Viseme aware realistic 3D face modeling from range images

Size: px
Start display at page:

Download "Viseme aware realistic 3D face modeling from range images"

Transcription

1 IJCSNS Internatonal Journal of Computer Scence and Network Securty, VOL.9 No.4, Aprl Vseme aware realstc 3D face modelng from range mages Ken Yano and Koch Harada, Graduate School of Engneerng Hroshma Unversty, Hroshma, Japan Summary In ths paper, we propose an example based realstc face modelng method wth vseme control. he vseme descrbes the partcular facal and oral postons and movements that occur alongsde the vocng of phonemes. In facal anmaton such as speech anmaton and talkng head, etc, a face model wth open mouth s often used and the model s anmated along wth the speech sound by synchronzng the change of mouth shape to that of the phoneme. We clam that vsemes as well as facal expressons are dosyncratc by nature, so they have to be modeled specfc to the subect to generate a truthful facal anmaton for the subect. he proposed method tres to automate the creaton of a realstc face model wth vseme control from a set of scanned data. Key words: Face modelng, surface reconstructon, morphng. Introducton Creatng face models that look and move realstcally s an mportant problem n computer graphcs. It s also one of the most dffcult ones, as even the mnute changes n facal expresson can reveal complex moods and emotons. he presence of very convncng characters n recent flms makes a strong case that these dffcultes can be overcome wth the ad of hghly sklled anmators. Because of the sheer amount of work requred to create such models, there s a clear need for more automated technques. A vseme s a supposed basc unt of speech n the vsual doman. It descrbes the partcular facal and oral postons and movements that occur alongsde the vocng of phonemes. We propose an example based realstc face modelng technque especally to create a morphable face model wth expressve vseme control. he nput to our system s a set of range and color mages of the specfc subect scanned by the scanner []. Note that at each pxel coordnate of the range mage contans a measured 3D coordnate and each pxel of the color mage contans RGB color ntensty. A par of the two mages s scanned smultaneously and has natural pxel-to-pxel" correspondence between them. A statc 3D face model s generated n two folds. Frst a general 2D facal template mesh s ftted to the mages of the subect usng RBF (Radal Bass Functon) method. he fttng s processed n 2D space, the mage plane, usng the constrants set by the user. he constrant s a set of fducal marker ponts, such as the tp of the nose and corners of the eyes, etc. After the RBF fttng, a 3D coordnate s sampled at each vertex of the deformed mesh from the range mage as well as a texture coordnate from the color mage. he connectvty of the sampled ponts s naturally gven by the template mesh. he face models generated offer "pont-to-pont" correspondence among them rrespectve of the knd of the vseme of the subect. hs feature makes t possble to statstcally analyze the geometrc changes by vseme. In facal modelng, sculptng expressve open mouths requres dauntng task even for a sklled modeler as well as expressve eyes. oreover, artculatng an open mouth n a realstc way s one of the dffcultes that an anmator encounters for facal anmaton. In order to estmate the geometrc changes of a face model due to vseme, especally the shape of a open mouth, we frst prepare the 3D face models for a set of vsemes as well as the base model (closed mouth poston). Gven a set of 3D face models for each vseme and for the closed mouth, the geometrc changes due to each vseme are estmated by subtractng the base model from the face model of the vseme. In order to acqure relable moton vectors, the geometrc changes, we create a stencl mask to remove undesred moton vectors for the upper part of the face. Gven the geometrc changes for each vseme, we construct a morphable face model wth expressve vseme control. wo types of morphng are expermented. he frst s the blend shapes n whch a novel face model s create as a convex lnear combnaton of $n$ bass vectors, each vector beng one of the blendshapes, e.g., the face tself. In the second method, we apply PCA (Prncpal Component Analyss) to the face space. PCA s a statstcal tool that transforms the space such that the covarance matrx s dagonal (.e., t decorrelates the data). anuscrpt receved Aprl 5, 2009 anuscrpt revsed Aprl 20, 2009

2 246 IJCSNS Internatonal Journal of Computer Scence and Network Securty, VOL.9 No.4, Aprl 2009 he morphable model by PCA provdes an another alternatve to generate a novel face model wth expressve vseme, however n most cases the bass vectors of the decomposed face space do not have clear semantcs, so t requres the user some tmes to fgure out what effects each bass vector mght have for the deformed face model. he man contrbuton of ths paper s to create a realstc morphable face model wth vseme control from a set of scanned data. he proposed method does not requre specal modelng or anmaton sklls and proceeds wth as lttle as user nterventons. ost of the relevant works pay lttle attenton or no attenton about the change of facal geometry due to the open mouth postons or the vsemes. he results ensure that our method provdes a novel way to generate a realstc face model wth expressve vseme control and the generated models can be used for further processng for downstream applcatons. 2. Related Works Correspondent face models are often used n graphcs and vson communty. Blanz et al [0] create a morphable face model by takng range mages of several faces usng a 3D scanner and puttng them nto "one-to-one" correspondence by expressng each shape usng the same mesh structure. Usng the morphable model, t s possble to group changes n vertex poston together for representng common changes n shape among several surfaces. Usng prncpal component analyss, they succeeded to fnd a bass for expressng shape changes between faces. odel wth full correspondence was also studed by Praun et al [9]. In order to establsh full correspondences between models, they create a base doman that s shared between models, and then apply a consstent parameterzaton. hey search for topologcally equvalent patch boundares to create base doman mesh. Wang et al [8] propose a novel approach for facal expresson analyss from mages. hey decompose facal expresson by Hgher-Order Sngular Value Decomposton (HOSVD), a natural generalzaton of matrx SVD and learn the expresson subspace. Several papers propose the method to learn facal expresson from vdeo [7] []. hose works typcally nvolves vsual trackng of: facal movement such as contours, optcal flow, or transent features such as furrows and wrnkles. Facal expressons are learned and recognzed n relaton to Facal Acton Codng System (FACS) [6], a popular representaton that codes vsually dstngushable facal movement n small unts. hese acton unts descrbe qualtatve measures such as "pullng lp corners", or "furrow deepenng". A facal expresson s descrbed as a combnaton of these unts. Vlasc et al [5] proposes a method for expressng changes n face shape usng a mult-lnear model, accountng for shape changes not only based on the subect's dentty but also based on varous expressons and vsemes. Yong et al [2] presents a novel approach to produce facal expresson anmaton for a new model. Instead of creatng new facal anmaton from scratch for each new model created, they take advantage of exstng anmaton data n the form of vertex moton vector. hey call ths process "expresson clonng" and t provdes an alternatve way for creatng facal anmaton for character models. 3. Realstc 3D face modelng he technques of realstc face modelng and anmaton have been spurred by emergence of 3D scannng devces at relatvely low cost. Snce range data contans artfacts such as spkes and holes, t needs to be repared for further processng. Furthermore only facal part of whole scanned data s relevant, unnecessary porton of the data has to be removed wth great care and sklls. he proposed modelng method tres to automate such lengthy operatons and produce a realstc face model for specfc subect. he detals of our method are descrbed as follows. 3. ethod - 2D emplate fttng he nput to our system s range and color mages of the subect scanned by range scanner []. he two mages are taken smultaneously by one scan so there s a natural correspondence between each par of the pxels of the two mages. At each pxel locaton of the range data contans a measured 3D coordnate (x,y,z) from the vew of the scanner. o dsplay range mages, the depth nformaton (z coordnate) of each pxel s converted to gray scale ntensty. Fgure : From left to rght; (a) A 2D template mesh s shown wth facal feature ponts. he template mesh contans 44 vertces and 804 faces. (b) he user s asked to select feature ponts nteractvely on the color mage. he selected feature ponts are shown n red. Each selected feature pont corresponds to the feature pont defned on the 2D template mesh.

