Object centered stereo: displacement map estimation using texture and shading

Size: px
Start display at page:

Download "Object centered stereo: displacement map estimation using texture and shading"

Transcription

1 Object centered stereo: dsplacement map estmaton usng texture and shadng Nel Brkbeck Dana Cobzas Martn Jagersand Computng Scence Unversty of Alberta Edmonton, AB T6G2E8, Canada Abstract Input Data Output Data We consder the problem of recoverng 3D surface dsplacements usng both shadng and mult-vew stereo cues. In contrast to tradtonal dsparty or depth map representatons, the object centered dsplacement map representaton enables the recovery of complete 3D objects whle also ensurng the reconstructon s not based towards a partcular mage. Although dsplacement mappng requres a base surface, ths base mesh s easly obtaned usng tradtonal computer vson technques e.g., shape-from-slhouette or structure-from-moton. Our method explots shadng varaton due to object rotaton relatve to the lght source, allowng the recovery of dsplacements n both textured and textureless regons n a common framework. In partcular, shadng cues are ntegrated nto a mult-vew stereo photoconsstency functon through the surface normals that are mpled by the dsplacement map. The analytc gradent of ths photo-consstency functon s used to drve a multresoluton conjugate gradent optmzaton. We demonstrate the geometrc qualty of the reconstructed dsplacements on several example objects ncludng a human face. 1. Introducton The automatc computaton of 3D geometrc and appearance models from mages s one of the most challengng and fundamental problems n computer vson. Whle a more tradtonal pont-based method provdes accurate results for camera geometry, a surface representaton s requred for modelng and vsualzaton applcatons. One of the most popular classes of surface-based reconstructon methods are mult-vew stereo approaches that represent the surface as a depth or dsparty map wth respect to one reference mage [17] or multple mages [11]. One of the man dsadvantage of those methods, referred to as mage-centered approaches, s that the reconstructon wll be based by the chosen reference mage. As a consequence, the results are often alased due to the lmted depth resoluton and they do not reconstruct parts of the object that were occluded n the reference mage. An object-centered model s therefore more sutable for mult-vew reconstructon. Examples of Input Images & Calbraton Base Mesh Dsplacement Albedo Fgure 1: An overvew of our method, whch recovers the surface dsplacement and albedo gven a set of nput mages, a base mesh, and calbraton data. object-centered surface representatons nclude voxels [18], level-sets [8], and meshes [10, 7]. We propose a surface reconstructon method that nvestgates a less explored object-centered representaton - a depth feld regstered wth a base mesh see Fg. 1. Our model s nspred by computer graphcs dsplacement maps [5] that nowadays have effcent HW mplementatons e.g., [15]. Compared to a deformable mesh-based representaton, our method s more stable as dspartes are constraned to move along the normal drectons w.r.t. the base mesh. For a purely mesh-based representaton, the vertexes are allowed to move freely, so the dsplacement drecton can become unstable. The method can also be regarded as an extenson of the stereo-based methods, where depth s calculated wth respect to an object-centered base mesh as opposed to the mage plane. Therefore reconstructon methods used for stereo can be easly generalzed to our representaton. One can, for example, make use of the effcent dscrete methods lke graph cuts or belef propagaton that cannot be used wth some other object-centered representatons e.g., mesh, level-sets. For our case, as the cost functon s based on shadng and uses the surface normals, we chose a contnuous representaton where the normals are mpled by the surface. We formulate the surface reconstructon of Lambertan scenes as an optmzaton of a photo-consstency functon that ntegrates stereo cues for textured regons wth shape 1

