Dynamic Neighborhoods in Active Surfaces for 3D Segmentation

Size: px
Start display at page:

Download "Dynamic Neighborhoods in Active Surfaces for 3D Segmentation"

Transcription

1 Internatonal Journal of Computatonal Vson and Bomechancs Internatonal Journal for Computatonal Vson and Bomechancs, Vol. 1, No., July-December 008 Vol.1 No. (July-December, 015 Serals Publcatons 008 ISSN Dynamc Neghborhoods n Actve Surfaces for 3D Segmentaton Julen Olver, Julen Mlle, Romuald Boné & Jean-Jacques Rousselle Unversté Franços Rabelas de Tours, Laboratore Informatque 64 Avenue Jean Portals, 3700 Tours, France E-mals: julen.olver, julen.mlle, romuald.bone, rousselle@unv-tours.fr Actve contours and surfaces are deformable models used for D and 3D mage segmentaton. In ths paper, we propose two methods developed n order to accelerate 3D mage segmentaton process. They are adaptatons on actve surfaces of two methods developed for D actve contour. We use them on a dscrete 3D surface model (mesh evolvng wth the greedy algorthm. Those methods wll be compared to the classcal greedy algorthm and to a recent fast adaptaton of the level set method. Keywords: Actve surface, mesh, greedy algorthm, fast level set 1. INTRODUCTION Actve contours or snakes, ntally developed by Kass et al. n [1], are powerful segmentaton tools thanks to ther nose robustness and ablty to generate lnked closed boundares. Ther 3D extensons, actve surfaces, were developed accordng to several mplementatons (see [] [3] for surveys on 3D deformable models. Among these mplementatons, meshes are explct dscrete representatons [4], whch represents the surface as a set of nterconnected vertces. The model s deformed by drect modfcatons of vertces coordnates. Several evoluton algorthms have been developed to deform them. One of the most popular s the greedy algorthm [5] because of ts effcency. An adaptaton of the greedy algorthm on 3D surface was proposed by Bulptt and Efford n [6]. Conversely, mplct mplementatons, based on the level set framework [7], handle the surface as the zero level of a hypersurface, defned on the same doman as the mage (for 3D mages, the hypersurface s a 3 applcaton. Level sets are often chosen for ther natural handlng of topologcal changes and adaptveness to an y dmenson. Ther algorthmc complexty s proportonal to the mage resoluton, makng them tme-consumng. Despte the development of acceleratng methods (lke the narrow band technc [7], the fast marchng method [8] and the recent fast level set [9], ther computatonal cost remans hgh, preventng ther use n tme-crtcal applcatons. Moreover, mesh surfaces have several advantages over ther mplct counterparts. Ther representaton s more ntutve, and thus allow easer modelng of a pror knowledge and user nteracton. The man drawback s that meshes do not modfy ther topology naturally (techncs for detecton of topologcal chan ges must be mplemented besde the evoluton algorthm. In many applcatons, the topology of the area of nterest s known n advance. When segmentng mages n whch pror knowledge about the object topology s avalable, we beleve that mesh-based approaches should be prvleged over mplct surfaces. In ths paper, we deal wth a 3D trangular mesh drven by greedy algorthm. The model s able to perform remeshng, thus provdng geometrcal versatlty (n the same manner that D reparameterzaton techncs overcome the lack of geometrcal flexblty of tradtonal snakes. Several methods have been developed n [10] n order to accelerate D actve contours. In ths paper, we study 3D adaptatons of these methods n order to accelerate dscrete actve surfaces, as an extenson of the work n [11]. The outlne of ths paper s as follows: secton presents the 3D model and ts energes, the greedy algorthm for actve surfaces and the remeshng prncple. Secton 3 and 4 descrbe the shfted neghborhood method and the lne search method. Secton 5 descrbes the Fast Level Set mplementaton for actve contours. Secton 6 shows our expermental results on 3D models, comparng the performances of our acceleraton methods wth the basc greedy algorthm and the recent Fast Level Set mplementaton method. Secton 7 concludes wth our work expectng the future developments.. THE 3D ACTIVE SURFACE MODEL In a contnuous doman, a 3D deformable model s represented by a parameterzed surface S mappng a couple of parameters (u, v to a space pont (x, y, z T : S : 3 (u, v (x(u, v, y(u, v, z(u, v T (1 The parameter doman s normalzed : = [0, 1]. The surface s attached to the mage I, whch s a 3 functon, mappng each voxel (x, y, z to a gray level I(x, y, z. Basng ourselves on the work by [5], the surface s endowed wth the energy functonal E(S. S S S E( S u v uv

2 174 Internatonal Journal of Computatonal Vson and Bomechancs S S 0 0 P( S dudv ( u v Segmentaton of a regon of nterest n mage I s performed by determnng the optmal surface S * mnmzng E. The energy functonal depends on two knds of energes. The frst one quantfes the geometrcal regularty of the surface, whereas the second one s the external term, dependng on the dstance between the surface and the salent edges of the mage. Frst and second order partal dervatves of S wth respect to u and v are smoothng terms. Coeffcents j weght the sgnfcance of regularzng components wth respect to the external term. 10 and 01 are the elastcty coeffcents, 0 and 0 are the rgdty coeffcents and 11 s the resstance to twst. Snce our segmentaton s boundarybased, P s the external term attractng surface ponts towards salent boundares on the mage. It must decrease as the edge magntude ncreases, hence we choose P = I. It should be noted that other attractng functons could be chosen, e.g. a dstance map wth respect to the thresholded edge mage. To mplement the actve surface, we use the dscrete representaton descrbed n [1], whch s a trangular mesh made up of n vertces, denoted p = (x, y, z T 3, and edges connectng the vertces (makng a set of adjacent trangles. In order to represent the connectvty noton, each vertex p has a set of adjacent vertces, denoted A. The mesh s bult from successve subdvsons of an cosahedron [13] [14], th us leadng to a sphere-lke surface wth a homogeneous vertex dstrbuton (see fgure 1. Fgure 1: The basc cosahedron and ts frst two tesselatons Intally developped for D actve contours by Wllams and Shah n [5], the greedy algorthm s an energy mnmzng method frst proposed as an alternatve to the varatonnal method [1] and the dynamc programmng [15]. It has been recently used for D segmentaton n [16] and [17]. In [6] one may fnd an extenson of ths algorthm for 3D meshes. Global energy mnmzaton s performed va successve local optmzatons. Each vertex s endowed wth ts own energy. Hence, unlke n eq., the energy functonal s the sum of vertex energes. n E( S E( p (3 1 At each teraton, a cubc neghborhood of sde length w around each vertex s consdered (see fg.. The energy s computed at each voxel belongng to the neghborhood and the vertex s moved to the locaton leadng to the lowest energy. Ths approach avods to perform gradent descent of a partal dervatves equaton, derved from the energy functonal usng classcal Euler-Lagrange scheme [13]. Hence, t s not necessary to derve analytcally the energes, wth respect to vertex coordnates. Fgure : Cubc centered neghborhood Unlke n the classcal greedy algorthm, our methods wll deal wth neghborhoods whch are not necessarly (t centered around the vertces. We defne, the neghborhood of the th vertex at teraton t: ( t ( t ( t p r s r[0, 1] 3 (4 s ( s, s, s ( t T x y z w s the shft vector, representng the coordnates of the th vertex at teraton t relatvely to an orgnal voxel chosen on the corner of the neghborhood. At the begnnng, all vertces are centered n ther neghborhood, (0 T hence we have s ( w /, w /, w /. The ntal poston beng p, we denote p a tested locaton n the neghborhood. Once all energes have been computed the new locaton of vertex p s chosen: p ( t 1 arg mn E( pk p (5 k The energy of a vertex at locaton p s a weghted sum of dscrete nternal and external energes, normalzed on the whole neghborhood. E(p = E cont (p + E curv (p + E grad (p + E bal (p (6 The coeffcents,, and are the weghts defnng the relatve nfluence of the energes. The contnuty E cont and the curvature E curv are dscrete mplementatons of frst and second order surface dervatves of eq., respectvely. Parameters controls the surface elastcty whereas s the rgdty. They have a smlar role than coeffcents j n the contnuous model. Let us descrbe the adaptaton of the dfferent energes to our 3D model. The energes are ntutve extensons of the D actve contour ones, sutable to our mesh representaton. Whle the dscretzaton of nternal energes s smple for a parametrc snake, t s not obvous how to mplement the surface dervatves of eq. n a trangular mesh. For a dscrete planar contour, the contnuty s the dstance between successve neghbors. However, lke n [5], ( t