3 IJCSNS Internatonal Journal of Computer Scence and Network Securty, VOL.9 No.4, Aprl Frst, a 2D template mesh (wth zero depth) shown n Fgure s prepared on whch a set of fducal feature ponts ( FP ) s defned. wenty feature ponts n total are emprcally defned. he user s asked to select facal feature ponts ( F P ) nteractvely on color mage (Fgure ). As a result, there s a one-to-one mappng, Φ, from the ponts set ( FP ) to the pont set ( F P ). After establshng the mappng, Φ, the template mesh s deformed and ftted to the face mage usng the mappng as the constrants. he fttng s done usng RBF (Radal Bass Functon) technque, a well known scattered data nterpolaton technque, or some authors call t PS (hn Plate Splne) nterpolaton. he detal of the algorthm as follows. Gven a par of pont patterns wth known correspondences U = ( u u,..., ) and ( ), 2 V = v, v2,..., v m, where U ( FP), V ( FP ), and m = 20 n our case, we need to establsh correspondences between other ponts on the template mesh to the pxel locaton on the facal mage. he warpng functon, F, whch warps the pont set U to the pont set V subect to perfect algnment, s gven by the condtons F ( u ) = v for =,2,..., m. he nterpolaton deformaton model s gven n terms of the warpng functon F (u), wth m F( u) = A( u) + = λ φ( u u ), u R 2,where φ ( u ) = r log( r) s a radcally symmetrc bass functon, whch s known as thn plane splne functon, λ s a real-valued weght, and u u s Eucldean dstance. he functon, A (u), s a degree one polynomal that accounts for the lnear and constant porton of F (u) whch s defned as A ( u) = a0 + a* u x + a2* u y for 2D nterpolaton. Solvng for the weghts λ and the coeffcents of A (u) subect to the gven constrants yelds a functon that both 2 u m nterpolates the constrants and mnmzes the Equaton above n least-square sense. In order to solve for the unknown weghts λ and the coeffcents of A (u), the constrants become v m = A( u ) + = λ φ( u u Snce the equaton s lnear wth respect to the unknowns, λ and the coeffcents of A (u), t can be formulated as a lnear system. Fgure 2: Fttng the 2D template mesh to the mage. From left to rght; (a) range mage (b) deformed 2D template mesh (low resoluton) (c) deformed 2D template mesh (hgh resoluton) he resultng functon F (u) exactly nterpolates the gven facal features ( F P ) and s not subect to approxmaton or dscretzaton errors. Note that the number of weghts to be determned does not grow wth the number of vertces n the template mesh; rather t s only dependent on the number of constrants. he 2D template mesh s deformed by the functon, F (u), and t s shown n Fgure 2. As the fgure ndcates, the template mesh s ncely deformed and ftted to the facal area on the mage wth all the feature ponts exactly nterpolated. In order to generate a 3D face model, 3D coordnate at each vertex poston of the deformed 2D template mesh s calculated from the range mage. As mentoned before, at each pxel coordnate of the range mage contans measured 3D coordnate of the subect, then the 3D coordnate s calculated as a blnear nterpolaton of the 3D coordnates at the four neghborng pxel coordnate of the vertex poston. Note that each vertex of the deformed mesh should not necessarly resde on the grd pxel coordnate of the mage; we use blnear nterpolaton to correctly calculate the 3D coordnates. One of the well known ssues pertanng to 3D scanned data s that there are holes or mssng data. For varous )

4 248 IJCSNS Internatonal Journal of Computer Scence and Network Securty, VOL.9 No.4, Aprl 2009 reasons reflected laser beam may be obscured or dspersed so that the sensors see no range or color data at some surface ponts. Especally for frontal face, those mssng data commonly occur on the undersde of the aw, the head har, the eyebrows, open mouth, and pupls of the eyes, etc. here are some technques to fll n those mssng data. For example [3] ntroduces the followng algorthm [Step 0] he range data s converted to a floatng pont array wth the mssng values set to 0.0 [Step ] A second matte array s created and set to.0 where vald range data exst and to 0.0 where gaps exst. [Step 2] In the regons around 0.0 matte value, a small blurrng flter kernel s appled to both the matte data and the surface data [Step 3] Where the blurred matte value ncreases above a threshold value, the surface sample s replaced by the blurred surface value dvded by blurred matte value. [Step 4] he flterng n recursvely appled untl no matte values are below the threshold We have ntegrated the above algorthm to our system and appled to the range mage before samplng the 3D coordnates. Note that ths algorthm only flls n the small gaps and cannot cope wth large mssng area such as the head har and the background. he resoluton of the ntal 2D template mesh s consdered to be low, so as the resoluton of the generated 3D face model. In order to ncrease the resoluton, we apply the one-to-four subdvson method to the deformed 2D template mesh before samplng 3D coordnates. he subdvson method subdvdes each trangle face nto four trangle faces, thus ncreases the resoluton by the factor of four. Fgure 2 shows the deformed template mesh before and after the subdvson. Snce the color mage and the range mage have "one-toone" pxel correspondence between them, as we re-sample 3D coordnates from the range mage, we can also obtan texture coordnate at each vertex of the 3D face model. Fgure 3 shows the generated 3D face model n varous renderng modes for a subect. Fgure 3: Generated 3D face model. From left to rght; (a) shadng model (b) texture model (c) wre frame model 4. odelng of vseme odelng a realstc face wth open mouth s a challengng task. In order to create a fully functonal face model, t s common that face, teeth and tongue, etc are modeled separately and are placed later at approprate postons of the entre face model. Snce we only concern about the dynamcs of facal geometry due to vseme, we do not dscuss the ssues of teeth, tongue, etc. he dynamcs of mouth shape s dosyncratc as well as the facal expresson, so we clam that vsemes should dffer from person to person. Our modelng method of vsemes s an example based and the changes of the geometry of the mouth are learned and modeled from a set of scanned data taken from the same subect from whom we model the base (closed mouth) face model n Secton 3. he detals of the method are descrbed as follows. 4. ethod Obtan a scanned data for each vseme from the subect. he subect s asked to make natural mouth shape n fve postons one by one by pronouncng these fve words then haltng the mouth poston at the underlned two characters (. car. man. she v. eel v. too) Step: Generate a 3D face model for each vseme usng the same technque n Secton 3. However we use a dfferent set of feature ponts to handle the deformaton of the open mouth. he new set of the feature ponts and the result of the fttng s shown n Fgure 4.

5 IJCSNS Internatonal Journal of Computer Scence and Network Securty, VOL.9 No.4, Aprl Fgure 4: Fttng the 2D template mesh to the mage of open mouth. From left to rght; (a) color mage wth markers. A new set of markers are defned to model open mouth (b) deformed 2D template mesh (low resoluton) (c) deformed 2D template mesh ( hgh resoluton) Step2: Estmate 3D moton vectors from the base model (closed mouth) to the specfc model of the vseme. Snce the two models are generated usng the same template mesh, there s a natural correspondence between each par of the vertces. In order to obtan the 3D moton vectors, we frst algn the two model (base model and the vseme specfc model) usng ICP (Iteratve Closest Pont) method [4]. Step3: he ntal correspondences of the ponts of the two models are gven by the user. As shown n Fgure 5, sx pars of ponts are selected for each model. hese ponts are selected because they are not lkely to be transformed due to the deformaton of the mouth shapes and sx ponts are enough to calculate the rotaton and the translaton of the rgd transformaton between the two models. Fgure 6 shows the algnment of the two models before and after ICP method s appled. Fgure 5: Pont correspondence between the two model for ICP he 3D moton vectors can be obtaned from the algned models by subtractng the base model from the specfc model of each vseme. However, we do not want undesred effects of the moton vectors over the upper face due to the changes of the mouth shape; we suppress them by makng a stencl mask over the face. Fgure 6: Rgd algnment of the two model by ICP. Frst row: ntal pose of the two models, second row: algned poses after ICP s appled he stencl mask s created n these steps.. Generate seeds ponts for the stencl mask. he seeds ponts are generated by tracng the shortest path between selected two ponts. he seeds ponts are all the vertces along the shortest path generated. In order to make a vald mask, a set of fve shortest paths s generated usng Dkstra algorthm. he start and the goal pont of each shortest path are defned emprcally. hese steps are llustrated n Fgure After obtanng the seed ponts, we make vrtual spheres centered at each seed pont then assgn every vertces whch s nsde of one of the spheres to the stencl mask. he radus of the sphere s decded as 20mm for the subect. Fgure 7 (d) shows the generated fnal stencl mask (vertces n orange color). 3. he 3D moton vectors under the stencl mask are suppressed to zero vectors. Compared wth the base model, the vseme specfc model can reveal nosy measured data n the open mouth area. hs s because at the area around the open mouth, the reflected laser beam may be obscured n the hollow area between the lps or dspersed from the teeth. Fgure 8 shows unprocessed mesh models from range data and the mesh models generated by our method.

