversion, and then adjust their scaling further.Two steps all with the optimization being expressed as second energy, thus are converted into the Solve problems of a linear system by them, reach in real time.Sheffer and Kraevoy ^[2]and the people such as Sorkine ^[3]there is the coordinate of specific invariable attribute to catch the local shape of triangle gridding by introducing, in deformation process, keep these coordinates constant thus the local shape of triangle gridding is kept.Some researchists are also had to propose various non-mesh method, the Free Transform technology that Sederberg and Parry proposes ^[4]and the thin plate spline deformation technology that Bookstein proposes ^[5]be some early stage examples, follow-up researchist devises the generalized barycenter coordinate of various shape sensitive, such as HCCI combustion ^[6,7]and their various improved forms ^[8,9,10]; For intermediate frame, because in whole animation sequence, role is different, be all generally by carrying out automatic interpolation generation to two key frames.Selecting which type of geometric sense to carry out interpolation is key problem wherein.Fairly simple method carries out a linear interpolation to each corresponding point, but this mode easily causes the problems such as contraction.The people such as Xu ^[11]the deformation gradient of diabolo carries out linear interpolation, and rebuilds the apex coordinate after interpolation eventually through solving Poisson equation, but the method there will be flaw when speed ratio is larger.The people such as Winkler ^[12]the mode adopting the length of side of diabolo and dihedral angle to carry out non-linear interpolation solves the large problem rotated, but the method is very consuming time.Fr ¨ ohlich and Botsch ^[13]propose one approximation method fast.

Although the relative production technology of 2 D animation obtains discussing lastingly and widely in figure educational circles, but it is pointed out that existing research more concentrates on and how to improve in the effect of animation.The interactive interface of existing animation system still more complicated and abstract, especially amateurish fan is not suitable for: (1) is although as long as the progress of deformation technology allows artist just can realize the natural deformation to object entirety by some simple handles of editor, thus design high-quality key frame, but the degree of freedom of these handles remains higher, and some editing operation such as rotation etc. adopts mouse mode still to seem directly perceived not; (2) mode that existing animation system mostly provide only timeline allows the time course of artist to animation sequence control, realize F.F. to go out soon or slow-motion such as to go out slowly at the effect, such mode of operation seems particularly abstract, and the obvious limitation of the edit format of this one dimension is also very large.

List of references:

[1]TakeoIgarashi,TomerMoscovich,andJohnF.Hughes,“As-rigid-as-possibleshapemanipulation,”ACMTrans.Graph.,vol.24,no.3,pp.107–116,2005.

[2]A.ShefferandV.Kraevoy,“Pyramidcoordinatesformorphinganddeformation,”inProceedingsofthe3DDataProcessing,Visualization,andTransmission,2004,pp.68–75.

[3]OlgaSorkine,YaronLipman,andDanielCohen-Or,“LaplacianSurfaceEditing,”inProceedingsoftheEurographics/ACMSIGGRAPHSymposiumonGeometryProcessing2004,2004,pp.179–188.

[4]T.W.SederbergandS.R.Parry,“Free-formdeformationofsolidgeometricmodels,”inProceedingsofACMSIGGRAPH,1986,pp.151–160.

[5]F.L.Bookstein,“Principalwarps:Thin-platesplinesandthedecompositionofdeformations,”IEEETrans.PatternAnal.Mach.Intell.,pp.567–585,1989.

[6]M.S.Floater,“Meanvaluecoordinates,”Comput.AidedGeom.Des.,vol.20,no.1,pp.19–27,2003.

[7]TaoJu,ScottSchaefer,andJoeWarren,“Meanvaluecoordinatesforclosedtriangularmeshes,”ACMTrans.Graph.,vol.24,no.3,pp.561–566,2005.

[8]P.Joshi,M.Meyer,T.DeRose,B.Green,andT.Sanocki,“Harmoniccoordinatesforcharacterarticulation,”ACMTrans.Graph.,vol.26,no.3,p.ArticleNo.71,2007.

[9]YaronLipman,DavidLevin,andDanielCohen-Or,“GreenCoordinates,”ACMTrans.Graph.,vol.27,no.3,2008.