2 from shadng cues for texture-less regons. The cost functon s calculated over a dscretzaton of the base mesh. Even though the method assumes an a-pror base mesh, ths can be easly obtaned n practce usng shape-fromslhouette or trangulaton of structure-from-moton ponts. Due to the hgh dmensonalty, reconstructon can be dffcult and slow, whle requrng a substantal amount of mage data. To amelorate these problems, we propose a multresoluton algorthm. There exst other approaches that combne stereo for textured regons wth shape from shadng cues for texture-less regons [10, 14], but, n those works, the two scores are separate terms n the cost functon and the combnaton s acheved ether usng weghts [10] or by parttonng the surface nto regons [14]. Lke photometrc stereo, our method s able to reconstruct the surface of spatally varyng or unform materal objects by assumng that the object s movng relatve to the lght source. A smlar constrant was used by Zhang et al. [25] and Weber et al. [22] but wth a dfferent surface representaton. To summarze, the man contrbutons of the paper are: We propose a method that reconstructs surfaces dscretzed as a depth feld wth respect to a base mesh dsplacement map; the representaton s sutable for both closed and open surfaces and, unlke tradtonal stereo, reconstructs whole objects; We desgned a photo-consstency functon sutable for surfaces wth textured and unform Lambertan reflectance by ntegratng shadng cues on mpled surface normals; We desgned a practcal setup that provdes the necessary lght varaton, camera and lght calbraton and requres only commonly avalable hardware: a lght source, a camera, and a glossy whte sphere. 2 Related Work To our knowledge, the work of Vogatzs et al. [21] s the only source that deals wth reconstructon of depth from a base mesh. Most prevous approaches reconstruct depth from stereo and then ft planes to the resultng 3D ponts [6]. In contrast, Vogatzs et al. [21] estmate the dsplacement for sample ponts on the base mesh usng a belef propagaton technque. Ther method assumes there s no llumnaton varaton n the mages the scene s fxed w.r.t. llumnaton and the cost functon s just a varance of the colors that a sample pont project to. The method s therefore smlar to mult-vew stereo reconstructon, requrng scenes wth good texture. A related representaton was proposed by Zeng et al. [23] but, n ther case, the base geometry s a collecton of planes ftted to a set of 3D ponts gven a-pror that do not necessarly form a mesh. The depth of the surface patch along the plane normal s reconstructed usng a globally-optmal graph cut optmzaton that ensures that the patch surface goes through any neghborng 3D ponts. A smlar surface patch approach was used to determne voxel consstency n a more recent work by Zheng and hs colleagues [24]. Other approaches use parametrc models for local depth representaton e.g., planar dsks [9], quadratc patches [13]. The patches are then ntegrated and nterpolated to form the fnal surface. The mesh-based methods share some smlartes to the chosen representaton, but they operate by teratvely evolvng and refnng an ntal mesh untl t fts the set of mages [10, 7, 12, 2]. In these approaches large and complex meshes have to be mantaned n order to represent fne detal. In contrast we assume a fxed base mesh and compute surface detal as a heght feld wth respect to t. The chosen representaton s an extenson of the tradtonal dsparty map where the depth s regstered wth respect to a base geometry nstead of the reference mage plane. Thus t has smlartes to the mult-vew stereo reconstructon technques. Lke n the global stereo methods e.g., [20, 11] we formulated the problem as mnmzng a global energy functon that takes nto account both matchng and smoothness costs. But, unlke stereo methods, n our case the dscretzaton and regularzaton s vewpont ndependent. Addtonally, our approach s able to reconstruct complete scenes as every pont that s vsble n at least one mage wll be reconstructed n the case of stereo ponts that are occluded n the reference mage are not reconstructed. 3 Problem Defnton and Formulaton Base Mesh True Surface sample ponts dsplaced sample pont ˆp = p + µ d sample pont p Fgure 2: An overvew of the representaton and notaton used n ths paper. We assume that we are gven a set of n mages, I, the correspondng calbraton matrces, P, the correspondng llumnaton parameters, L, and a base mesh consstng of a set of vertces V = {v R 3 } and a trangulaton of these vertces T = {v 1, v 2, v 3 v V }. The shape reconstructon problem s then cast as an optmzaton problem, 2