3 Dynamc Neghbourhoods n Actve Surfaces for 3D Segmentaton 175 t s preferable to use the absolute dfference between ths dstance and a default length d, whch may be the mean dstance computed over all vertces. Extendng ths prncple to the mesh, E cont mantans the vertces evenly spaced along the surface. Mnmzng t reduces the gap between the mean squared dstance d and the dstance between the consdered vertex and ts adjacent vertces. E ( p d p p cont j ja d 1 1 n j n 1 A ja p p (7 The second nternal energy s the curvature E curv, whch mnmzaton results n a local smoothng effect, by makng the vertex get closer to the centrod of ts adjacent vertces. In a D snake, curvature s equvalent to the squared dstance between the vertex and the mddle of ts two neghbors. By extenson to 3D, the curvature of the tested pont p s the squared dstance between p and the centrod of the neghbors of vertex p. E 1 ( p p p (8 A j A curv j Note that for a gven mesh vertex p, E curv (p = 0 f p and all ts adjacent vertces le on the same plane. To attract vertces towards salent edges, the external energy E grad s a functon of normalzed gradent magntude g of mage I. In presence of nosy data, the mage s smoothed wth a gaussan flter pror to gradent operaton. In the followng equatons, G s a gaussan kernel wth standard devaton, s the convoluton operator and g max s the maxmal edge ntensty n the mage. g(p = I(p G /g max E grad (p = g(p (9 As regards gradent magntude, real 3D edge detecton s obtaned by convolvng the mage wth the Zucker- Hummel operator [18], whch usually yelds better edge localzaton than usng drect fnte dfferences. It s made up of three masks ZH x, ZH y and ZH z, each mask flterng the mage n one dmenson. k1 0 k1 k 0 k k1 0 k1 ZH k 0 k k 0 k k 0 k x 3 3 k1 0 k 1 k 0 k k1 0 k 1 k1 k k1 k k3 k k1 k k1 ZH y k k k k k k k k k k1 k k k1 k k1 ZH k k k k k k z 3 3 k1 k k k1 k k 1 (10 (11 (1 3 k1 ; k ; k3 1 (13 3 To ncrease the capture range, we ntroduce a balloon energy E bal derved from the nflaton force proposed n [19]. It allows the mesh to be ntalzed far from the object boundares. E bal ( p p ( p kn (14 where n s the unt nward normal, defned at vertex p. The normal of vertex p s the normalzed sum of the normals of the neghborng trangles [14]. Rgorously, the normal of a trangle t s the unt vector orthogonal to the plane defned by t. In the followng expressons, T s the set of neghborng trangles of p. The normal n t of a gven trangle s the normalzed cross product between two vectors belongng to the correspondng plane. nt tt ( p ( t p t p 1 t p 3 t1 nt ; nt st ( pt p ( t p 1 t p 3 t1 n (15 tt t where p tj, j = are vertces of trangle t (p must be one of them. s t = ±1 s the sgn changng the orentaton of n t nsurng that t ponts towards the nteror of the surface. Such a calculaton of the normal vector s necessary to a correct balloon mplementaton. The moton resultng from the balloon energy mnmzaton s ether an expanson or a retracton of the surface, dependng on the sgn of coffecent. Ths one must be chosen regardng the ntal poston of the surface wth respect to the target object. In order to adapt local topology, remeshng s performed after each teraton of the greedy algorthm. The mesh s allowed to add or delete vertces to keep the dstance between adjacent vertces homogeneous, resultng n a stable vertex dstrbuton [0] [14] [1]. It nsures that every couple of adjacent vertces (p, p j satsfes the constrant: d mn p p j d max (16 where d mn and d max are two user-defned thresholds, such that d max d mn. We choose ther values near the neghborhood wdth w, so that the surface samplng s consstent wth the moton range of the vertces. Addng or deletng vertces modfes local topology, thus topologcal constrants should be verfed. To perform vertex addng or deletng, p and p j should share exactly two common adjacent vertces: A A j =. When p p j > d max, a new vertex s created at the mddle of lne segment p p j and connected to p a and p b (see mddle part of fgure 3. When p p j < d mn, p j s deleted and p s translated to the mddle locaton (see rght part of fgure THE SHIFTED NEIGHBORHOOD METHOD In order to mprove actve surfaces completon tme, we used the shfted neghborhood method developed for D snakes