[10]Xian-YingLi,TaoJu,andShi-MinHu,“CubicMeanValueCoordinates,”ACMTrans.Graph.,vol.32,no.4,p.ArticleNo.126,2013.

[11]DongXu,HongxinZhang,QingWang,andH.Bao,“Poissonshapeinterpolation,”inProcofthe2005ACMSymposiumonsolidandphysicalmodeling,2005,pp.267–274.

[12]T.Winkler,J.Drieseberg,M.Alexa,andK.Hormann,“Multi-scalegeometryinterpolation,”Comput.Graph.Forum,vol.29,no.2,pp.309–318,2010.

[13]S. andM.Botsch,“Example-drivendeformationsbasedondiscreteshells,”Comput.Graph.Forum,vol.30,no.8,pp.2246–2257,2011.

Summary of the invention

The object of the present invention is to provide the method for making of simple, intuitive, powerful a kind of multi-point touch 2 D animation.

The present invention includes following steps:

1) handle is specified: user adopts handler definition gesture to specify various types of editor's handle over an input image;

2) animation producing: to specify handle for constraint structure triangle gridding, calculate each handle conversion weight.User handles each handle of gesture editing by handle, according to the position after each vertex edit of Transformation Weight re-computation grid, upgrades the image embedding grid finally by barycentric coordinates interpolation;

3) signal decomposition: the animation that dynamic triangle gridding sequence represents is resolved into shape, posture and time three relatively independent signals;

4) signal edit: user can be individually-edited to isolated shape, posture and time three relatively independent signals respectively by the multi-point touch gesture of good definition;

5) animation exports: call storehouse and animation sequence is stored as the output of GIF picture.

In step 1) in, described handler definition gesture can be the handler definition gesture of a set of multi-point touch, and the handle editor gesture of described a set of multi-point touch allows the how single or multiple handle of user to carry out the conversion such as translation, rotation simultaneously; Described editor's handle comprises the various forms handles such as a handle, line handle, region handle.

In step 3) in, described dynamic triangle gridding sequence is s represents the shape space of this warp mesh, and dimension is d=2V, V is grid vertex number, z _ifor the vector that the coordinate on all summits of this grid is formed, be decomposed into by sub-shape signal attitude signal and time signal the compact representation formed, by one group of function f, g, h are by these signal contacts together.

In step 4) in, describedly individually-editedly multi-point touch mode can be adopted intuitively to edit to isolated shape, posture and time three relatively independent signal, described editor directly perceived by naturally multi-point touch gesture directly perceived to support localization time domain editor to 2 D animation and resampling, shape territory editor and posture territory editor.

The present invention is directed to existing two-dimension cartoon making system interaction interface complicated, abstract, creation threshold is high, the problems such as creation efficiency is low, make full use of the multiple degrees of freedom input of multi-touch device, devise a kind of key frame production method, a kind of efficiently controlled middle frame generating method and the later stage compilation technology intuitively enriched easily naturally.The method can reduce the threshold of mobile phone cartoon content production greatly after industrialization, improves efficiency and the quality of mobile phone cartoon content production, promotes the flourish of industry.

The present invention utilizes the multiple degrees of freedom input mode of the multi-touch device of current popular, by multi-point touch gestures such as design nature, handle appointment intuitively, handle editor and signal edits, animation sequence is decomposed into independently shape, posture and time signal, provides a kind of simple, intuitive, powerful two-dimension cartoon making and edit methods.

Accompanying drawing explanation

Fig. 1 is overall flow figure of the present invention;

Fig. 2 generates animation scheme process flow diagram in Fig. 1;

Fig. 3 is signal decomposition protocol procedures figure in Fig. 1;

Fig. 4 is that pattern switches gesture figure;

Fig. 5 is handler definition gesture figure;

Fig. 6 is handle conversion gesture figure;

Fig. 7 is signal edit gesture figure.