3 that recovers a set of dsplacements µ j along a drecton d j from a dscrete set of m sample ponts, p j, taken on the surface of each trangle. A pont on the dsplaced surface wll be denoted as ˆp j = p j + µ j d j, whch s a functon of the current estmate of the dsplacement µ j see Fg. 2. The dsplacements, µ j, are then related to the mage measurements through a photo-consstency functon. The photo-consstency functon measures the smlarty between the reflectance of a pont on the surface and the mages n whch t s observed. The goal s then to fnd a set of dsplacements, µ j, and the reflectance parameters, α j, of the sample ponts that mnmze the value of the followng cost functon: F data = j = j 1 f µ j, α j V j V j 1 1 R ˆp j, α j I ΠP ˆp j 2 V j V j 2 In the above cost functon, V j denotes the set of mages that observe dsplaced pont ˆp j, and R denotes the renderng functon that produces a color value for the dsplaced pont observed under the llumnaton and vewng parameters of camera. In other words, the cost functon expresses the dfference between the rendered surface representaton and the nput mages. We used a smlar cost functon n our mesh-based mplementaton [2] although n ths work we currently take no precautonary measures to flter out specular hghlghts or address mage samplng ssues. Agan, nstead of usng the entre set of cameras that observe a sample pont, a subset of the vsble cameras closest to the medan camera are used. Ths adjustment partally accounts for mage samplng problems e.g., cameras that vew the surface at a grazng vew. We assume that the nput mages and llumnaton nformaton s suffcent for determnng the reflectance parameters α j. That s, f we fx the surface dsplacements, there exsts a closed form soluton for reflectance parameters that mnmze Eq. 4. If ths s the case, then ths mnmzaton process can be done for each α j ndependent of the other ones lke n photometrc stereo. In our specfc work, we assume that the surface s Lambertan, mplyng that the surface reflectance parameters at a pont are smply the albedo e.g., α j s a RGB color. Furthermore, we assume that the llumnaton for each frame s represented as a drectonal lght source plus an ambent term 1. We use l to denote the lght drecton, l to denote the lght color, and a to denote the ambent color. Under these condtons, the renderng functon for mage s smply the Lambertan shadng model for the case of a sngle lght source: 1 Actually, we model the lght as a pont lght source, mplyng that the drecton to the lght source s dependent on the 3D poston of the pont. n j R = α j l l + a where n j s the surface normal at dsplaced pont j. We use the mpled surface normal, obtaned as a functon of the sample ponts that are wthn a neghborhood of sample pont j, denoted N j. Unlke tradtonal bnocular stereo, the use of the mplct surface normal n the above cost functon acts a regularzer, but the mnmzaton s stll senstve to nose. As s the case n other mult-vew stereo reconstructons, addtonal regularzaton s necessary. In ths work, we take the approach often used n bnocular methods, where the regularzaton s added as a separate term n the mnmzaton, gvng 3 F = F data + λf smooth 4 Currently, we use a smple quadratc regularzer that ensures that neghborng sample ponts, have smlar dsplacements: F smooth = µ j µ k 2 5 j m k N j where N j s a neghborhood around sample pont j. Another potental smoothness term would be to use the Eucldean dstance between neghborng ponts, as was done by Vogatzs et al. [21]. An alternatve approach s to keep the regularzaton mplct multplcatve regularzer by consderng the weghted mnmal surface lke n the level set approaches e.g., Faugeras and Kerven [8]. However, researchers e.g., Soatto et al. [19] made the observaton that the mplct level set approach mght suffer from over-smoothness due to hgh order dervatves nvolved nto unknowns and also could change the mage varablty dependng on the surface. Therefore, for our system, we used an addtve regularzer, gvng a more precse and easly nterpretable effect. 4 Samplng & Optmzaton We now detal the procedures requred n mnmzng the cost functon presented n Secton 3. Frst we dscuss how we dscretze the base mesh nto sample ponts, and defne a specfc neghborhood functon for the sample ponts. Wth these detals n place we then present a closed form soluton for the reflectance parameters for a set of surface ponts, followed by detals of the optmzaton procedure. 4.1 Samplng Clearly, t s desrable to have a set of sample ponts that are evenly dstrbuted on the base mesh. The resoluton of the mesh should be chosen so that the dstance between sam- 3

4 ple ponts roughly corresponds to one pxel n mage space. One smple soluton would be to choose the sample ponts on a regular grd wthn each base trangle. However, for vsblty and neghborhood purposes t s useful to have a trangulaton of the base ponts. Unfortunately, obtanng a water tght trangulaton of these regularly spaced ponts may become complcated. Moreover, the samplng rate may be dfferent on the edges of the base mesh. To allevate these samplng ssues we nstead use a subdvson approach. In ths approach, startng from the base mesh and gven a desred sample spacng, any edges n the mesh that are longer than twce the desred sample spacng are splt n two. Splttng an edge turns the two trangles that contan the volatng edge nto four smaller trangles. Ths operaton s performed untl all edges n the mesh are less than twce the sample spacng e.g., splttng an edge would produce two sample ponts that closer than the desred sample spacng. Smply usng ths approach may produce sample ponts wth an rregular valence.e., an rregular number of neghbors and may create trangles wth poor aspect ratos. To crcumvent these problems, we use the topologcal operators of Lachaud and Montanvert [16]. These operators ensure that edge lengths are wthn a certan range and that non-neghborng vertces are suffcently spaced. The vertces of the subdvded mesh are the sample ponts p j, and the neghborhood functon N j s defned as those sample ponts that share an edge n the subdvded mesh. Ths approach ensures that the neghborhood functon extends across base trangles. Fnally, we take the dsplacement drecton d j of a sample pont to be the correspondng nterpolated surface normal at the poston p j on the base mesh. As our current mplementaton does not account for self-ntersecton durng the refnement, usng the nterpolated normal opposed to the base plane normal reduces the occurance of these self-ntersectons. 4.2 Reflectance Parameters Recall that we requred a closed form soluton for the reflectance parameters for any partcular choce of dsplacements. The Lambertan shadng functon n Eq. 3 s dependent on the mpled surface normal, n j, for surface pont j. In ths work, we compute the mpled surface normal, n j, as an area weghted average of the trangle normals of the sample pont trangulaton, whch s gven below n unnormalzed form: n j = ˆp k ˆp j ˆp ˆp j 6 j,k, p j where p j denotes the trangles that contan p j. Notce that the mpled surface normal s a functon of the dsplaced surface ponts. Usng the above defnton of a surface normal for a partcular nstantaton of surface dsplacements, we can compute a closed form soluton for the α j. For each partcular surface pont, we collect all mage observatons of the pont nto a system of equatons: n l j l 1 1 n j + a 1 n l j l 2 2 n j + a 2.. l m n j l m n j + a m α j = I 1 ΠP 1 ˆp j I 2 ΠP 2 ˆp j. I n ΠP n ˆp j 7 A j α j = I j 8 As both the left and rght hand sdes are n 1 vectors, the least squares soluton for the albedo s easly obtaned as: α j = A j I j A j A j 4.3 Optmzaton = A j I j A j A j 9 One possble method for optmzaton would be to dscretze the dsplacement values nto a set of dscrete labels and then use a combnatoral optmzaton technque to refne the labels e.g., smlar to the method of Vogatzs et al. [21]. As the surface normals rely on a current estmate of the neghborng dsplacements, ths approach can not drectly mnmze our cost functon see Secton 4.4.1, for more detals about a drect dscretzaton of the cost functon. Instead, we perform a contnuous optmzaton of the dsplacements. As both the memory and computatonal cost for computng a Hessan of the objectve functon s prohbtve, we chose the conjugate gradent method. We use the analytc dervatves of the cost functon n the optmzaton. These dervatves are summarzed below. For convenence we defne the ndvdual terms of the cost metrc for a gven sample pont n a gven vew: c j = 2 n j l α j l + a I ΠP ˆp j The value of the objectve functon for sample pont j n mage e.g., c j s dependent only on ts dsplacement, µ j, and the dsplacements of the neghbors N j p j. Therefore, c j = 0 f k j&k / N j p j 10 c j n j l = 2 α j l + a I ΠP ˆp j n j l αj n j l l + a + α j l I P ˆp j The dervatve of the least squares computaton for the 4