4 176 Internatonal Journal of Computatonal Vson and Bomechancs Fgure 3: Remeshng operatons: vertex addng and deletng n [10]. We adapted ths greedy-based method on actve surfaces. For each vertex and at each teraton, we modfy the neghborhood n order to drect the searchng space of each vertex to the drectons that seems the most nterestng. To defne where these drectons are, we use the nformaton of the drecton followed by each vertex durng the last teraton. So each neghborhood wll be shfted from one voxel n the drecton followed prevously. At each teraton, we compute the next shft appled to the vertex wth: ( t 1 ( t 1 ( t d ( 1, 1, p p (17 ( t The vector quantty d represents the dsplacement appled on the neghborhood of the th vertex at teraton t. s a shft lmtng functon, boundng the vector coordnates between two scalars: max( b1, mn( b, ux ( b1, b, u max( b1, mn( b, uy (18 max( b1, mn( b, uz ( t 1 The dsplacement d gven by equaton (17 allows ( t 1 us to defne the new shft vector s for each vertex of the actve surface. Thus we have: ( t 1 ( ( 1 (1, w, t t s s d (19 Algorthm 1 Shfted Neghborhood Method: 3D Model 1: for t 1 to T do : for 1 to n do (t+1 3: p = arg mn E( p ( t k pk ( t 1 ( t 1 ( t 4: d ( 1,1, p p ( 1 ( ( 1 5. (1,, s w s d t N t p t r s r[0, w 1] ( 1 ( 1 ( 3 6: 7: end for 8: end for The next teraton of the greedy algorthm wll be helded wth these new neghborhoods. At ths stage, we can defne the algorthm for the shfted neghborhood method. Ths last conssts n computng the new neghborhood (t+1 wth equatons (4, (17 and (19 at the end of each teraton, once all the vertces of the actve surface have been modfed. When ncluded n the greedy algorthm for an actve surface of n vertces and T teratons, the shfted neghborhood method s descrbed n algorthm THE LINE SEARCH METHOD The lne search method [10] orgnally appled for twodmensonal actve contours allows to reduce completon tme effcently. We adapted ths method on 3D actve surfaces. The prncple of ths approach s to antcpate on the next teraton of the greedy algorthm usng the nformaton taken from the prevous one. Ths method s launched at the end of each teraton of the greedy algorthm, once all the vertces have been translated. The drecton followed by each vertex p s memorzed and we look toward t for a fxed number of voxels, whch creates a lnear neghborhood. These lasts are compared by computng ther global energes n a smlar way as t s done wth the cubc neghborhood (see equaton (6. The voxel gvng the lowest energy s then chosen for the new locaton of the current vertex. As a result, for each vertex, two neghborhoods are scanned consecutvely: the cubc centered neghborhood and the lnear neghborhood. The second algorthm descrbes the lne search method ntegrated n the greedy algorthm for actve surfaces. Let T be the number of teratons to be done by the greedy algorthm and l the number of voxels to be explored (length of the lnear neghborhood. 5. LEVEL SET AND FAST LEVEL SET IMPLEMENTATION In order to compare our methods wth recent breaktroughs n the mage segmentaton doman we adapted the Fast Level Set mplementaton method to our 3D segmentaton problem. Dynamc neghborhoods n actve surfaces for 3D segmentaton 9 Algorthm : Lne Search Method: 3D Model 1: for t 1 to T do : for 1 to n do p arg mn E( p ( t 1 3: ( t pk p 4: Determne the drecton v p k p ( t 1 ( t ( t 1 ( t p 5: Lne Search: m = ( t arg mn E( p kv k[0, l] ( t 1 ( t1 6: Update: p p v 7: end for 8: end for m Basng ourselves on the work of Osher et al. [] and [7], we mplemented our actve surface model wth level sets. We consder the parameterzed surface S defned n secton, on whch we add a tme dependency. S : + 3 (u, v, t (x(u, v, t, y(u, v, t, z(u, v, t T (0 The surface evolves accordng to the followng partal dfferental equaton:

5 Dynamc Neghbourhoods n Actve Surfaces for 3D Segmentaton 177 S( u, v, t Fn (1 t where n s the nward normal vector and F s the speed functon appled on the surface ponts. The surface boundary s mplctly represented as the zero-level of a functon : 3 +. It comes: (S(u, v, t, t = 0 (u, v, t 0 ( Functon evolves on ts zero-level accordng to the equaton: ( S( u, v, t, t F( u, v, t ( S( u, v, t, t (3 t If the speed functon F s defned over the entre doman 3 + then eq. 3 can be extended such as: ( x, t 3 F( x, t ( x, t x, t 0 (4 t The man advantage of the level-set mplementaton s ts ablty to automatcally handle topologcal changes. Indeed, the curve can naturally splt or merge wth others wthout any addtonnal mplementaton. Unfortunately the level-set method requres sgnfcant computatonnal tme. Several methods have been developed n order to accelerate the level-set mplemen taton. Th e n arrow ban d mplementaton [8] allows to decrease the complexty of the algorthm from O(n to O(n, n beng the sze of the mage grd. The author consders a band around the zero-level of the level-set functon. The partal derental equaton (PDE s solved only nsde ths regon and not n the entre defnton doman of. Although ths technque allows to sgnfcally reduce computatonal tme of the level-set mplementaton, t stll lmts the use of the level-set n real tme applcatons such as object trackng. Sh et al have recently developed n [9] an acceleraton method based on the narrow band mplementaton called the Fast Level Set method. As t was orgnally desgned for D segmentaton, we extended t to 3D. The Fast Level Set method uses two lsts to represent the surface: the lst of outsde boundary ponts L out and the lst of nsde boundary ponts L n. L out = {x (x > 0, y N(x, (y < 0} L n = {x (x < 0, y N(x, (y > 0} (5 where N(x s the dscrete neghborhood of x. The authors assume (x to take only four nteger values, accordng to the poston of x: 1 f xlout 1 f x Ln (x 3 f x soutsdes and xlout (6 3 f x snsde S and xln The authors defne the dcrete optmalty condton for the surface as: The surface S wth boundary ponts L n and L out s optmal f the speed functon F satsfes: F(x < 0 x L out and F(x > 0 x L n (7 Whle ths optmalty condton s not reached, the speed functon F s calculated for every pont n L n and L out. If F(x > 0 at a pont n L out the surface s moved outward. If F(x < 0 at a pont n L n the surface s moved nward. Once the surface has moved, the two lsts are updated. The man dea of ths method s thus to make the surface evolve wthout solvng the PDE of 4 whch requres sgnfcant computatonnal tme but only by computng the speed functon F. Agan, segmentaton s boundary-based, hence the surface should stop on strong edges. As a result, F s a functon of curvature and gradent magntude. F(x = I(x (x (8 The curvature regularzes and s weghted by coecent. It s expressed as follows: dv = xx ( y z yy ( x z zz ( x y xy x y xz x z yz y z wth notatons x ( 3/ x y z, xx,... x x For more detals and testngs on the method one can refer to [9]. The next secton descrbes and compares the results obtaned on tested mages wth our greedy acceleraton technques. 6. EXPERIMENTAL RESULTS In ths secton, we present our experments on 3D mages. We compare the shfted neghborhood method and the lne search method to the classcal greedy algorthm. Ths comparatve study also ncludes the results obtaned wth mplct surface modelng mplemented wth the fast level set method. Each tested mage s made of several slces of gray objects embedded n whte backgrounds, hghly corrupted wth gaussan nose. Derent values of neghborhood wdth w are tested (obvously, the shfted neghborhood s not expermented wth w = 3. For each mage, the surface s ntalzed as a sphere wth dentcal center and radus for all evoluton methods, ndependently from ts mplementaton (a mesh or a level set. In order to evaluate segmentaton, we use a functon takng nto account the overall dstance between the estmated boundary and ground truth. Let be the set of voxels belongng to the real boundary, and the set of voxels belongng to the estmated boundary. For each voxel on the estmated boundary, we consder the dstance to the nearest voxel on the real boundary, and conversely. The modfed Hausdor dstance mean ntroduced n [3] measures the average fttng of the surface to the real boundary. mean (, = max(h mean (,, h mean (, h mean (, = 1 mn p q q p (30