6 250 IJCSNS Internatonal Journal of Computer Scence and Network Securty, VOL.9 No.4, Aprl 2009 to get the face model. Fgure 0 shows a set of generated face models for each vseme. he procedure s appled to two dfferent subects and the results are shown n Fgure 7 Fgure 7: ask generaton to extract 3D moton vector. From left to rght, (a) seed ponts from the frst path (b) seed ponts from the second and the thrd paths (c) seed ponts from the fourth and ffth paths (d) generated mask. he mask s the vertces n orange. he uncovered vertces are colored n black. In order to smooth the nosy 3D moton vectors, we use Laplacan smoothng, e.g., a sgnal processng approach proposed by [2]. Laplacan smoothng removes hgh frequency nose. he formula s gven as Fgure 9: he result of deformaton. arget model (blue), deformed model (green) d v ) / t = wδd( v ) ( Δd ( v ) = N( v ) v N ( v ) ( d( v ) d( v )),where d ( v ) s a moton vector at the vertex v, N ( v ) s the -rng neghbors of the vertex v and w s a scalng factor 0 w. We set w = 0. and the smoothng operaton s terated thrty tmes to get the fnal 3D moton vectors. Fgure 0: he result of deformaton for all targets. From left to rght (a) vseme( car ) (b) vseme( she ) (c) vseme( man ) (d) vseme( eel ) (e) vseme( too) 5. orphable face model In ths secton example-based morphable 3D face models are created. wo methods are expermented. he frst method represents the morphable model as a blendng of example models. A lnear combnaton of the examples models of specfc vseme descrbes a realstc change of vsemes. he second method apples PCA to the face space. PCA defnes a bass transformaton from the set of example models to a bass that has two propertes relevant to our morphable face model. Frst, the bass vectors are ordered accordng to the varances n the dataset along each vector. Second the bass vectors are orthogonal to each other. he detals of two methods are descrbed as follows. Fgure 8: Nosy measured data around the open mouth area. Frst row shows the unprocessed 3D model obtaned from the range data. Second row shows the generated 3D model usng the template fttng. hose undesred nosy samplng data around the mouth s the result of nosy scannng data. 4.2 Generated models Fgure 9 shows generated face model of specfc vseme. he moton vectors obtaned are appled to the base model 5. orphng by Blend-Shapes In order to create a morphable face model, a set of the vseme models generated n Secton 4 s algned to the base model usng ICP method. We let S0 be the shape vector of the base model and S be the shape vectors of the model of the vseme, the morphable face model s represented as

7 IJCSNS Internatonal Journal of Computer Scence and Network Securty, VOL.9 No.4, Aprl S = w S + (.0 w ) S () = = 0, where w are non-negatve blendng weghts and (= 6) s the number of vseme poses. he shape vectors S are formed from the 3D coordnates of surface ponts: S =..., x, y, z ) ( x, y, z, n n Fgure shows the gradual deformaton from the base model to the models of three specfc vsemes. Each deformed model s created by changng the weght relevant to the specfc vseme and other weghts are set to zero. n Fgure 2: Random blendng of varous vsemes. Left: 0.6 * vseme(man) * vseme(eel) + 0. * base, Center: 0.5 * vseme(too) * vseme(car) * base, Rght: 0.8 * vseme(too) + 0. * vseme(eel) + 0. * vseme(car) 5.2 orphng by PCA Let a set of algned 3D face model represented as a matrx X, where [ S S S ] X = S S of each vseme be,and X s of dmenson 3 n, where n s the number of vertces. he dfference from the average face model (the sample mean) S s the matrx : [( S S )( S S )( S S )..( S S )] X = 0 = S S S S Prncpal component analyss seeks a set of orthogonal vectors, e, whch best descrbes the dstrbuton of the nput data n least-square sense,.e., the Eucldean proecton error s mnmzed. he typcal method of computng the prncpal components s to fnd the egenvectors of the covarance matrxc, where 2 Fgure : Gradual deformaton of each vseme. he base model s gradually deformed to the target vseme model by changng the weght of the target (w=0.3: left column, w=0.6: center column, and w=0.9: rght column). arget vseme model; frst row (vseme(car)), second row (vseme(eel)), thrd row (vseme(too)) he effect of blendng of multple vseme examples s shown n Fgure 2. C = = 0 S S = X X s 3 n 3n. hs wll normally be a huge matrx, and a full egenvector calculaton s mpractcal. Fortunately, there are only nonzero egenvalues, and they can be computed effcently wth an egenvector calculaton. It s easy to show the relatonshp between the two. he egenvectors e and egenvalues λ of C are such that Ce = λ e

8 252 IJCSNS Internatonal Journal of Computer Scence and Network Securty, VOL.9 No.4, Aprl 2009 hese are related to the egenvectors ê and egenvalues μ of the matrx D = X X n the followng way: Deˆ = μ eˆ, X CX eˆ = μ X eˆ, X eˆ = μ eˆ, X X C( X eˆ ) = μ ( X eˆ ), X eˆ = μ X eˆ Ce = λ e, showng that the egenvectors and egenvalues of $C$ can be computed as e = ( X ˆ ), e λ = μ In other words, the egenvectors of the large matrx C are equal to the egenvectors of the much smaller matrx D, premultpled by the matrx X. he nonzero egenvalues of C are equal to the egenvalues of D. Once the egenvectors of C are found, they are sorted accordng to ther correspondng egenvalues; a larger egenvalue means that more of the varance n the data s captured by the egenvector. Egenvectors correspondng to the top K evenvalues defne the face space of the vsemes. he result of PCA s descrbed as follows. Fgure 3 shows the average model, S, and Fgure 4 shows top fve prncpal egen face models. Fgure 3: Average face model he CR (Contrbuton Rate) and CCR (Cumulatve Contrbuton Rate) of a egen face model, e l, are defned as CR λ l = =, CCR = = λ l = λ λ able shows the egenvalue, CR and CCR of top fve prncpal egen models. From these results, the CCR of the top three prncpal models s over 0.95 and the last two prncpal models are less mportant. From ths analyss, we can conclude that only the top three prncpal models may be used to defne the face space thus compressng the entre model space. able : he result of PCA analyss Egen value CR CCR Frst egen model Second egen model hrd egen model Fourth egen model Ffth egen model able 2: PCA coeffcents of each vseme model ω ω 2 ω ω 3 4 ω 5 Base Vseme(too) Vseme(man) Vseme(she) Vseme(eel) Vseme(car) In order to proect each face model to ths egen face space, we frst subtract the average face model from the face model then proect t nto the egen face space as followng ( w w... w ) = ( e e2... e ) ( S ) 2 k k S, where w s the PCA coeffcent of the face model for Fgure 4: A set of prncpal egen face models. From left to rght; (a) frst egen model (b) second egen model (c) thrd egen model (d) fourth egen model (e) ffth egen model (he egen models are actually moton vectors, here we add them to the mean face model for dsplay) the th prncpal face model e. able 2 shows the PCA coeffcents of each face model and the results are plotted n Fgure 5 graphcally. From ths result, we can assume that the frst prncpal face model descrbes natural open mouth by aw rotaton, the second prncpal model descrbes puckered mouth and the thrd prncpal model descrbes mouth pulled out sdeways.

9 IJCSNS Internatonal Journal of Computer Scence and Network Securty, VOL.9 No.4, Aprl able 3: orphable model by PCA. he values are the blendng parameters, α, n Equaton 2. α α 2 α 3 Blend model Blend model Blend model Fgure 5: PCA Coeffcents of each vseme model. Graphcal dsplay of able 3 hanks to the face space decomposed by PCA, t gves the user an alternatve method to anmate the face model or to blend the prncpal face models. For nstance, exaggerated open puckered mouth can be ntutvely generated by usng the frst and second prncpal face model, whch s otherwse dffcult to be made by usng the method descrbed n Subsecton 5.. Compared wth the Equaton, the morphable face model by PCA s formulated as K S = α λ e + S ( 3 α 3) (2) = λ s the standard devaton for the th prncpal face model. Fgure 6 shows example face models generated by blendng the top three prncpal face models. he parameters of the blendng are shown n able 3. Fgure 6: Blendng face models generated from the morphable face model by PCA. he parameters of the blendng are descrbed n able 3. From left to rght; blend model, blend model2 and blend model3 he modelng by PCA offers better controllablty than the former method. Snce the bass shapes, the prncpal face models, are orthogonal wth each other, the blendng of one bass shape does not nterfere wth the blendng of the other bass shapes. However there s a drawback, e.g., the prncpal component does not have a clear semantcs of the shape tself and requre the user some tme to get the feelng of the effects that the prncpal component mght have on the result of the deformaton. 6. Concluson In ths paper, we propose an example based face modelng technque especally to create a realstc face model wth vseme control. odelng and anmatng a realstc human face requres sklls and consderable tme. he proposed method lessens the costs requred by utlzng the scanned data taken from the subect. he process runs almost automatc except that the user needs to select a set of fducal markers on the facal mages provded by the scanner. In preparaton, we generate a set of face models for the base (closed mouth) and for each vseme. he face models generated have "pont-to-pont" correspondence among them and make t straght forward to extract the moton vectors, the dsplacement vectors from the base model to the specfc model of the vseme. In order to generate a morphable model, two methods are expermented. he frst method uses the correspondent face model as the bass of the face space. A novel face model s created by a convex lnear combnaton of the bass vectors. he other method apples PCA to the face space and a novel face model s created by usng the average face model and a set of prncpal face models. he statstcal modelng method by PCA provdes another alternatve to generate a novel face whch cannot be made by the former method. he proposed method can also be used to extract expresson from a subect and generate a morphable face model wth realstc expressons. We plan to extend the morphable face model by ncorporatng the facal expressons to the face model wth vseme control, thus provdes much greater varetes of facal deformaton.