5 α j albedo,, s straghtforward and can be expressed n terms of. These dervatons and those for the nj l n j I ΠP ˆp j term can be found n Appendx A. 4.4 Mult-Resoluton The local nature of the conjugate gradent optmzaton s senstve to the startng poston and s lkely to fall nto a local mnma f ths startng poston s far from the global mnmum. Followng other mage-based approaches [10], we use a mult-resoluton to lessen the dependence on a good startng poston. We start the optmzaton on downsampled mages and a correspondng down-sampled mesh. After convergence the mesh resoluton s ncreased and the conjugate descent optmzaton s run agan. After ths step converges the mage resoluton s ncreased. These steps are terated untl the true resoluton for the nput mages has been used. We use OpenGL depth bufferng to compute vsblty of the sample ponts. Ths operaton can be expensve, so the vsblty of sample ponts s only updated every 100 teratons or when the resoluton changes Intalzaton To further avod local mnma and because our method requres a good ntalzaton for convergence we provde an approxmate ntalzaton at the lowest resoluton. Specfcally, we use a samplng technque that s smlar to the other dsplacement recovery methods [21]. That s, the cost functon s evaluated at a dscrete set of dsplacements for each sample pont ndependently. There s no noton of a current surface n ths approach, so we must approxmate some measurements. Frst, we approxmate vsblty durng ths stage by usng the vsblty of the base mesh. Furthermore, each dscrete sample of the cost functon for each sample pont requres a surface normal. As there s no mpled surface at ths stage, the best we can do s assume that the normal can be arbtrary. Therefore, for each sample pont and each dscrete depth label we must ft a surface normal that reduces the F data cost. The resdual of ths fttng s the approxmate cost functon. The optmzaton problem s now dscrete, wth the goal of recoverng a set of depth labels that reduce the approxmated cost functon. Ths sampled cost functon s easly ncorporated wth the smoothness term and optmzed usng exstng combnatoral optmzaton technques, such as graphcuts or loopy belef propagaton. In our case, we adapted the expanson graph-cut proposed by Boykov et al. [3] for the dsplacement map representaton the orgnal algorthm estmates a dsparty map. To summarze ths secton, our algorthm takes a low resoluton base mesh along wth the mages and calbraton data as nput. The downsampled mages are created, and the correspondng low resoluton sample ponts are obtaned Fgure 3: An mage of the capture setup used n the experments. The lght s placed nearby the camera. We used a calbraton pattern to calbrate the camera and a whte specular sphere to calbrate the lght poston. usng the subdvson method that was outlned n Secton 4.1. The sample dsplacements on the low resoluton mesh are then ntalzed wth the method outlned above Secton Fnally, the mult-resoluton optmzaton s performed. The algorthm outputs the recovered dsplacements and the albedo at each sample pont on the base mesh. 5 System We now dscuss the practcal elements requred n provdng the nput to our system. Recall that we need a base mesh, mage calbraton matrces, and a lght drecton for each frame. Furthermore, the lght drectons need to vary suffcently durng the capture to ensure that the reflectance parameters can be ft to each dsplaced surface pont. To provde these requrements we use a turntable based setup, whch s llumnated by a regular desk lamp located by the camera see Fg. 3. Wthn ths setup we use a planar dot-based pattern for the calbraton [1]. A blue-screenng approach s used to obtan slhouettes that are then used as nput to shape-from-slhouette to provde an ntal shape. Other computer vson technques for sparse structure could be substtuted for shape-from-slhouette. As shape-fromslhouette may gve a farly dense mesh, we then use the smplfcaton method of Cohen-Stener et al. to obtan a low polygon base mesh [4]. Notce that the camera s fxed relatve to the lght postons, mplyng that the lght s actually movng wth respect to the turntable coordnate system. To avod shadows the lght source s postoned near the camera, and we capture two full rotatons at dfferent heghts to ensure that there s suffcent lght varaton. A glossy whte sphere rotates on the turntable and s used to calbrate the the exact poston of the lght source.e., our desk lamp cannot be postoned ex- 5