6 178 Internatonal Journal of Computatonal Vson and Bomechancs In what follows, we compare computatonal tmes obtaned on fnal surfaces havng equvalent qualtes, wth respect to the modfef Hausdo dstance. Typcally, H mean have values around 1, whch corresponds to good fttng of the real boundares. The frst mage s a data set representng a spral and was chosen n order to test the methods when vertex addng s enabled. The actve surface s ntalzed nsde the 3D model wth only 1 vertces. The fnal meshes for the three methods contan about one thousand vertces. We choose = 0, = 0.5, = and [ 0.6, 1.1]. The second mage s a model of a vase and s nterestng to test the nfltraton of the model nto the concavtes. For ths partcular 3D dataset, remeshng of the model s dsabled. We use = 0, = 0.5 for the greedy algorthm, 0.4 for the shfted neghborhood method and 0.3 for the lne search, = and = 0.8. We ntalze the meshes wth 56 vertces. The thrd mage represents three ellpsods and allows to have both salent and smooth angles on the same model. The mage sze s We prevent remeshng of the model and ntalze t wth 56 vertces. The parameters are = 0.5, = 0.3 for the greedy algorthm and 0.4 for the shfted neghborhood method and the lne search, = and [0.3, 0.7]. In order to compare boundary fttng ablty of the methods on real data, we tested them on computed tomography (CT data sets of the abdomen, n whch the actve surface was used to detect boundares of the aorta. Such segmentaton s done n the framework of abdomnal aortc aneurysm dagnoss. The mesh was ntalazed as a small sphere nsde the aorta and nflated thanks to the balloon energy. Fgure 4 shows one slce of each mage and the vsual 3D results obtaned wth the lne search method. Tables 1 and lst computatonal tmes obtaned on synthetc and CT mages, respectvely. Each method and each wndow wdth s systematcally tested on the mages (tests were made on a Pentum IV 1.7 GHz wth 51 Mb RAM. Mesh-based methods (ntal greedy algorthm, LS and SN need fewer teratons than the level set surface to reach the boundares, whch s manly due to the level set front. A voxel neghborng the front (the L out lst needs more than one teraton to change ts status, from outer to nner voxel. An teratons of the level set method s also more computatonally ntensve than one of the explct methods. The dscretzaton of the evolvng front s the same as the mage grd, so that each voxel located on the front needs to be consdered. On the mesh, the moton of a vertex does not aect ts coordnates but also all neghborng trangles, whch cover many voxels (the quantty of voxels depends on the samplng resoluton w. As regards dstance measures, t s nterestng to note that explct methods lead to more accurate results than the level set approach, except on the Vase data set. We may assume that level sets are globally more senstve to nose and prone to boundary leakage ssues, because of the mplct formulaton of the regularzng curvature energy, whch has a lmted range. On the mesh, the nternal energy of a vertex aects a larger porton of the surface. The better performance of the mplct surface on the Vase data set s explaned by the presence of sharp angles at the shape borders. Curvature prevents the mesh from fttng angular parts accurately, whereas the level set surface s not lmted thanks to ts dscretzaton. Indeed, t can easly grow voxels nto small concave parts of the boundary. Fgure 4: D slces of 3D nosy mages (top and surface results obtaned wth the lne search method (bottom

7 Dynamc Neghbourhoods n Actve Surfaces for 3D Segmentaton 179 Table 1 Comparson of Completon Tmes between Classcal Greedy Algorthm, Lne Search Method (LS, Shfted Neghborhood Method (SN and Fast Level Set on Synthetc Data Sets Image Neghborhood Method # teratons Tme (s Hmean w=3 Greedy Spral LS ( w=5 Greedy voxels LS SN w=7 Greedy LS SN Fast level set ellpsods w=3 Greedy ( LS voxels Greedy w=5 LS SN w=7 Greedy LS SN Fast level set Vase w=3 Greedy ( LS voxels w=5 Greedy LS SN w=7 Greedy LS SN Fast level set Table Comparson of Completon Tmes between Classcal Greedy Algorthm, lne Search Method (LS, Shfted Neghborhood Method (SN and Fast Level Set on CT (Computed Tomography Data Sets Image Neghborhood Method # teratons Tme (s Hmean CT 1 w=3 Greedy ( LS voxels w=5 Greedy LS SN w=7 Greedy LS SN Fast level set CT w=3 Greedy ( LS voxels w=5 Greedy LS SN w=7 Greedy LS SN Fast level set CT 3 w=3 Greedy ( LS voxels w=5 Greedy LS SN w=7 Greedy LS SN Fast level set CONCLUSION AND FUTURE WORK In ths artcle we have descrbed two acceleraton methods for actve surfaces evolvng wth the greedy algorthm. The frst one s based on a smart orentaton of the neghbourhood grd of each vertex regardng the drectons followed n the precedng teratons. The second one uses the same drecton nformaton to make each vertex search for a better poston along an exploraton lne. The man applcaton doman of our methods s tme-dependent segmentaton wth a pror knowledge about the topology of the object, such as 3D vdeos. We compared our two acceleraton methods wth a recent fast mplct mplementaton of actve contour based on the level-set approach. As shown n table 1 and, our methods allowed us to accelerate the greedy algorthm for actve surfaces. Meshbased methods (the greedy algorthm and our acceleraton methods leads to performances turnng out to be far beyond the level set-based method ones. Acceleratons methods tend to make completon tme fall below 1s (for the best confguraton whereas the fast level set approach exceeds 30s. The best acceleraton method for D actve contours was the shfted neghborhood but we detected the lne search as the best to be appled on 3D actve surfaces. The explcaton s that n three dmensons an exploraton lne stays a lne whereas a square neghborhood becomes cubc, rsng the completon tmes added by the shfted neghborhood method. We also tested the Deformed Neghborhood method dscrbed n [10] but t was not ecent on actve surfaces for the same reasons. We can also notce that the contrbuton of the shfted neghborhood method s better wth a large neghborhood. Indeed, the shftngs are dependent of ts sze. We are developng an hybrd model based on the shfted neghborhood method and the physc-based approach of parametrc actve contours [4]. The man dea s to use the nformaton of the force vector to drect the shft of the neghborhood grd. We also plan to upgrade the performances of our methods by usng a mult-resoluton approach. REFERENCES [1] Kass, M., Wtkn, A., Terzopoulos, D.: Snakes: actve contour models. Internatonal Journal of Computer Vson 1 (1988, [] Montagnat, J., Delngette, H., Ayache, N.: A revew of deformable surfaces: topology, geometry, and deformaton. Image and Vson Computng 19 (001, [3] Mlle, J., Boné, R., Makrs, P., Cardot, H.: 3D segmentaton usng actve surface: a survey and a new model. In: 5th Int. Conf. on Vsualzaton, Imagng & Image Processng (VIIP, Bendorm, Span (005, [4] Zhang, Z., Braun, F.: Fully 3D actve surface models wth selfnflaton and self-deflaton forces. In: IEEE Computer Vson and Pattern Recognton (CVPR, San Juan, Puerto Rco (1997, [5] Wllams, D., Shah, M.: A fast algorthm for actve contours and curvature estmaton. Computer Vson, Graphcs, and Image Processng: Image Understandng 55 (199, 14 6.