Embodiment

See Fig. 1 ~ 7, a kind of multi-point touch control type two-dimension cartoon making and edit methods: first, user adopts the handler definition gesture of definition to generate the control handle of various form according to object to be manipulated; Then, user can carry out suitable map function by the handle conversion gesture of definition to these handles, and these effects operated is diffused on whole object in the natural mode of one, thus makes each key frame easily; According to these key frames, system can generate intermediate frame automatically, and user can pass through the time parameter of the meticulousr each intermediate frame of editor of multi-point touch gesture in this process, the animation sequence that final generation one is complete; System will be decomposed this animation sequence automatically, and form independently sub-shape signal, attitude signal and time signal, user can be edited freely by the mode of multi-point touch.Key of the invention process has 4 points: handle appointment, animation producing, signal decomposition, signal edit.Lower mask body introduction key realize details:

1. handle is specified

User can specify various editor's handle by the handler definition gesture (Fig. 4) of design, user can defining point, limit, lines and region handle.

2. animation producing

User can by the handle editor gesture (Fig. 5) of design to handle position, direction etc. effectively edit.An edited result affined transformation T of each handle _jrepresent, also need the affined transformation natural diffuseness of each handle on whole object.The conversion of each some p ∈ Ω on object is expressed as the weighted sum of the conversion of all relevant handles:

$T_p=\underset{j=1}{\overset{m}{\Σ}}{w}_j\left(p\right)T_j---\left(1\right)$

Wherein, for with each handle H _jrelevant weight function.This function is defined as the smooth function of the shape sensitive of a high-order:

${}^{\underset{w_j,j=1,...,m}{\arg \min }\underset{j=1}{\overset{m}{\Σ}}\frac{1}{2}{\∫}_{\mathrm{\Ω}}\left|\right|{\mathrm{\Δw}}_j|{|}^{2}dV---\left(2\right)}$

And meet:

$w_j{|}_{H_k}={\mathrm{\δ}}_{jk}---\left(3\right)$

W _j| _ffor linearly, $\∀F\∈F_C---\left(4\right)$

$\underset{j=1}{\overset{m}{\Σ}}{w}_j\left(p\right)=1,\∀p\∈\mathrm{\Ω}---\left(5\right)$

$0\≤w_j\left(p\right)\≤1,j=1,...,m,\∀p\∈\mathrm{\Ω}---\left(6\right)$

Wherein F _cfor all controlling polygon boundary edge, δ _jkfor Kroneckerdelta value.

By the linear finite of standard by this variational problem discretize:

$\begin{array}{c}{}^{\underset{j=1}{\overset{m}{\Σ}}\frac{1}{2}{\∫}_{\mathrm{\Ω}}\left|\right|{\mathrm{\Δw}}_j|{|}^{2}dV\≈\underset{j=1}{\overset{m}{\Σ}}\frac{1}{2}{\left(M^{-1}{\mathrm{Lw}}_j\right)}^{T}M\left(M^{-1}{\mathrm{Lw}}_j\right)}\\ =\frac{1}{2}\underset{j=1}{\overset{m}{\Σ}}{w}_j^T\left({\mathrm{LM}}^{-1}L\right)w_j.\end{array}---\left(7\right)$

Finally, QUADRATIC PROGRAMMING METHOD FOR is adopted to carry out solving the weight distribution function that can obtain each handle.

For intermediate frame, obtain by carrying out non-linear interpolation to key frame at the whole story.For each triangle of beginning key frame triangle gridding and at the corresponding triangle of last key frame the mapping A between them can be constructed _{{ i, j, k}}, make Q _{{ i, j, k}}=A _{{ i, j, k}}p _{{ i, j, k}}, and for the corresponding triangle of intermediate frame there is V (t) _{{ i, j, k}}=A (t) P _{{ i, j, k}}.By A _{{ i, j, k}}be decomposed into the combination of a rotation matrix and a symmetrical matrix:

$A_{\{i,j,k\}}=R_{\mathrm{\α}}{\mathrm{DR}}_{\mathrm{\β}}=R_{\mathrm{\α}}\left(R_{\mathrm{\β}}{R}_{\mathrm{\β}}^T\right){\mathrm{DR}}_{\mathrm{\β}}=\left(R_{\mathrm{\α}}{R}_{\mathrm{\β}}\right)\left(R_{\mathrm{\β}}^T{\mathrm{DR}}_{\mathrm{\β}}\right)=R_{\mathrm{\γ}}S---\left(8\right)$

Based on this decomposition, interpolation can be carried out to the free parameter of this decomposition and obtain generating intermediate frame V (t) _{{ i, j, k}}conversion A (t):

A _α,β(t)＝R _tα((1-t)I+tD)R _tβ，(9)

A _γ(t)＝R _tγ((1-t)I+tS)。(10)

Like this, according to different time t, transformation matrix A (t) is applied to source grid, can intermediate frame V (t) be obtained _{{ i, j, k}}grid.