10 254 IJCSNS Internatonal Journal of Computer Scence and Network Securty, VOL.9 No.4, Aprl 2009 [2] Ulrch, Jun Yong and Noh. Expresson clonng th annual conference on Computer graphcs and nteractve technques. pp Ken Yano s a PhD student of the Graduate School of Engneerng at Hroshma Unversty after workng at Sharp Corporaton. He receved the BE n 992 from Waseda Unversty and S n 998 from Central chgan Unversty n U.S. Hs current research s manly n the area of computer graphcs and computer vson. Specal nterests nclude 2D and 3D data processng and machne learnng. He s a member of AC, IEEE, IPS of Japan. Fgure 7: Realstc face models wth expressve vseme control for two subects. Subect: frst two rows, Subect2: second two rows. From the left most column to the rght most column { base model, vseme(car), vseme(eel), vseme(man), vseme(she), vseme(too) } References [] INOLA, KONIKA. Vvd 700 : Non-contact 3D laser scanner. [2] aubn, Gabrel. A Sgnal Processng Approach to Far Surface Desgn nd annual conference on Computer graphcs and nteractve technques. [3] L.Wllams. Performance drven facal anmaton. 990, Computer Graphcs, pp [4] ckay, Paul J. Besl and Nel D. A ethod for Regstraton of 3-D Shapes. 992, IEEE ransactons on Pattern Analyss and achne Intellgence, pp [5] Popovc, Danel Vlasc and atthew Brand and Hanspeter Pfster and Jovan. Face ransfer wth ultlnear odels. 2005, AC ransactons on Graphcs, pp [6] Paul Ekman, Wallace V.Fressen. Facal Acton Codng System. [7] Blake, B Bascle. Separablty of pose and expresson n facal trackng and anmaton Int. Conf. Computer Vson. pp [8] Ahua,H.Wang. Facal expresson decomposton. IEEE Internatonal Conference on Computer Vson. pp [9] Schroder, Eml Praun and Wm Sweldens and Peter.Consstent esh Parameterzatons Computer graphcs and nteractve technques. pp [0] Vetter, Volker Blanz and homas. Face Recognton Based on Fttng a 3D orphable odel. 2003, IEEE ransactons on Pattern Analyss and achne Intellgence, pp [] Bregler, E S Chuang and H Deshpande. Facal expresson space learnng Pacfc Conference on Computer Graphcs and Applcatons. pp Koch Harada s a professor of the Graduate School of Engneerng at Hroshma Unversty. He receved the BE n 973 from Hroshma Unversty, and S and PhD n 975 and 978, respectvely, from okyo Insttute of echnology. Hs current research s manly n the area of computer graphcs. Specal nterests nclude man-machne nterface through graphcs; 3D data nput technques, data converson between 2D and 3D geometry, effectve nteractve usage of curved surfaces. He s a member of AC, IPS of Japan, and IEICE of Japan.

Range images. Range image registration. Examples of sampling patterns. Range images and range surfaces

Range images. Range image registration. Examples of sampling patterns. Range images and range surfaces Range mages For many structured lght scanners, the range data forms a hghly regular pattern known as a range mage. he samplng pattern s determned by the specfc scanner. Range mage regstraton 1 Examples

More information

Face Recognition University at Buffalo CSE666 Lecture Slides Resources:

Face Recognition University at Buffalo CSE666 Lecture Slides Resources: Face Recognton Unversty at Buffalo CSE666 Lecture Sldes Resources: http://www.face-rec.org/algorthms/ Overvew of face recognton algorthms Correlaton - Pxel based correspondence between two face mages Structural

More information

Cluster Analysis of Electrical Behavior

Cluster Analysis of Electrical Behavior Journal of Computer and Communcatons, 205, 3, 88-93 Publshed Onlne May 205 n ScRes. http://www.scrp.org/ournal/cc http://dx.do.org/0.4236/cc.205.350 Cluster Analyss of Electrcal Behavor Ln Lu Ln Lu, School

More information

Improvement of Spatial Resolution Using BlockMatching Based Motion Estimation and Frame. Integration

Improvement of Spatial Resolution Using BlockMatching Based Motion Estimation and Frame. Integration Improvement of Spatal Resoluton Usng BlockMatchng Based Moton Estmaton and Frame Integraton Danya Suga and Takayuk Hamamoto Graduate School of Engneerng, Tokyo Unversty of Scence, 6-3-1, Nuku, Katsuska-ku,

More information

TN348: Openlab Module - Colocalization

TN348: Openlab Module - Colocalization TN348: Openlab Module - Colocalzaton Topc The Colocalzaton module provdes the faclty to vsualze and quantfy colocalzaton between pars of mages. The Colocalzaton wndow contans a prevew of the two mages

More information

A Binarization Algorithm specialized on Document Images and Photos

A Binarization Algorithm specialized on Document Images and Photos A Bnarzaton Algorthm specalzed on Document mages and Photos Ergna Kavalleratou Dept. of nformaton and Communcaton Systems Engneerng Unversty of the Aegean kavalleratou@aegean.gr Abstract n ths paper, a

More information

Computer Animation and Visualisation. Lecture 4. Rigging / Skinning

Computer Animation and Visualisation. Lecture 4. Rigging / Skinning Computer Anmaton and Vsualsaton Lecture 4. Rggng / Sknnng Taku Komura Overvew Sknnng / Rggng Background knowledge Lnear Blendng How to decde weghts? Example-based Method Anatomcal models Sknnng Assume

More information

Analysis of Continuous Beams in General

Analysis of Continuous Beams in General Analyss of Contnuous Beams n General Contnuous beams consdered here are prsmatc, rgdly connected to each beam segment and supported at varous ponts along the beam. onts are selected at ponts of support,

More information

Content Based Image Retrieval Using 2-D Discrete Wavelet with Texture Feature with Different Classifiers

Content Based Image Retrieval Using 2-D Discrete Wavelet with Texture Feature with Different Classifiers IOSR Journal of Electroncs and Communcaton Engneerng (IOSR-JECE) e-issn: 78-834,p- ISSN: 78-8735.Volume 9, Issue, Ver. IV (Mar - Apr. 04), PP 0-07 Content Based Image Retreval Usng -D Dscrete Wavelet wth

More information

Lecture #15 Lecture Notes

Lecture #15 Lecture Notes Lecture #15 Lecture Notes The ocean water column s very much a 3-D spatal entt and we need to represent that structure n an economcal way to deal wth t n calculatons. We wll dscuss one way to do so, emprcal

More information

Model-Based Bundle Adjustment to Face Modeling

Model-Based Bundle Adjustment to Face Modeling Model-Based Bundle Adjustment to Face Modelng Oscar K. Au Ivor W. sang Shrley Y. Wong oscarau@cs.ust.hk vor@cs.ust.hk shrleyw@cs.ust.hk he Hong Kong Unversty of Scence and echnology Realstc facal synthess

More information

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points;

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points; Subspace clusterng Clusterng Fundamental to all clusterng technques s the choce of dstance measure between data ponts; D q ( ) ( ) 2 x x = x x, j k = 1 k jk Squared Eucldean dstance Assumpton: All features

More information

SLAM Summer School 2006 Practical 2: SLAM using Monocular Vision

SLAM Summer School 2006 Practical 2: SLAM using Monocular Vision SLAM Summer School 2006 Practcal 2: SLAM usng Monocular Vson Javer Cvera, Unversty of Zaragoza Andrew J. Davson, Imperal College London J.M.M Montel, Unversty of Zaragoza. josemar@unzar.es, jcvera@unzar.es,

More information

Reading. 14. Subdivision curves. Recommended:

Reading. 14. Subdivision curves. Recommended: eadng ecommended: Stollntz, Deose, and Salesn. Wavelets for Computer Graphcs: heory and Applcatons, 996, secton 6.-6., A.5. 4. Subdvson curves Note: there s an error n Stollntz, et al., secton A.5. Equaton

More information

MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION

MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION Paulo Quntlano 1 & Antono Santa-Rosa 1 Federal Polce Department, Brasla, Brazl. E-mals: quntlano.pqs@dpf.gov.br and

More information

Structure from Motion

Structure from Motion Structure from Moton Structure from Moton For now, statc scene and movng camera Equvalentl, rgdl movng scene and statc camera Lmtng case of stereo wth man cameras Lmtng case of multvew camera calbraton

More information

Mesh Editing in ROI with Dual Laplacian

Mesh Editing in ROI with Dual Laplacian Mesh Edtng n ROI wth Dual Laplacan Luo Qong, Lu Bo, Ma Zhan-guo, Zhang Hong-bn College of Computer Scence, Beng Unversty of Technology, Chna lqngng@sohu.com, lubo@but.edu.cn,mzgsy@63.com,zhb@publc.bta.net.cn

More information

S1 Note. Basis functions.