8 180 Internatonal Journal of Computatonal Vson and Bomechancs [6] Julen Olver, Julen Mlle, Romuald Boné, Jean-Jacques Rousselle 6. Bulptt, A., Efford, N.: An effcent 3D deformable model wth a self-optmsng mesh. Image and Vson Computng 14 (1996, [7] Mallad, R., Sethan, J., Vemur, B.: Shape modelng wth front propagaton: a level set approach. IEEE Transactons on Pattern Analyss and Machne Intellgence 17 (1995, [8] Adalstensson, D., Sethan, J.: A fast level set method for propagatng nterfaces. Journal of Computatonal Physcs 118 (1995, [9] Sh, Y., Karl, W.: A fast level set method wthout solvng PDEs. In: IEEE Internatonal Conference on Acoustcs, Speech, and Sgnal Processng (ICASSP. Volume., Phladelpha, USA (005, [10] Olver, J., Boné, R., Rousselle, J.J.: Comparson of actve contour acceleraton methods. In: 5th Int. Conf. on Vsualzaton, Imagng & Image Processng (VIIP, Bendorm, Span (005. [11] Olver, J., Mlle, J., Boné, R., Rousselle, J.J.: Actve surfaces acceleraton methods. In: Computatonal Modellng of Objects Represented n Images: fundamentals, methods and applcatons (CompIMAGE, Combra, Portugal (006. [1] Mlle, J., Boné, R., Makrs, P., Cardot, H.: Greedy algorthm and physcs-based method for actve contours and surfaces: a comparatve study. In: IEEE Internatonal Conference on Image Processng (ICIP, Atlanta, USA (006, [13] McInerney, T., Terzopoulos, D.: A dynamc fnte element surface model for segmentaton and trackng n multdmensonal medcal mages wth applcaton to cardac 4D mage analyss. Computerzed Medcal Imagng and Graphcs 19 (1995, [14] Park, J.Y., McInerney, T., Terzopoulos, D.: A non-selfntersectng adaptve deformable surface for complex boundary extracton from volumetrc mages. Computer & Graphcs 5 (001, [15] Amn, A., Weymouth, T., Ran, R.: Usng dynamc programmng for solvng varatonal problem n vson. IEEE Transactons on Pattern Analyss and Machne Intellgence 1 (1991, [16] J, L., Yan, H.: Attractable snakes based on the greedy algorthm for contour extracton. Pattern Recognton 35 (00, [17] Lam, C., Yuen, S.: An unbased actve contour algorthm for object trackng. Pattern Recognton Letters 19 (1998, [18] Zucker, S., Hummel, R.: A three-dmensonal edge operator. IEEE Transactons on Pattern Analyss and Machne Intellgence 3 (1981, [19] Cohen, L.: On actve contour models and balloons. Computer Vson, Graphcs, and Image Processng: Image Understandng 53 (1991, [0] Lachaud, J., Montanvert, A.: Deformable meshes wth automated topology changes for coarse-to-fne threedmensonal surface extracton. Medcal Image Analyss 3 (1999, [1] Slabaugh, G., Unal, G.: Actve polyhedron: surface evoluton theory appled to deformable meshes. In: IEEE Computer Vson and Pattern Recognton (CVPR. Volume., San Dego, USA ( [] Osher, S., Sethan, J.: Fronts propagaton wth curvaturedependent speed: algorthms based on hamlton-jacob formulatons. Journal of Computatonal Physcs 79 (1988, [3] Dubusson, M. P., Jan, A.: A modfed Hausdor dstance for object matchng. In: 1th Internatonal Conference on Pattern Recognton (ICPR, Jerusalem, Israel (1994, [4] Lobregt, S., Vergever, M.: A dscrete dynamc contour model. IEEE Transactons on Medcal Imagng 14 (1995, 1 4. [5] L. Cohen and I. Cohen, Fnte element methods for models and balloons for D and 3D mages, IEEE Transactons on Pattern analyss Machne Intellgence, 15(1993,

Active Contours/Snakes

Active Contours/Snakes Actve Contours/Snakes Erkut Erdem Acknowledgement: The sldes are adapted from the sldes prepared by K. Grauman of Unversty of Texas at Austn Fttng: Edges vs. boundares Edges useful sgnal to ndcate occludng

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

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

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

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

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

n i P i φ i t i P i-1 i+1 P i+1 P i i-1 i+1 2 r i

n i P i φ i t i P i-1 i+1 P i+1 P i i-1 i+1 2 r i New Algorthms for Controllng Actve Contours Shape and Topology H. Delngette and J. Montagnat Projet Epdaure I.N.R.I.A. 06902 Sopha-Antpols Cedex, BP 93, France Abstract In recent years, the eld of actve-contour

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

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

Active Contour Models

Active Contour Models Actve Contour Models By Taen Lee A PROJECT submtted to Oregon State Unversty n partal fulfllment of The requrements for the Degree of Master of Scence n Computer Scence Presented September 9 005 Commencement

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

Accounting for the Use of Different Length Scale Factors in x, y and z Directions

Accounting for the Use of Different Length Scale Factors in x, y and z Directions 1 Accountng for the Use of Dfferent Length Scale Factors n x, y and z Drectons Taha Soch (taha.soch@kcl.ac.uk) Imagng Scences & Bomedcal Engneerng, Kng s College London, The Rayne Insttute, St Thomas Hosptal,

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

Support Vector Machines

Support Vector Machines /9/207 MIST.6060 Busness Intellgence and Data Mnng What are Support Vector Machnes? Support Vector Machnes Support Vector Machnes (SVMs) are supervsed learnng technques that analyze data and recognze patterns.