6 actly at the camera center as well as to estmate the color of the source. Several spheres are not necessary as the rotaton of the sphere gves several temporal vews. We trangulate the poston of the lght source usng the specular hghlght on the sphere. From ths pont lght source we can obtan the drecton to the source for any 3D pont n the scene to use n the renderng functon. Assumng that the sphere s whte, after we know the poston of the sphere and the poston of the lght source, each non-specular pxel observaton of the sphere gves an equaton n the 2 unknowns ambent lght and lght source color for each color channel. 6 Experments In a frst experment we demonstrate that the method s accurate on synthetc objects wth varyng degrees of texture Fgure 4. We have smulated the capture setup usng a synthetc object texture mapped wth two dfferent textures. In each case the concavtes of the object were accurately recovered. We have also computed the dstance from the recovered dsplacements to the ground truth dsplacements we used ray tracng to recover the ground truth dsplacements from a trangulated mesh. A color mapped verson of ths dstance on the ground truth s also portrayed n Fgure 4. The reconstructon s qute good wth an average geometrc error s about unts on an object of sze 2 unts 1.6% accuracy. The results of the refnement on two real objects are shown n Fgure 5 2. Notce that n each case the base mesh and the ntal dsplacements usng the method descrbed n Secton are coarse approxmatons of the true object. After the dsplacements are refned, the fne scale detal of the 3D geometry becomes apparent. Notce the method performs well on the face sequence, whch would be consdered hard to capture wth tradtonal methods. The face also exhbts a surface reflectance that devates from the Lambertan model, but our Lambertan cost functon s successful at recoverng vsually appealng dsplacements. 7 Conclusons We have presented a method for reconstructng the dsplacements from a base mesh usng mage data. The method couples the surface normals n the cost functon, mplyng that surface shadng cues nfluence the reconstructon. The experments valdated the effectveness of ths approach on objects wth unform and non-unform Lambertan surface propertes. Expermental evdence also suggests that our method s capable of dealng wth slght reflectance devatons from the Lambertan model e.g., the face example. 2 The human head data set was obtaned by placng the calbraton pattern upsde down on the subject s head whle the subject rotated on an offce char. Currently, one of the man lmtatons of ths approach s the nfluence that the base mesh has on the fnal results. For example, the recovered mesh s restrcted to have the same topology as the base mesh. Also, t s possble that the true surface cannot be approxmated by dsplacement mappng the base mesh. In future work, we would lke to address these problems by refnng the base mesh durng the optmzaton. A Gradent Terms I ΠP ˆp j = 0 for k j 11 P = p 00 p 10 p 01 p 11 p 02 p 12 p 03 p 13 p 20 p 21 p 22 p I ΠP ˆp j = I ΠP ˆp j ΠP ˆp j [ ] p = I 00 d j,x + p 01d j,y + p 02d j,z /w p 10 d j,x + p 11 d j,y + p 12 d j,z/w [ up I 20 d j,x + p 21 d j,y + p 22 d ] j,z/w 2 vp 20d j,x + p 21d j,y + p 22d j,z /w 2 wth [u, v, w] = P ˆp j. In practce a Gaussan blurred verson of the mage gradent s substtuted for I. nj l = 1 n j l 1 + n j l nj n j Recall that: n j = = µ k n j l j,k, p j n j l 3 µ k n j n j ˆp k ˆp j ˆp ˆp j Lettng e 1 = ˆp k ˆp j = x k x j, y k y j, z k z j and e 2 = ˆp ˆp j = x x j, y y j, z z j, e 1 e 2 = e 1ye 2z e 2y e 1z e 1x e 2z + e 2x e 1z = e 1x e 2y e 2x e 1y y k z y k z j y j z y z k + y z j + y j z k x k z x k z j x j z x z k + x z j + x j z k x k y x k y j x j y x y k + x y j + x j y k Therefore µj n j = j,k, p j y y k d j,z + z k z d j,y z z k d j,x + x k x d j,z x x k d j,y + y k y d j,x 6