S1 Note. Basis functions. S1 Note. Bass functons. Contents Types of bass functons...1 The Fourer bass...2 B-splne bass...3 Power and type I error rates wth dfferent numbers of bass functons...4 Table S1. Smulaton results of type

More information

Registration of Expressions Data using a 3D Morphable Model

Registration of Expressions Data using a 3D Morphable Model Regstraton of Expressons Data usng a 3D Morphable Model Curzo Basso DISI, Unverstà d Genova, Italy Emal: curzo.basso@ds.unge.t Thomas Vetter Departement Informatk, Unverstät Basel, Swtzerland Emal: thomas.vetter@unbas.ch

More information

Multi-stable Perception. Necker Cube

Multi-stable Perception. Necker Cube Mult-stable Percepton Necker Cube Spnnng dancer lluson, Nobuuk Kaahara Fttng and Algnment Computer Vson Szelsk 6.1 James Has Acknowledgment: Man sldes from Derek Hoem, Lana Lazebnk, and Grauman&Lebe 2008

More information

A Fast Visual Tracking Algorithm Based on Circle Pixels Matching

A Fast Visual Tracking Algorithm Based on Circle Pixels Matching A Fast Vsual Trackng Algorthm Based on Crcle Pxels Matchng Zhqang Hou hou_zhq@sohu.com Chongzhao Han czhan@mal.xjtu.edu.cn Ln Zheng Abstract: A fast vsual trackng algorthm based on crcle pxels matchng

More information

Type-2 Fuzzy Non-uniform Rational B-spline Model with Type-2 Fuzzy Data

Type-2 Fuzzy Non-uniform Rational B-spline Model with Type-2 Fuzzy Data Malaysan Journal of Mathematcal Scences 11(S) Aprl : 35 46 (2017) Specal Issue: The 2nd Internatonal Conference and Workshop on Mathematcal Analyss (ICWOMA 2016) MALAYSIAN JOURNAL OF MATHEMATICAL SCIENCES

More information

An Image Fusion Approach Based on Segmentation Region

An Image Fusion Approach Based on Segmentation Region Rong Wang, L-Qun Gao, Shu Yang, Yu-Hua Cha, and Yan-Chun Lu An Image Fuson Approach Based On Segmentaton Regon An Image Fuson Approach Based on Segmentaton Regon Rong Wang, L-Qun Gao, Shu Yang 3, Yu-Hua

More information

Some Tutorial about the Project. Computer Graphics

Some Tutorial about the Project. Computer Graphics Some Tutoral about the Project Lecture 6 Rastersaton, Antalasng, Texture Mappng, I have already covered all the topcs needed to fnsh the 1 st practcal Today, I wll brefly explan how to start workng on

More information

3D Virtual Eyeglass Frames Modeling from Multiple Camera Image Data Based on the GFFD Deformation Method

3D Virtual Eyeglass Frames Modeling from Multiple Camera Image Data Based on the GFFD Deformation Method NICOGRAPH Internatonal 2012, pp. 114-119 3D Vrtual Eyeglass Frames Modelng from Multple Camera Image Data Based on the GFFD Deformaton Method Norak Tamura, Somsangouane Sngthemphone and Katsuhro Ktama

More information

Modeling, Manipulating, and Visualizing Continuous Volumetric Data: A Novel Spline-based Approach

Modeling, Manipulating, and Visualizing Continuous Volumetric Data: A Novel Spline-based Approach Modelng, Manpulatng, and Vsualzng Contnuous Volumetrc Data: A Novel Splne-based Approach Jng Hua Center for Vsual Computng, Department of Computer Scence SUNY at Stony Brook Talk Outlne Introducton and

More information

Lecture 5: Multilayer Perceptrons

Lecture 5: Multilayer Perceptrons Lecture 5: Multlayer Perceptrons Roger Grosse 1 Introducton So far, we ve only talked about lnear models: lnear regresson and lnear bnary classfers. We noted that there are functons that can t be represented

More information

FEATURE EXTRACTION. Dr. K.Vijayarekha. Associate Dean School of Electrical and Electronics Engineering SASTRA University, Thanjavur

FEATURE EXTRACTION. Dr. K.Vijayarekha. Associate Dean School of Electrical and Electronics Engineering SASTRA University, Thanjavur FEATURE EXTRACTION Dr. K.Vjayarekha Assocate Dean School of Electrcal and Electroncs Engneerng SASTRA Unversty, Thanjavur613 41 Jont Intatve of IITs and IISc Funded by MHRD Page 1 of 8 Table of Contents

More information

An Accurate Evaluation of Integrals in Convex and Non convex Polygonal Domain by Twelve Node Quadrilateral Finite Element Method

An Accurate Evaluation of Integrals in Convex and Non convex Polygonal Domain by Twelve Node Quadrilateral Finite Element Method Internatonal Journal of Computatonal and Appled Mathematcs. ISSN 89-4966 Volume, Number (07), pp. 33-4 Research Inda Publcatons http://www.rpublcaton.com An Accurate Evaluaton of Integrals n Convex and

More information

Lecture 4: Principal components

Lecture 4: Principal components /3/6 Lecture 4: Prncpal components 3..6 Multvarate lnear regresson MLR s optmal for the estmaton data...but poor for handlng collnear data Covarance matrx s not nvertble (large condton number) Robustness

More information

High-Boost Mesh Filtering for 3-D Shape Enhancement

High-Boost Mesh Filtering for 3-D Shape Enhancement Hgh-Boost Mesh Flterng for 3-D Shape Enhancement Hrokazu Yagou Λ Alexander Belyaev y Damng We z Λ y z ; ; Shape Modelng Laboratory, Unversty of Azu, Azu-Wakamatsu 965-8580 Japan y Computer Graphcs Group,

More information

An efficient method to build panoramic image mosaics

An efficient method to build panoramic image mosaics An effcent method to buld panoramc mage mosacs Pattern Recognton Letters vol. 4 003 Dae-Hyun Km Yong-In Yoon Jong-Soo Cho School of Electrcal Engneerng and Computer Scence Kyungpook Natonal Unv. Abstract

More information

LECTURE : MANIFOLD LEARNING

LECTURE : MANIFOLD LEARNING LECTURE : MANIFOLD LEARNING Rta Osadchy Some sldes are due to L.Saul, V. C. Raykar, N. Verma Topcs PCA MDS IsoMap LLE EgenMaps Done! Dmensonalty Reducton Data representaton Inputs are real-valued vectors

More information

12/2/2009. Announcements. Parametric / Non-parametric. Case-Based Reasoning. Nearest-Neighbor on Images. Nearest-Neighbor Classification

12/2/2009. Announcements. Parametric / Non-parametric. Case-Based Reasoning. Nearest-Neighbor on Images. Nearest-Neighbor Classification Introducton to Artfcal Intellgence V22.0472-001 Fall 2009 Lecture 24: Nearest-Neghbors & Support Vector Machnes Rob Fergus Dept of Computer Scence, Courant Insttute, NYU Sldes from Danel Yeung, John DeNero

More information

Load Balancing for Hex-Cell Interconnection Network

Load Balancing for Hex-Cell Interconnection Network Int. J. Communcatons, Network and System Scences,,, - Publshed Onlne Aprl n ScRes. http://www.scrp.org/journal/jcns http://dx.do.org/./jcns.. Load Balancng for Hex-Cell Interconnecton Network Saher Manaseer,

More information

Detection of an Object by using Principal Component Analysis

Detection of an Object by using Principal Component Analysis Detecton of an Object by usng Prncpal Component Analyss 1. G. Nagaven, 2. Dr. T. Sreenvasulu Reddy 1. M.Tech, Department of EEE, SVUCE, Trupath, Inda. 2. Assoc. Professor, Department of ECE, SVUCE, Trupath,