3. signal decomposition

In order to better, the animation that abovementioned steps generates is edited, will the dynamic grid sequence of this animation process be described (S represents the shape space of this warp mesh, and dimension is d=2V, V is grid vertex number, z _ivector for the coordinate on all summits of this grid is formed) be decomposed into by sub-shape signal attitude signal and time signal form compact representation, by one group of function f, g, h by these signal contacts together:

These functions jointly define a continuous print and map Ψ=h ο g ο f.(12)

In order to make full use of the shape redundancy existed in dynamic grid sequence, first to S dimensionality reduction, with a linearity configuration subspace replace, each frame z of dynamic grid sequence _iby a vector x _i∈ R ^mrepresent, this vector is that this frame is relative to base vector coordinate, finally, can defined function h be:

H is the matrix of d × m, and each row is by base vector h _jform.Definition by solving following optimization problem to determine parameter in h

$\arg min\underset{i=1}{\overset{n}{\Σ}}\left|\right|z_i-h\left(x_i\right)|{|}^2.---\left(14\right)$

Meet the minimum H of this energy for skew frame order be the best approximation of m.Skew frame vector is carried out svd to this matrix, the best approximation of the order that namely singular vector corresponding with individual maximum singular value obtained will be found as row formation matrix m.

Similarly, by constructing an attitude redundancy eliminated from two-dimentional posture subspace to the mapping g of shape subspace animation sequence.First by minimizing following energy to determine the two-dimensional coordinate of each frame at configuration space

$\underset{i=1}{\overset{n-1}{\Σ}}\underset{j=i+1}{\overset{n}{\Σ}}{\left(\right||p_i-p_j||-||x_i-x_j|\left|\right)}^2.---\left(15\right)$

Then, the problem of approaching by solving a discrete data constructs this Nonlinear Mapping g.For given one radial basis function φ: R ₊one group of basic point in → R and plane then g may be defined as:

Make Radial basis kernel function φ (r)=r ²logr, offset vector g in formula ₀with matrix G=(g _l+1, g _l+2) ∈ R ^{m × 2}ensure that this radial basis approaches and has linear precision, also ensure the uniqueness of this expression simultaneously.Adopt a kind of greedy algorithm from configuration space coordinate middle l the basic point selecting optimum make x _i-g (p _i) maximum:

(1) point set is calculated optimal fitting plane.

(2) order set represent the index value to four maximum summits of this plan range;

(3) l is made to be increased to n-1 from 4

A) with l basic point calculate g,

B) σ is made _l+1under current approaching, there is maximum error x corresponding to non-basic point _i-g (p _i) the index value of point, that is:

${\mathrm{\σ}}_{l+1}=\underset{i\∉{\left\{{\mathrm{\σ}}_j\right\}}_1^l}{\mathrm{argmax}}(x_i-g(p_i\left)\right)---\left(17\right)$

Finally, adopt cubic spline curve to the coordinate points p corresponding to each frame in two-dimensional attitude space _icarry out matching, obtain a continuous print and approach to extract temporal redundancy:

In formula for B-spline function, f _jfor the control vertex of correspondence.This k cubic B-spline function is defined in and comprises k+2 non-decreasing node knot vector u on, and p _ican by calculating n parameter worth go out.In order to obtain optimum approaching, have employed (the distance t between neighboring parameter values of non-uniform spacing _i+1-t _i=|| p _i+1-p _i|| ^α), but this n frame is but equally distributed in time, for this reason, to function carry out Reparameterization.Introduce a monotonic quantity make ρ (i)=t _i.Order final:

4. signal edit

After completing the signal decomposition to animation sequence, just very intuitively can edit three signals respectively, thus effectively edit whole animation sequence.