7 Fgure 4: A synthetc object wth dfferent textures used n the experments. From left to rght: an nput mage, the recovered shaded model, the recovered textured model, and dstance to ground truth dark means close, whte means far. The average dstance of the recovered dsplacement to the ground truth was and for the untextured and textured sequences respectvely the sze of the whole object s around 2 unts. The dervaton of µk n j and µ n j are smlar. Fnally, usng the defnton of A j,i j and α j from Equatons 8 and 9, we have α j = 1 A j A j A j I j A j A j 2 A j I j A j A j A j I j = = I ΠP ˆp j l A j A j = = nj l n j l + l n j l l + a I ΠP ˆp j n j l l n j l 2l l + a References + a I ΠP ˆp j nj l + a 2 [1] Adam Baumberg, Alex Lyons, and Rchard Taylor. 3D S.O.M. - a commercal software soluton to 3d scannng. In Vson, Vdeo, and Graphcs VVG 03, pages 41 48, July [2] Nel Brkbeck, Dana Cobzas, Peter Sturm, and Martn Jagersand. Varatonal shape and reflectance estmaton under changng lght and vewponts. In ECCV2006, [3] Yur Boykov, Olga Veksler, and Ramn Zabh. Fast approxmate energy mnmzaton va graph cuts. IEEE Trans. Pattern Anal. Mach. Intell., 2311: , [4] Davd Cohen-Stener, Perre Allez, and Matheu Desbrun. Varatonal shape approxmaton. ACM Trans. Graph., 233: , [5] Robert L. Cook. Shade trees. In SIGGRAPH 84: Proceedngs of the 11th annual conference on Computer graphcs and nteractve technques, pages , New York, NY, USA, ACM Press. [6] Paul E. Debevec, Camllo J. Taylor, and Jtendra Malk. Modelng and renderng archtecture from photographs. In SIGGRAPH, [7] Y. Duan, L. Yang, H. Qn, and D. Samaras. Shape reconstructon from 3d and 2d data usng pde-based deformable surfaces. In ECCV, [8] O. Faugeras and R. Kerven. Varatonal prncples, surface evoluton, pde s, level set methods and the stereo problem. IEEE Trans. Image Processng, 73: , [9] P. Fua. Reconstructng complex surfaces from multple stereo vews. In ICCV, pages , [10] P. Fua and Y. Leclerc. Object-centered surface reconstructon: combnng mult-mage stereo shadng. In Image Understandng Workshop, pages ,

8 Fgure 5: The results of the dsplacement map estmaton for a model house made from wood and bark and a human face. From left to rght, an nput mage, the base mesh, the low-resoluton startng pont usng dscrete optmzaton, two shaded vews of the object, and a textured lt renderng of the recovered object. [11] Pau Gargallo and Peter Sturm. Bayesan 3d modelng from mages usng multple depth maps. In CVPR, pages , [12] C. Hernández and F. Schmtt. Slhouette and stereo fuson for 3d object modelng. Computer Vson and Image Understandng, specal ssue on Model-based and magebased 3D Scene Representaton for Interactve Vsualzaton, 963: , December [13] Wllam A. Hoff and Narendra Ahuja. Surfaces from stereo: Integratng feature matchng, dsparty estmaton, and contour detecton. IEEE Trans. Pattern Anal. Mach. Intell., 112: , [14] H. Jn, A. Yezz, and S. Soatto. Stereoscopc shadng: Integratng shape cues n a varatonal framework. In CVPR, pages , [15] Jan Kautz and Hans-Peter Sedel. Hardware accelerated dsplacement mappng for mage based renderng. In Proccedngs of Graphcs nterface, pages 61 70, [16] J.-O. Lachaud and A. Montanvert. Deformable meshes wth autom. topology changes for coarse-to-fne 3D surface extracton. Medcal Im. Anal., 32: , [17] Danel Scharsten and Rchard Szelsk. A taxonomy and evaluaton of dense two-frame stereo correspondence algorthms. Int. J. Comput. Vson, 471-3:7 42, [18] Gregory G. Slabaugh, W. Bruce Culbertson, Thomas Malzbender, and Ronald W. Schafer. A survey of methods for volumetrc scene reconstructon from photographs. In Internatonal Workshop on Volume Graphcs, [19] S. Soatto, A. Yezz, and H. Jn. Tales of shape and radance n mult-vew stereo. In ICCV, [20] C. Strecha, T. Tuytelaars, and L.J. Van Gool. Dense matchng of multple wde-baselne vews. In ICCV, pages , [21] George Vogatzs, Phlp Torr, Steve Setz, and Roberto Cpolla. Reconstructng relef surfaces. In BMVC, [22] Martn Weber, Andre Blake, and Roberto Cpolla. Towards a complete dense geometrc and photometrc reconstructon under varyng pose and llumnaton. In BMVC, [23] G. Zeng, S. Pars, L. Quan, and M. Lhuller. Surface reconstructon by propagatng 3d stereo data n multple 2d mages. In ECCV, [24] Gang Zeng, Sylvan Pars, Long Quan, and Francos Sllon. Progressve surface reconstructon from mages usng a local pror. In ICCV, pages , [25] L. Zhang, B. Curless, A. Hertzmann, and S. M. Setz. Shape and moton under varyng llumnaton: Unfyng structure from moton, photometrc stereo, and mult-vew stereo. In ICCV,