More information

PCA Based Gait Segmentation

PCA Based Gait Segmentation Honggu L, Cupng Sh & Xngguo L PCA Based Gat Segmentaton PCA Based Gat Segmentaton Honggu L, Cupng Sh, and Xngguo L 2 Electronc Department, Physcs College, Yangzhou Unversty, 225002 Yangzhou, Chna 2 Department

More information

Image Alignment CSC 767

Image Alignment CSC 767 Image Algnment CSC 767 Image algnment Image from http://graphcs.cs.cmu.edu/courses/15-463/2010_fall/ Image algnment: Applcatons Panorama sttchng Image algnment: Applcatons Recognton of object nstances

More information

Recognizing Faces. Outline

Recognizing Faces. Outline Recognzng Faces Drk Colbry Outlne Introducton and Motvaton Defnng a feature vector Prncpal Component Analyss Lnear Dscrmnate Analyss !"" #$""% http://www.nfotech.oulu.f/annual/2004 + &'()*) '+)* 2 ! &

More information

Scan Conversion & Shading

Scan Conversion & Shading Scan Converson & Shadng Thomas Funkhouser Prnceton Unversty C0S 426, Fall 1999 3D Renderng Ppelne (for drect llumnaton) 3D Prmtves 3D Modelng Coordnates Modelng Transformaton 3D World Coordnates Lghtng

More information

Feature Reduction and Selection

Feature Reduction and Selection Feature Reducton and Selecton Dr. Shuang LIANG School of Software Engneerng TongJ Unversty Fall, 2012 Today s Topcs Introducton Problems of Dmensonalty Feature Reducton Statstc methods Prncpal Components

More information

Scan Conversion & Shading

Scan Conversion & Shading 1 3D Renderng Ppelne (for drect llumnaton) 2 Scan Converson & Shadng Adam Fnkelsten Prnceton Unversty C0S 426, Fall 2001 3DPrmtves 3D Modelng Coordnates Modelng Transformaton 3D World Coordnates Lghtng

More information

Classifier Selection Based on Data Complexity Measures *

Classifier Selection Based on Data Complexity Measures * Classfer Selecton Based on Data Complexty Measures * Edth Hernández-Reyes, J.A. Carrasco-Ochoa, and J.Fco. Martínez-Trndad Natonal Insttute for Astrophyscs, Optcs and Electroncs, Lus Enrque Erro No.1 Sta.

More information

Facial Animation by Optimized Blendshapes from Motion Capture Data

Facial Animation by Optimized Blendshapes from Motion Capture Data Facal Anmaton by Optmzed Blendshapes from oton Capture Data Xuecheng Lu,2 Tanlu ao Shhong Xa Zhaoq Wang Yong Yu,2 Insttute of Computng Technology, Chnese Academy of Scences 2 Graduate Unversty of Chnese

More information

MOTION PANORAMA CONSTRUCTION FROM STREAMING VIDEO FOR POWER- CONSTRAINED MOBILE MULTIMEDIA ENVIRONMENTS XUNYU PAN

MOTION PANORAMA CONSTRUCTION FROM STREAMING VIDEO FOR POWER- CONSTRAINED MOBILE MULTIMEDIA ENVIRONMENTS XUNYU PAN MOTION PANORAMA CONSTRUCTION FROM STREAMING VIDEO FOR POWER- CONSTRAINED MOBILE MULTIMEDIA ENVIRONMENTS by XUNYU PAN (Under the Drecton of Suchendra M. Bhandarkar) ABSTRACT In modern tmes, more and more

More information

R s s f. m y s. SPH3UW Unit 7.3 Spherical Concave Mirrors Page 1 of 12. Notes

R s s f. m y s. SPH3UW Unit 7.3 Spherical Concave Mirrors Page 1 of 12. Notes SPH3UW Unt 7.3 Sphercal Concave Mrrors Page 1 of 1 Notes Physcs Tool box Concave Mrror If the reflectng surface takes place on the nner surface of the sphercal shape so that the centre of the mrror bulges

More information

User Authentication Based On Behavioral Mouse Dynamics Biometrics

User Authentication Based On Behavioral Mouse Dynamics Biometrics User Authentcaton Based On Behavoral Mouse Dynamcs Bometrcs Chee-Hyung Yoon Danel Donghyun Km Department of Computer Scence Department of Computer Scence Stanford Unversty Stanford Unversty Stanford, CA

More information

Hermite Splines in Lie Groups as Products of Geodesics

Hermite Splines in Lie Groups as Products of Geodesics Hermte Splnes n Le Groups as Products of Geodescs Ethan Eade Updated May 28, 2017 1 Introducton 1.1 Goal Ths document defnes a curve n the Le group G parametrzed by tme and by structural parameters n the

More information

AP PHYSICS B 2008 SCORING GUIDELINES

AP PHYSICS B 2008 SCORING GUIDELINES AP PHYSICS B 2008 SCORING GUIDELINES General Notes About 2008 AP Physcs Scorng Gudelnes 1. The solutons contan the most common method of solvng the free-response questons and the allocaton of ponts for

More information

Image warping and stitching May 5 th, 2015

Image warping and stitching May 5 th, 2015 Image warpng and sttchng Ma 5 th, 2015 Yong Jae Lee UC Davs PS2 due net Frda Announcements 2 Last tme Interactve segmentaton Feature-based algnment 2D transformatons Affne ft RANSAC 3 1 Algnment problem

More information

Face Recognition Based on SVM and 2DPCA

Face Recognition Based on SVM and 2DPCA Vol. 4, o. 3, September, 2011 Face Recognton Based on SVM and 2DPCA Tha Hoang Le, Len Bu Faculty of Informaton Technology, HCMC Unversty of Scence Faculty of Informaton Scences and Engneerng, Unversty

More information

Real-time Motion Capture System Using One Video Camera Based on Color and Edge Distribution

Real-time Motion Capture System Using One Video Camera Based on Color and Edge Distribution Real-tme Moton Capture System Usng One Vdeo Camera Based on Color and Edge Dstrbuton YOSHIAKI AKAZAWA, YOSHIHIRO OKADA, AND KOICHI NIIJIMA Graduate School of Informaton Scence and Electrcal Engneerng,

More information

Generating a Mapping Function from one Expression to another using a Statistical Model of Facial Shape

Generating a Mapping Function from one Expression to another using a Statistical Model of Facial Shape Generatng a Mappng Functon from one Expresson to another usng a Statstcal Model of Facal Shape John Ghent Computer Scence Department Natonal Unversty of Ireland, Maynooth Maynooth, Co. Kldare, Ireland

More information

Proper Choice of Data Used for the Estimation of Datum Transformation Parameters

Proper Choice of Data Used for the Estimation of Datum Transformation Parameters Proper Choce of Data Used for the Estmaton of Datum Transformaton Parameters Hakan S. KUTOGLU, Turkey Key words: Coordnate systems; transformaton; estmaton, relablty. SUMMARY Advances n technologes and

More information

Stitching of off-axis sub-aperture null measurements of an aspheric surface

Stitching of off-axis sub-aperture null measurements of an aspheric surface Sttchng of off-axs sub-aperture null measurements of an aspherc surface Chunyu Zhao* and James H. Burge College of optcal Scences The Unversty of Arzona 1630 E. Unversty Blvd. Tucson, AZ 85721 ABSTRACT

More information

Vectorization of Image Outlines Using Rational Spline and Genetic Algorithm

Vectorization of Image Outlines Using Rational Spline and Genetic Algorithm 01 Internatonal Conference on Image, Vson and Computng (ICIVC 01) IPCSIT vol. 50 (01) (01) IACSIT Press, Sngapore DOI: 10.776/IPCSIT.01.V50.4 Vectorzaton of Image Outlnes Usng Ratonal Splne and Genetc

More information

Image Fusion With a Dental Panoramic X-ray Image and Face Image Acquired With a KINECT

Image Fusion With a Dental Panoramic X-ray Image and Face Image Acquired With a KINECT Image Fuson Wth a Dental Panoramc X-ray Image and Face Image Acqured Wth a KINECT Kohe Kawa* 1, Koch Ogawa* 1, Aktosh Katumata* 2 * 1 Graduate School of Engneerng, Hose Unversty * 2 School of Dentstry,

More information

3D vector computer graphics

3D vector computer graphics 3D vector computer graphcs Paolo Varagnolo: freelance engneer Padova Aprl 2016 Prvate Practce ----------------------------------- 1. Introducton Vector 3D model representaton n computer graphcs requres

More information

Local Quaternary Patterns and Feature Local Quaternary Patterns

Local Quaternary Patterns and Feature Local Quaternary Patterns Local Quaternary Patterns and Feature Local Quaternary Patterns Jayu Gu and Chengjun Lu The Department of Computer Scence, New Jersey Insttute of Technology, Newark, NJ 0102, USA Abstract - Ths paper presents

More information

Parallelism for Nested Loops with Non-uniform and Flow Dependences

Parallelism for Nested Loops with Non-uniform and Flow Dependences Parallelsm for Nested Loops wth Non-unform and Flow Dependences Sam-Jn Jeong Dept. of Informaton & Communcaton Engneerng, Cheonan Unversty, 5, Anseo-dong, Cheonan, Chungnam, 330-80, Korea. seong@cheonan.ac.kr

More information

Video Object Tracking Based On Extended Active Shape Models With Color Information

Video Object Tracking Based On Extended Active Shape Models With Color Information CGIV'2002: he Frst Frst European Conference Colour on Colour n Graphcs, Imagng, and Vson Vdeo Object rackng Based On Extended Actve Shape Models Wth Color Informaton A. Koschan, S.K. Kang, J.K. Pak, B.