More information

Smoothing Spline ANOVA for variable screening

Smoothing Spline ANOVA for variable screening Smoothng Splne ANOVA for varable screenng a useful tool for metamodels tranng and mult-objectve optmzaton L. Rcco, E. Rgon, A. Turco Outlne RSM Introducton Possble couplng Test case MOO MOO wth Game Theory

More information

MOTION BLUR ESTIMATION AT CORNERS

MOTION BLUR ESTIMATION AT CORNERS Gacomo Boracch and Vncenzo Caglot Dpartmento d Elettronca e Informazone, Poltecnco d Mlano, Va Ponzo, 34/5-20133 MILANO boracch@elet.polm.t, caglot@elet.polm.t Keywords: Abstract: Pont Spread Functon Parameter

More information

Quality Improvement Algorithm for Tetrahedral Mesh Based on Optimal Delaunay Triangulation

Quality Improvement Algorithm for Tetrahedral Mesh Based on Optimal Delaunay Triangulation Intellgent Informaton Management, 013, 5, 191-195 Publshed Onlne November 013 (http://www.scrp.org/journal/m) http://dx.do.org/10.36/m.013.5601 Qualty Improvement Algorthm for Tetrahedral Mesh Based on

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

Lobachevsky State University of Nizhni Novgorod. Polyhedron. Quick Start Guide

Lobachevsky State University of Nizhni Novgorod. Polyhedron. Quick Start Guide Lobachevsky State Unversty of Nzhn Novgorod Polyhedron Quck Start Gude Nzhn Novgorod 2016 Contents Specfcaton of Polyhedron software... 3 Theoretcal background... 4 1. Interface of Polyhedron... 6 1.1.

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

A 3D Laplacian-Driven Parametric Deformable Model

A 3D Laplacian-Driven Parametric Deformable Model A 3D Laplacan-Drven Parametrc Deformable Model Tan Shen 1,2 Xaole Huang 1 Hongsheng L 1 Edward Km 1 Shaotng Zhang 3 Junzhou Huang 4 1 Department of Computer Scence and Engneerng, Lehgh Unversty, USA 2

More information

2x x l. Module 3: Element Properties Lecture 4: Lagrange and Serendipity Elements

2x x l. Module 3: Element Properties Lecture 4: Lagrange and Serendipity Elements Module 3: Element Propertes Lecture : Lagrange and Serendpty Elements 5 In last lecture note, the nterpolaton functons are derved on the bass of assumed polynomal from Pascal s trangle for the fled varable.

More information

A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS

A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS Proceedngs of the Wnter Smulaton Conference M E Kuhl, N M Steger, F B Armstrong, and J A Jones, eds A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS Mark W Brantley Chun-Hung

More information

Simplification of 3D Meshes

Simplification of 3D Meshes Smplfcaton of 3D Meshes Addy Ngan /4/00 Outlne Motvaton Taxonomy of smplfcaton methods Hoppe et al, Mesh optmzaton Hoppe, Progressve meshes Smplfcaton of 3D Meshes 1 Motvaton Hgh detaled meshes becomng

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

Skew Angle Estimation and Correction of Hand Written, Textual and Large areas of Non-Textual Document Images: A Novel Approach

Skew Angle Estimation and Correction of Hand Written, Textual and Large areas of Non-Textual Document Images: A Novel Approach Angle Estmaton and Correcton of Hand Wrtten, Textual and Large areas of Non-Textual Document Images: A Novel Approach D.R.Ramesh Babu Pyush M Kumat Mahesh D Dhannawat PES Insttute of Technology Research

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

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

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

Chapter 6 Programmng the fnte element method Inow turn to the man subject of ths book: The mplementaton of the fnte element algorthm n computer programs. In order to make my dscusson as straghtforward

More information

A Region based Active Contour Approach for Liver CT Image Analysis Driven by Local likelihood Image Fitting Energy

A Region based Active Contour Approach for Liver CT Image Analysis Driven by Local likelihood Image Fitting Energy Internatonal Journal of Engneerng and Advanced Technology (IJEAT) ISSN: 49 8958, Volume-6 Issue-5, June 07 A Regon based Actve Contour Approach for Lver CT Image Analyss Drven by Local lkelhood Image Fttng

More information

The Greedy Method. Outline and Reading. Change Money Problem. Greedy Algorithms. Applications of the Greedy Strategy. The Greedy Method Technique

The Greedy Method. Outline and Reading. Change Money Problem. Greedy Algorithms. Applications of the Greedy Strategy. The Greedy Method Technique //00 :0 AM Outlne and Readng The Greedy Method The Greedy Method Technque (secton.) Fractonal Knapsack Problem (secton..) Task Schedulng (secton..) Mnmum Spannng Trees (secton.) Change Money Problem Greedy

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

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

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

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

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

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

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

A Graphical Model Framework for Coupling MRFs and Deformable Models

A Graphical Model Framework for Coupling MRFs and Deformable Models A Graphcal Model Framework for Couplng MRFs and Deformable Models Ru Huang, Vladmr Pavlovc, and Dmtrs N. Metaxas Dvson of Computer and Informaton Scences, Rutgers Unversty {ruhuang, vladmr, dnm}@cs.rutgers.edu

More information

Application of Learning Machine Methods to 3 D Object Modeling

Application of Learning Machine Methods to 3 D Object Modeling Applcaton of Learnng Machne Methods to 3 D Object Modelng Crstna Garca, and José Al Moreno Laboratoro de Computacón Emergente, Facultades de Cencas e Ingenería, Unversdad Central de Venezuela. Caracas,

More information

Simulation: Solving Dynamic Models ABE 5646 Week 11 Chapter 2, Spring 2010

Simulation: Solving Dynamic Models ABE 5646 Week 11 Chapter 2, Spring 2010 Smulaton: Solvng Dynamc Models ABE 5646 Week Chapter 2, Sprng 200 Week Descrpton Readng Materal Mar 5- Mar 9 Evaluatng [Crop] Models Comparng a model wth data - Graphcal, errors - Measures of agreement

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

A Novel Adaptive Descriptor Algorithm for Ternary Pattern Textures

A Novel Adaptive Descriptor Algorithm for Ternary Pattern Textures A Novel Adaptve Descrptor Algorthm for Ternary Pattern Textures Fahuan Hu 1,2, Guopng Lu 1 *, Zengwen Dong 1 1.School of Mechancal & Electrcal Engneerng, Nanchang Unversty, Nanchang, 330031, Chna; 2. School

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

S.P.H. : A SOLUTION TO AVOID USING EROSION CRITERION?