(1) localize time domain editor

Owing to have employed the B-spline curves of local support to represent time-domain signal f, this allows can edit time-domain curve by the control vertex editing B-spline curves easily, and these editors will only have influence on some adjacent frames.In addition, due to postural cue flatness guarantee that such editor can not cause the jump of grid sequence to change.Multiple control vertex can be edited by the mode of multi-point touch simultaneously.

(2) resampling

New postural position can also be obtained by carrying out resampling to time-domain curve, and then obtain one group of new Deformation Series.

(3) shape territory editor

By the editor to shape territory, animation process can be moved to another grid sequence from a grid sequence.

(4) posture territory editor

A dynamic grid sequence can be generated with growing out of nothing by the editor in posture territory.

The invention discloses a kind of multi-point touch control type two-dimension cartoon making and edit methods, comprise the following steps: 1) handle is specified: user adopts handler definition gesture to specify various types of handle over an input image; 2) animation producing: to specify handle for constraint structure triangle gridding, calculate each handle conversion weight.User handles each handle of gesture editing by handle, according to the position after each vertex edit of Transformation Weight re-computation grid, upgrades the image embedding grid finally by barycentric coordinates interpolation; 3) signal decomposition: the animation that dynamic triangle gridding sequence represents is resolved into shape, posture and time three relatively independent signals; 4) signal edit: user can be individually-edited to isolated several signal respectively by the multi-point touch gesture of good definition; 5) animation exports: call storehouse and animation sequence is stored as the output of GIF picture.

The present invention is directed to existing two-dimension cartoon making system interaction interface complicated, abstract, creation threshold is high, the problems such as creation efficiency is low, make full use of the multiple degrees of freedom input of multi-touch device, devise a kind of key frame production method, a kind of efficiently controlled middle frame generating method and the later stage compilation technology intuitively enriched easily naturally.The method can reduce the threshold of mobile phone cartoon content production greatly after industrialization, improves efficiency and the quality of mobile phone cartoon content production, promotes the flourish of industry.

Claims (4)

1. a method for making for multi-point touch 2 D animation, is characterized in that comprising the following steps:

1) handle is specified: user adopts handler definition gesture to specify various types of editor's handle over an input image;

2) animation producing: to specify handle for constraint structure triangle gridding, calculate each handle conversion weight.User handles each handle of gesture editing by handle, according to the position after each vertex edit of Transformation Weight re-computation grid, upgrades the image embedding grid finally by barycentric coordinates interpolation;

3) signal decomposition: the animation that dynamic triangle gridding sequence represents is resolved into shape, posture and time three relatively independent signals;

4) signal edit: user is individually-edited to isolated shape, posture and time three relatively independent signals respectively by the multi-point touch gesture of good definition;

5) animation exports: call storehouse and animation sequence is stored as the output of GIF picture.

2. the method for making of a kind of multi-point touch 2 D animation as claimed in claim 1, it is characterized in that in step 1) in, described handler definition gesture is the handler definition gesture of a set of multi-point touch, and the handle editor gesture of described a set of multi-point touch allows the how single or multiple handle of user to carry out translation, rotational transform simultaneously; Described editor's handle comprises a handle, line handle, region handle various forms handle.

3. the method for making of a kind of multi-point touch 2 D animation as claimed in claim 1, is characterized in that in step 3) in, described dynamic triangle gridding sequence is s represents the shape space of this warp mesh, and dimension is d=2V, V is grid vertex number, z _ifor the vector that the coordinate on all summits of this grid is formed, be decomposed into by sub-shape signal attitude signal and time signal the compact representation formed, by one group of function f, g, h are by these signal contacts together.

4. the method for making of a kind of multi-point touch 2 D animation as claimed in claim 1, it is characterized in that in step 4) in, described individually-edited be adopt multi-point touch mode intuitively to edit to isolated shape, posture and time three relatively independent signal, described editor directly perceived by naturally multi-point touch gesture directly perceived to support localization time domain editor to 2 D animation and resampling, shape territory editor and posture territory editor.