More information

An Optimal Algorithm for Prufer Codes *

An Optimal Algorithm for Prufer Codes * J. Software Engneerng & Applcatons, 2009, 2: 111-115 do:10.4236/jsea.2009.22016 Publshed Onlne July 2009 (www.scrp.org/journal/jsea) An Optmal Algorthm for Prufer Codes * Xaodong Wang 1, 2, Le Wang 3,

More information

Machine Learning 9. week

Machine Learning 9. week Machne Learnng 9. week Mappng Concept Radal Bass Functons (RBF) RBF Networks 1 Mappng It s probably the best scenaro for the classfcaton of two dataset s to separate them lnearly. As you see n the below

More information

A B-Snake Model Using Statistical and Geometric Information - Applications to Medical Images

A B-Snake Model Using Statistical and Geometric Information - Applications to Medical Images A B-Snake Model Usng Statstcal and Geometrc Informaton - Applcatons to Medcal Images Yue Wang, Eam Khwang Teoh and Dnggang Shen 2 School of Electrcal and Electronc Engneerng, Nanyang Technologcal Unversty

More information

The Research of Ellipse Parameter Fitting Algorithm of Ultrasonic Imaging Logging in the Casing Hole

The Research of Ellipse Parameter Fitting Algorithm of Ultrasonic Imaging Logging in the Casing Hole Appled Mathematcs, 04, 5, 37-3 Publshed Onlne May 04 n ScRes. http://www.scrp.org/journal/am http://dx.do.org/0.436/am.04.584 The Research of Ellpse Parameter Fttng Algorthm of Ultrasonc Imagng Loggng

More information

3D Face Reconstruction With Local Feature Refinement

3D Face Reconstruction With Local Feature Refinement ternatonal Journal of Multmeda and Ubqutous Engneerng Vol.9, No.8 (014), pp.59-7 http://dx.do.org/10.1457/jmue.014.9.8.06 3D Face Reconstructon Wth Local Feature Refnement Rudy Adpranata 1, Kartka Gunad

More information

3D Face Reconstruction With Local Feature Refinement. Abstract

3D Face Reconstruction With Local Feature Refinement. Abstract , pp.6-74 http://dx.do.org/0.457/jmue.04.9.8.06 3D Face Reconstructon Wth Local Feature Refnement Rudy Adpranata, Kartka Gunad and Wendy Gunawan 3, formatcs Department, Petra Chrstan Unversty, Surabaya,

More information

Real-time Joint Tracking of a Hand Manipulating an Object from RGB-D Input

Real-time Joint Tracking of a Hand Manipulating an Object from RGB-D Input Real-tme Jont Tracng of a Hand Manpulatng an Object from RGB-D Input Srnath Srdhar 1 Franzsa Mueller 1 Mchael Zollhöfer 1 Dan Casas 1 Antt Oulasvrta 2 Chrstan Theobalt 1 1 Max Planc Insttute for Informatcs

More information

Image Representation & Visualization Basic Imaging Algorithms Shape Representation and Analysis. outline

Image Representation & Visualization Basic Imaging Algorithms Shape Representation and Analysis. outline mage Vsualzaton mage Vsualzaton mage Representaton & Vsualzaton Basc magng Algorthms Shape Representaton and Analyss outlne mage Representaton & Vsualzaton Basc magng Algorthms Shape Representaton and

More information

Edge Detection in Noisy Images Using the Support Vector Machines

Edge Detection in Noisy Images Using the Support Vector Machines Edge Detecton n Nosy Images Usng the Support Vector Machnes Hlaro Gómez-Moreno, Saturnno Maldonado-Bascón, Francsco López-Ferreras Sgnal Theory and Communcatons Department. Unversty of Alcalá Crta. Madrd-Barcelona

More information

The Codesign Challenge

The Codesign Challenge ECE 4530 Codesgn Challenge Fall 2007 Hardware/Software Codesgn The Codesgn Challenge Objectves In the codesgn challenge, your task s to accelerate a gven software reference mplementaton as fast as possble.

More information

3D Novel Face Sample Modeling for Face Recognition

3D Novel Face Sample Modeling for Face Recognition JOURAL OF MULTIMEDIA, VOL. 6, O. 5, OCTOBER 20 467 3D ovel Face Sample Modelng for Face Recognton Yun Ge, Yanfeng Sun, Baoca Yn, Henglang Tang Beng Key Laboratory of Multmeda and Intellgent Software Technology

More information

Machine Learning: Algorithms and Applications

Machine Learning: Algorithms and Applications 14/05/1 Machne Learnng: Algorthms and Applcatons Florano Zn Free Unversty of Bozen-Bolzano Faculty of Computer Scence Academc Year 011-01 Lecture 10: 14 May 01 Unsupervsed Learnng cont Sldes courtesy of

More information

Feature-Preserving Mesh Denoising via Bilateral Normal Filtering

Feature-Preserving Mesh Denoising via Bilateral Normal Filtering Feature-Preservng Mesh Denosng va Blateral Normal Flterng Ka-Wah Lee, Wen-Png Wang Computer Graphcs Group Department of Computer Scence, The Unversty of Hong Kong kwlee@cs.hku.hk, wenpng@cs.hku.hk Abstract

More information

A Fast Content-Based Multimedia Retrieval Technique Using Compressed Data

A Fast Content-Based Multimedia Retrieval Technique Using Compressed Data A Fast Content-Based Multmeda Retreval Technque Usng Compressed Data Borko Furht and Pornvt Saksobhavvat NSF Multmeda Laboratory Florda Atlantc Unversty, Boca Raton, Florda 3343 ABSTRACT In ths paper,

More information

A PATTERN RECOGNITION APPROACH TO IMAGE SEGMENTATION

A PATTERN RECOGNITION APPROACH TO IMAGE SEGMENTATION 1 THE PUBLISHING HOUSE PROCEEDINGS OF THE ROMANIAN ACADEMY, Seres A, OF THE ROMANIAN ACADEMY Volume 4, Number 2/2003, pp.000-000 A PATTERN RECOGNITION APPROACH TO IMAGE SEGMENTATION Tudor BARBU Insttute

More information

Ecient Computation of the Most Probable Motion from Fuzzy. Moshe Ben-Ezra Shmuel Peleg Michael Werman. The Hebrew University of Jerusalem

Ecient Computation of the Most Probable Motion from Fuzzy. Moshe Ben-Ezra Shmuel Peleg Michael Werman. The Hebrew University of Jerusalem Ecent Computaton of the Most Probable Moton from Fuzzy Correspondences Moshe Ben-Ezra Shmuel Peleg Mchael Werman Insttute of Computer Scence The Hebrew Unversty of Jerusalem 91904 Jerusalem, Israel Emal:

More information

What are the camera parameters? Where are the light sources? What is the mapping from radiance to pixel color? Want to solve for 3D geometry

What are the camera parameters? Where are the light sources? What is the mapping from radiance to pixel color? Want to solve for 3D geometry Today: Calbraton What are the camera parameters? Where are the lght sources? What s the mappng from radance to pel color? Why Calbrate? Want to solve for D geometry Alternatve approach Solve for D shape

More information

Outline. Self-Organizing Maps (SOM) US Hebbian Learning, Cntd. The learning rule is Hebbian like:

Outline. Self-Organizing Maps (SOM) US Hebbian Learning, Cntd. The learning rule is Hebbian like: Self-Organzng Maps (SOM) Turgay İBRİKÇİ, PhD. Outlne Introducton Structures of SOM SOM Archtecture Neghborhoods SOM Algorthm Examples Summary 1 2 Unsupervsed Hebban Learnng US Hebban Learnng, Cntd 3 A

More information

A Robust Method for Estimating the Fundamental Matrix

A Robust Method for Estimating the Fundamental Matrix Proc. VIIth Dgtal Image Computng: Technques and Applcatons, Sun C., Talbot H., Ourseln S. and Adraansen T. (Eds.), 0- Dec. 003, Sydney A Robust Method for Estmatng the Fundamental Matrx C.L. Feng and Y.S.