S.P.H. : A SOLUTION TO AVOID USING EROSION CRITERION? S.P.H. : A SOLUTION TO AVOID USING EROSION CRITERION? Célne GALLET ENSICA 1 place Emle Bloun 31056 TOULOUSE CEDEX e-mal :cgallet@ensca.fr Jean Luc LACOME DYNALIS Immeuble AEROPOLE - Bat 1 5, Avenue Albert

More information

Fair Triangle Mesh Generation with Discrete Elastica

Fair Triangle Mesh Generation with Discrete Elastica Far Trangle Mesh Generaton wth Dscrete Elastca Shn Yoshzawa, and Alexander G. Belyaev, Computer Graphcs Group, Max-Planck-Insttut für Informatk, 66123 Saarbrücken, Germany Phone: [+49](681)9325-414 Fax:

More information

CS 534: Computer Vision Model Fitting

CS 534: Computer Vision Model Fitting CS 534: Computer Vson Model Fttng Sprng 004 Ahmed Elgammal Dept of Computer Scence CS 534 Model Fttng - 1 Outlnes Model fttng s mportant Least-squares fttng Maxmum lkelhood estmaton MAP estmaton Robust

More information

NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS

NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS ARPN Journal of Engneerng and Appled Scences 006-017 Asan Research Publshng Network (ARPN). All rghts reserved. NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS Igor Grgoryev, Svetlana

More information

Learning the Kernel Parameters in Kernel Minimum Distance Classifier

Learning the Kernel Parameters in Kernel Minimum Distance Classifier Learnng the Kernel Parameters n Kernel Mnmum Dstance Classfer Daoqang Zhang 1,, Songcan Chen and Zh-Hua Zhou 1* 1 Natonal Laboratory for Novel Software Technology Nanjng Unversty, Nanjng 193, Chna Department

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

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

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

A Modified Median Filter for the Removal of Impulse Noise Based on the Support Vector Machines

A Modified Median Filter for the Removal of Impulse Noise Based on the Support Vector Machines A Modfed Medan Flter for the Removal of Impulse Nose Based on the Support Vector Machnes H. GOMEZ-MORENO, S. MALDONADO-BASCON, F. LOPEZ-FERRERAS, M. UTRILLA- MANSO AND P. GIL-JIMENEZ Departamento de Teoría

More information

Very simple computational domains can be discretized using boundary-fitted structured meshes (also called grids)

Very simple computational domains can be discretized using boundary-fitted structured meshes (also called grids) Structured meshes Very smple computatonal domans can be dscretzed usng boundary-ftted structured meshes (also called grds) The grd lnes of a Cartesan mesh are parallel to one another Structured meshes

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

A Gradient Difference based Technique for Video Text Detection

A Gradient Difference based Technique for Video Text Detection A Gradent Dfference based Technque for Vdeo Text Detecton Palaahnakote Shvakumara, Trung Quy Phan and Chew Lm Tan School of Computng, Natonal Unversty of Sngapore {shva, phanquyt, tancl }@comp.nus.edu.sg

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

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

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

A Gradient Difference based Technique for Video Text Detection

A Gradient Difference based Technique for Video Text Detection 2009 10th Internatonal Conference on Document Analyss and Recognton A Gradent Dfference based Technque for Vdeo Text Detecton Palaahnakote Shvakumara, Trung Quy Phan and Chew Lm Tan School of Computng,

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

Distance Calculation from Single Optical Image

Distance Calculation from Single Optical Image 17 Internatonal Conference on Mathematcs, Modellng and Smulaton Technologes and Applcatons (MMSTA 17) ISBN: 978-1-6595-53-8 Dstance Calculaton from Sngle Optcal Image Xao-yng DUAN 1,, Yang-je WEI 1,,*

More information

Hierarchical clustering for gene expression data analysis

Hierarchical clustering for gene expression data analysis Herarchcal clusterng for gene expresson data analyss Gorgo Valentn e-mal: valentn@ds.unm.t Clusterng of Mcroarray Data. Clusterng of gene expresson profles (rows) => dscovery of co-regulated and functonally

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

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

Wavefront Reconstructor

Wavefront Reconstructor A Dstrbuted Smplex B-Splne Based Wavefront Reconstructor Coen de Vsser and Mchel Verhaegen 14-12-201212 2012 Delft Unversty of Technology Contents Introducton Wavefront reconstructon usng Smplex B-Splnes

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

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

Dynamic wetting property investigation of AFM tips in micro/nanoscale

Dynamic wetting property investigation of AFM tips in micro/nanoscale Dynamc wettng property nvestgaton of AFM tps n mcro/nanoscale The wettng propertes of AFM probe tps are of concern n AFM tp related force measurement, fabrcaton, and manpulaton technques, such as dp-pen

More information

Nonlocal Mumford-Shah Model for Image Segmentation

Nonlocal Mumford-Shah Model for Image Segmentation for Image Segmentaton 1 College of Informaton Engneerng, Qngdao Unversty, Qngdao, 266000,Chna E-mal:ccluxaoq@163.com ebo e 23 College of Informaton Engneerng, Qngdao Unversty, Qngdao, 266000,Chna E-mal:

More information

Monte Carlo Rendering

Monte Carlo Rendering Monte Carlo Renderng Last Tme? Modern Graphcs Hardware Cg Programmng Language Gouraud Shadng vs. Phong Normal Interpolaton Bump, Dsplacement, & Envronment Mappng Cg Examples G P R T F P D Today Does Ray

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

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

Corner-Based Image Alignment using Pyramid Structure with Gradient Vector Similarity

Corner-Based Image Alignment using Pyramid Structure with Gradient Vector Similarity Journal of Sgnal and Informaton Processng, 013, 4, 114-119 do:10.436/jsp.013.43b00 Publshed Onlne August 013 (http://www.scrp.org/journal/jsp) Corner-Based Image Algnment usng Pyramd Structure wth Gradent

More information

Polyhedral Surface Smoothing with Simultaneous Mesh Regularization

Polyhedral Surface Smoothing with Simultaneous Mesh Regularization olyhedral Surface Smoothng wth Smultaneous Mesh Regularzaton Yutaka Ohtake The Unversty of Azu Azu-Wakamatsu Cty Fukushma 965-8580 Japan d800@u-azu.ac.jp Alexander G. Belyaev The Unversty of Azu Azu-Wakamatsu