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

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

Color in OpenGL Polygonal Shading Light Source in OpenGL Material Properties Normal Vectors Phong model

Color in OpenGL Polygonal Shading Light Source in OpenGL Material Properties Normal Vectors Phong model Color n OpenGL Polygonal Shadng Lght Source n OpenGL Materal Propertes Normal Vectors Phong model 2 We know how to rasterze - Gven a 3D trangle and a 3D vewpont, we know whch pxels represent the trangle

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

Angle-Independent 3D Reconstruction. Ji Zhang Mireille Boutin Daniel Aliaga

Angle-Independent 3D Reconstruction. Ji Zhang Mireille Boutin Daniel Aliaga Angle-Independent 3D Reconstructon J Zhang Mrelle Boutn Danel Alaga Goal: Structure from Moton To reconstruct the 3D geometry of a scene from a set of pctures (e.g. a move of the scene pont reconstructon

More information

Interactive Rendering of Translucent Objects

Interactive Rendering of Translucent Objects Interactve Renderng of Translucent Objects Hendrk Lensch Mchael Goesele Phlppe Bekaert Jan Kautz Marcus Magnor Jochen Lang Hans-Peter Sedel 2003 Presented By: Mark Rubelmann Outlne Motvaton Background

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

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

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

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

Outline. Type of Machine Learning. Examples of Application. Unsupervised Learning

Outline. Type of Machine Learning. Examples of Application. Unsupervised Learning Outlne Artfcal Intellgence and ts applcatons Lecture 8 Unsupervsed Learnng Professor Danel Yeung danyeung@eee.org Dr. Patrck Chan patrckchan@eee.org South Chna Unversty of Technology, Chna Introducton

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

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

Implementation of a Dynamic Image-Based Rendering System

Implementation of a Dynamic Image-Based Rendering System Implementaton of a Dynamc Image-Based Renderng System Nklas Bakos, Claes Järvman and Mark Ollla 3 Norrköpng Vsualzaton and Interacton Studo Lnköpng Unversty Abstract Work n dynamc mage based renderng has

More information

Prof. Feng Liu. Spring /24/2017

Prof. Feng Liu. Spring /24/2017 Prof. Feng Lu Sprng 2017 ttp://www.cs.pd.edu/~flu/courses/cs510/ 05/24/2017 Last me Compostng and Mattng 2 oday Vdeo Stablzaton Vdeo stablzaton ppelne 3 Orson Welles, ouc of Evl, 1958 4 Images courtesy

More information

Geometric Primitive Refinement for Structured Light Cameras

Geometric Primitive Refinement for Structured Light Cameras Self Archve Verson Cte ths artcle as: Fuersattel, P., Placht, S., Maer, A. Ress, C - Geometrc Prmtve Refnement for Structured Lght Cameras. Machne Vson and Applcatons 2018) 29: 313. Geometrc Prmtve Refnement

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

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

Unsupervised Learning

Unsupervised Learning Pattern Recognton Lecture 8 Outlne Introducton Unsupervsed Learnng Parametrc VS Non-Parametrc Approach Mxture of Denstes Maxmum-Lkelhood Estmates Clusterng Prof. Danel Yeung School of Computer Scence and

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

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

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

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

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

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

High resolution 3D Tau-p transform by matching pursuit Weiping Cao* and Warren S. Ross, Shearwater GeoServices

High resolution 3D Tau-p transform by matching pursuit Weiping Cao* and Warren S. Ross, Shearwater GeoServices Hgh resoluton 3D Tau-p transform by matchng pursut Wepng Cao* and Warren S. Ross, Shearwater GeoServces Summary The 3D Tau-p transform s of vtal sgnfcance for processng sesmc data acqured wth modern wde

More information

Real-time. Shading of Folded Surfaces

Real-time. Shading of Folded Surfaces Rhensche Fredrch-Wlhelms-Unverstät Bonn Insttute of Computer Scence II Computer Graphcs Real-tme Shadng of Folded Surfaces B. Ganster, R. Klen, M. Sattler, R. Sarlette Motvaton http://www www.vrtualtryon.de

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

Surface Mapping One. CS7GV3 Real-time Rendering

Surface Mapping One. CS7GV3 Real-time Rendering Surface Mappng One CS7GV3 Real-tme Renderng Textures Add complexty to scenes wthout addtonal geometry Textures store ths nformaton, can be any dmenson Many dfferent types: Dffuse most common Ambent, specular,

More information

SIGGRAPH Interactive Image Cutout. Interactive Graph Cut. Interactive Graph Cut. Interactive Graph Cut. Hard Constraints. Lazy Snapping.