More information

Learning Ensemble of Local PDM-based Regressions. Yen Le Computational Biomedicine Lab Advisor: Prof. Ioannis A. Kakadiaris

Learning Ensemble of Local PDM-based Regressions. Yen Le Computational Biomedicine Lab Advisor: Prof. Ioannis A. Kakadiaris Learnng Ensemble of Local PDM-based Regressons Yen Le Computatonal Bomedcne Lab Advsor: Prof. Ioanns A. Kakadars 1 Problem statement Fttng a statstcal shape model (PDM) for mage segmentaton Callosum segmentaton

More information

Constructing Minimum Connected Dominating Set: Algorithmic approach

Constructing Minimum Connected Dominating Set: Algorithmic approach Constructng Mnmum Connected Domnatng Set: Algorthmc approach G.N. Puroht and Usha Sharma Centre for Mathematcal Scences, Banasthal Unversty, Rajasthan 304022 usha.sharma94@yahoo.com Abstract: Connected

More information

Radial Basis Functions

Radial Basis Functions Radal Bass Functons Mesh Reconstructon Input: pont cloud Output: water-tght manfold mesh Explct Connectvty estmaton Implct Sgned dstance functon estmaton Image from: Reconstructon and Representaton of

More information

Tsinghua University at TAC 2009: Summarizing Multi-documents by Information Distance

Tsinghua University at TAC 2009: Summarizing Multi-documents by Information Distance Tsnghua Unversty at TAC 2009: Summarzng Mult-documents by Informaton Dstance Chong Long, Mnle Huang, Xaoyan Zhu State Key Laboratory of Intellgent Technology and Systems, Tsnghua Natonal Laboratory for

More information

Development of an Active Shape Model. Using the Discrete Cosine Transform

Development of an Active Shape Model. Using the Discrete Cosine Transform Development of an Actve Shape Model Usng the Dscrete Cosne Transform Kotaro Yasuda A Thess n The Department of Electrcal and Computer Engneerng Presented n Partal Fulfllment of the Requrements for the

More information

3D Face Modeling Using the Multi-Deformable Method

3D Face Modeling Using the Multi-Deformable Method Sensors 2012, 12, 12870-12889; do:10.3390/s121012870 Artcle OPEN ACCESS sensors ISSN 1424-8220 www.mdp.com/journal/sensors 3D Face Modelng Usng the Mult-Deformable Method Jnkyu Hwang, Sunjn Yu, Joongrock

More information

Multiresolution Modeling for Real-Time Facial Animation

Multiresolution Modeling for Real-Time Facial Animation Multresoluton Modelng for Real-me Facal Anmaton Soo-Kyun Km, Jong-In Cho and Chang-Hun Km Department of Computer Scence and Engneerng, Korea Unersty {ncesk, cho, chkm}@cgr.korea.ac.kr Fgure 1. Same expresson

More information

An AAM-based Face Shape Classification Method Used for Facial Expression Recognition

An AAM-based Face Shape Classification Method Used for Facial Expression Recognition Internatonal Journal of Research n Engneerng and Technology (IJRET) Vol. 2, No. 4, 23 ISSN 2277 4378 An AAM-based Face Shape Classfcaton Method Used for Facal Expresson Recognton Lunng. L, Jaehyun So,

More information

Visual Thesaurus for Color Image Retrieval using Self-Organizing Maps

Visual Thesaurus for Color Image Retrieval using Self-Organizing Maps Vsual Thesaurus for Color Image Retreval usng Self-Organzng Maps Chrstopher C. Yang and Mlo K. Yp Department of System Engneerng and Engneerng Management The Chnese Unversty of Hong Kong, Hong Kong ABSTRACT

More information

y and the total sum of

y and the total sum of Lnear regresson Testng for non-lnearty In analytcal chemstry, lnear regresson s commonly used n the constructon of calbraton functons requred for analytcal technques such as gas chromatography, atomc absorpton

More information

Fitting: Deformable contours April 26 th, 2018

Fitting: Deformable contours April 26 th, 2018 4/6/08 Fttng: Deformable contours Aprl 6 th, 08 Yong Jae Lee UC Davs Recap so far: Groupng and Fttng Goal: move from array of pxel values (or flter outputs) to a collecton of regons, objects, and shapes.

More information

Fitting & Matching. Lecture 4 Prof. Bregler. Slides from: S. Lazebnik, S. Seitz, M. Pollefeys, A. Effros.

Fitting & Matching. Lecture 4 Prof. Bregler. Slides from: S. Lazebnik, S. Seitz, M. Pollefeys, A. Effros. Fttng & Matchng Lecture 4 Prof. Bregler Sldes from: S. Lazebnk, S. Setz, M. Pollefeys, A. Effros. How do we buld panorama? We need to match (algn) mages Matchng wth Features Detect feature ponts n both

More information

3D Rigid Facial Motion Estimation from Disparity Maps

3D Rigid Facial Motion Estimation from Disparity Maps 3D Rgd Facal Moton Estmaton from Dsparty Maps N. Pérez de la Blanca 1, J.M. Fuertes 2, and M. Lucena 2 1 Department of Computer Scence and Artfcal Intellgence ETSII. Unversty of Granada, 1871 Granada,

More information

Extraction of 3D Facial Motion Parameters from Mirror-Reflected Multi-View Video for Audio-Visual Synthesis

Extraction of 3D Facial Motion Parameters from Mirror-Reflected Multi-View Video for Audio-Visual Synthesis AVSP 00 Internatonal Conference on Audtory-Vsual Speech Processng Extracton of 3D Facal Moton Parameters from Mrror-Reflected Mult-Vew Vdeo for Audo-Vsual Synthess I-Chen Ln Jeng-Sheng Yeh Mng Ouhyoung

More information

Takahiro ISHIKAWA Takahiro Ishikawa Takahiro Ishikawa Takeo KANADE

Takahiro ISHIKAWA Takahiro Ishikawa Takahiro Ishikawa Takeo KANADE Takahro ISHIKAWA Takahro Ishkawa Takahro Ishkawa Takeo KANADE Monocular gaze estmaton s usually performed by locatng the pupls, and the nner and outer eye corners n the mage of the drver s head. Of these

More information

Laplacian Eigenmap for Image Retrieval

Laplacian Eigenmap for Image Retrieval Laplacan Egenmap for Image Retreval Xaofe He Partha Nyog Department of Computer Scence The Unversty of Chcago, 1100 E 58 th Street, Chcago, IL 60637 ABSTRACT Dmensonalty reducton has been receved much

More information

Articulated Motion Capture from Visual Hulls in High Dimensional Configuration Spaces

Articulated Motion Capture from Visual Hulls in High Dimensional Configuration Spaces Semnar Presentaton June 18th, 010 Artculated Moton Capture from Vsual Hulls n Hgh Dmensonal Confguraton Spaces Azawa Yamasak Lab D, 48-09741, 羅衛蘭 ABSTRACT In ths paper, we propose a novel approach for

More information

Graph-based Clustering

Graph-based Clustering Graphbased Clusterng Transform the data nto a graph representaton ertces are the data ponts to be clustered Edges are eghted based on smlarty beteen data ponts Graph parttonng Þ Each connected component

More information

Feature-Preserving Denoising of Point-Sampled Surfaces

Feature-Preserving Denoising of Point-Sampled Surfaces Feature-Preservng Denosng of Pont-Sampled Surfaces Jfang L College of Computer Scence and Informaton Technology Zhejang Wanl Unversty Nngbo 315100 Chna Abstract: Based on samplng lkelhood and feature ntensty,

More information

EECS 730 Introduction to Bioinformatics Sequence Alignment. Luke Huan Electrical Engineering and Computer Science

EECS 730 Introduction to Bioinformatics Sequence Alignment. Luke Huan Electrical Engineering and Computer Science EECS 730 Introducton to Bonformatcs Sequence Algnment Luke Huan Electrcal Engneerng and Computer Scence http://people.eecs.ku.edu/~huan/ HMM Π s a set of states Transton Probabltes a kl Pr( l 1 k Probablty

More information