More information

An Entropy-Based Approach to Integrated Information Needs Assessment

An Entropy-Based Approach to Integrated Information Needs Assessment Dstrbuton Statement A: Approved for publc release; dstrbuton s unlmted. An Entropy-Based Approach to ntegrated nformaton Needs Assessment June 8, 2004 Wllam J. Farrell Lockheed Martn Advanced Technology

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

Shape Representation Robust to the Sketching Order Using Distance Map and Direction Histogram

Shape Representation Robust to the Sketching Order Using Distance Map and Direction Histogram Shape Representaton Robust to the Sketchng Order Usng Dstance Map and Drecton Hstogram Department of Computer Scence Yonse Unversty Kwon Yun CONTENTS Revew Topc Proposed Method System Overvew Sketch Normalzaton

More information

6.854 Advanced Algorithms Petar Maymounkov Problem Set 11 (November 23, 2005) With: Benjamin Rossman, Oren Weimann, and Pouya Kheradpour

6.854 Advanced Algorithms Petar Maymounkov Problem Set 11 (November 23, 2005) With: Benjamin Rossman, Oren Weimann, and Pouya Kheradpour 6.854 Advanced Algorthms Petar Maymounkov Problem Set 11 (November 23, 2005) Wth: Benjamn Rossman, Oren Wemann, and Pouya Kheradpour Problem 1. We reduce vertex cover to MAX-SAT wth weghts, such that the

More information

Programming in Fortran 90 : 2017/2018

Programming in Fortran 90 : 2017/2018 Programmng n Fortran 90 : 2017/2018 Programmng n Fortran 90 : 2017/2018 Exercse 1 : Evaluaton of functon dependng on nput Wrte a program who evaluate the functon f (x,y) for any two user specfed values

More information

Snake-Based Segmentation of Teeth from Virtual Dental Casts

Snake-Based Segmentation of Teeth from Virtual Dental Casts Computer-Aded Desgn and Applcatons ISSN: (Prnt) 1686-4360 (Onlne) Journal homepage: http://www.tandfonlne.com/lo/tcad20 Snake-Based Segmentaton of Teeth from Vrtual Dental Casts Thomas Kronfeld, Davd Brunner

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

Mathematics 256 a course in differential equations for engineering students

Mathematics 256 a course in differential equations for engineering students Mathematcs 56 a course n dfferental equatons for engneerng students Chapter 5. More effcent methods of numercal soluton Euler s method s qute neffcent. Because the error s essentally proportonal to the

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

Visual Curvature. 1. Introduction. y C. IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), June 2007

Visual Curvature. 1. Introduction. y C. IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), June 2007 IEEE onf. on omputer Vson and Pattern Recognton (VPR June 7 Vsual urvature HaRong Lu, Longn Jan Lateck, WenYu Lu, Xang Ba HuaZhong Unversty of Scence and Technology, P.R. hna Temple Unversty, US lhrbss@gmal.com,

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

METRIC ALIGNMENT OF LASER RANGE SCANS AND CALIBRATED IMAGES USING LINEAR STRUCTURES

METRIC ALIGNMENT OF LASER RANGE SCANS AND CALIBRATED IMAGES USING LINEAR STRUCTURES METRIC ALIGNMENT OF LASER RANGE SCANS AND CALIBRATED IMAGES USING LINEAR STRUCTURES Lorenzo Sorg CIRA the Italan Aerospace Research Centre Computer Vson and Vrtual Realty Lab. Outlne Work goal Work motvaton

More information

Electrical analysis of light-weight, triangular weave reflector antennas

Electrical analysis of light-weight, triangular weave reflector antennas Electrcal analyss of lght-weght, trangular weave reflector antennas Knud Pontoppdan TICRA Laederstraede 34 DK-121 Copenhagen K Denmark Emal: kp@tcra.com INTRODUCTION The new lght-weght reflector antenna

More information

Multi-view 3D Position Estimation of Sports Players

Multi-view 3D Position Estimation of Sports Players Mult-vew 3D Poston Estmaton of Sports Players Robbe Vos and Wlle Brnk Appled Mathematcs Department of Mathematcal Scences Unversty of Stellenbosch, South Afrca Emal: vosrobbe@gmal.com Abstract The problem

More information

Problem Set 3 Solutions

Problem Set 3 Solutions Introducton to Algorthms October 4, 2002 Massachusetts Insttute of Technology 6046J/18410J Professors Erk Demane and Shaf Goldwasser Handout 14 Problem Set 3 Solutons (Exercses were not to be turned n,

More information

A Hierarchical Deformable Model Using Statistical and Geometric Information

A Hierarchical Deformable Model Using Statistical and Geometric Information A Herarchcal Deformable Model Usng Statstcal and Geometrc Informaton Dnggang Shen 3 and Chrstos Davatzkos Department of adology Department of Computer Scence 3 Center for Computer-Integrated Surgcal Systems

More information

Geodesic Active Regions for Supervised Texture Segmentation

Geodesic Active Regions for Supervised Texture Segmentation Geodesc Actve egons for Supervsed Texture Segmentaton Nkos Paragos achd Derche INIA BP 9, 00, oute des Lucoles 0690 Sopha Antpols Cedex, France e-mal: {nparago,der}@sophanrafr Abstract Ths paper presents

More information

A Volumetric Approach for Interactive 3D Modeling

A Volumetric Approach for Interactive 3D Modeling A Volumetrc Approach for Interactve 3D Modelng Dragan Tubć Patrck Hébert Computer Vson and Systems Laboratory Laval Unversty, Ste-Foy, Québec, Canada, G1K 7P4 Dens Laurendeau E-mal: (tdragan, hebert, laurendeau)@gel.ulaval.ca

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

Boundary Detection Using Open Spline Curve Based on Mumford-Shah Model

Boundary Detection Using Open Spline Curve Based on Mumford-Shah Model Vol. 35, o. 2 ACTA AUTOMATICA SIICA February, 2009 Boundary Detecton Usng Open Splne Curve Based on Mumford-Shah Model LI ao-mao 1, 2 ZHU Ln-Ln 1, 2 TAG Yan-Dong 1 Abstract Inspred by Cremers s work, ths

More information

Reducing Frame Rate for Object Tracking

Reducing Frame Rate for Object Tracking Reducng Frame Rate for Object Trackng Pavel Korshunov 1 and We Tsang Oo 2 1 Natonal Unversty of Sngapore, Sngapore 11977, pavelkor@comp.nus.edu.sg 2 Natonal Unversty of Sngapore, Sngapore 11977, oowt@comp.nus.edu.sg

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

Wishing you all a Total Quality New Year!

Wishing you all a Total Quality New Year! Total Qualty Management and Sx Sgma Post Graduate Program 214-15 Sesson 4 Vnay Kumar Kalakband Assstant Professor Operatons & Systems Area 1 Wshng you all a Total Qualty New Year! Hope you acheve Sx sgma

More information