SIGGRAPH Interactive Image Cutout. Interactive Graph Cut. Interactive Graph Cut. Interactive Graph Cut. Hard Constraints. Lazy Snapping. SIGGRAPH 004 Interactve Image Cutout Lazy Snappng Yn L Jan Sun Ch-Keung Tang Heung-Yeung Shum Mcrosoft Research Asa Hong Kong Unversty Separate an object from ts background Compose the object on another

More information

Interactive Virtual Relighting of Real Scenes

Interactive Virtual Relighting of Real Scenes Frst submtted: October 1998 (#846). Edtor/revewers please consult accompanyng document wth detaled responses to revewer comments. Interactve Vrtual Relghtng of Real Scenes Célne Loscos, George Drettaks,

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

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

Discussion. History and Outline. Smoothness of Indirect Lighting. Irradiance Calculation. Irradiance Caching. Advanced Computer Graphics (Fall 2009)

Discussion. History and Outline. Smoothness of Indirect Lighting. Irradiance Calculation. Irradiance Caching. Advanced Computer Graphics (Fall 2009) Advanced Computer Graphcs (Fall 2009 CS 283, Lecture 13: Recent Advances n Monte Carlo Offlne Renderng Rav Ramamoorth http://nst.eecs.berkeley.edu/~cs283/fa10 Dscusson Problems dfferent over years. Intally,

More information

Discussion. History and Outline. Smoothness of Indirect Lighting. Irradiance Caching. Irradiance Calculation. Advanced Computer Graphics (Fall 2009)

Discussion. History and Outline. Smoothness of Indirect Lighting. Irradiance Caching. Irradiance Calculation. Advanced Computer Graphics (Fall 2009) Advanced Computer Graphcs (Fall 2009 CS 29, Renderng Lecture 6: Recent Advances n Monte Carlo Offlne Renderng Rav Ramamoorth http://nst.eecs.berkeley.edu/~cs29-13/fa09 Dscusson Problems dfferent over years.

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

A Range Image Refinement Technique for Multi-view 3D Model Reconstruction

A Range Image Refinement Technique for Multi-view 3D Model Reconstruction A Range Image Refnement Technque for Mult-vew 3D Model Reconstructon Soon-Yong Park and Mural Subbarao Electrcal and Computer Engneerng State Unversty of New York at Stony Brook, USA E-mal: parksy@ece.sunysb.edu

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

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

Multi-Resolution Geometric Fusion

Multi-Resolution Geometric Fusion Internatonal Conference on Recent Advances n 3-D Dgtal Imagng and Modellng, Ottawa, Canada May 12 15, 1997 Mult-Resoluton Geometrc Fuson Adran Hlton and John Illngworth Centre for Vson, Speech and Sgnal

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

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

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

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

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

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

CS434a/541a: Pattern Recognition Prof. Olga Veksler. Lecture 15

CS434a/541a: Pattern Recognition Prof. Olga Veksler. Lecture 15 CS434a/541a: Pattern Recognton Prof. Olga Veksler Lecture 15 Today New Topc: Unsupervsed Learnng Supervsed vs. unsupervsed learnng Unsupervsed learnng Net Tme: parametrc unsupervsed learnng Today: nonparametrc

More information

Determining the Optimal Bandwidth Based on Multi-criterion Fusion

Determining the Optimal Bandwidth Based on Multi-criterion Fusion Proceedngs of 01 4th Internatonal Conference on Machne Learnng and Computng IPCSIT vol. 5 (01) (01) IACSIT Press, Sngapore Determnng the Optmal Bandwdth Based on Mult-crteron Fuson Ha-L Lang 1+, Xan-Mn

More information

Computer Vision I. Xbox Kinnect: Rectification. The Fundamental matrix. Stereo III. CSE252A Lecture 16. Example: forward motion

Computer Vision I. Xbox Kinnect: Rectification. The Fundamental matrix. Stereo III. CSE252A Lecture 16. Example: forward motion Xbox Knnect: Stereo III Depth map http://www.youtube.com/watch?v=7qrnwoo-8a CSE5A Lecture 6 Projected pattern http://www.youtube.com/watch?v=ceep7x-z4wy The Fundamental matrx Rectfcaton The eppolar constrant

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

Simulation Based Analysis of FAST TCP using OMNET++

Simulation Based Analysis of FAST TCP using OMNET++ Smulaton Based Analyss of FAST TCP usng OMNET++ Umar ul Hassan 04030038@lums.edu.pk Md Term Report CS678 Topcs n Internet Research Sprng, 2006 Introducton Internet traffc s doublng roughly every 3 months

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

Problem Definitions and Evaluation Criteria for Computational Expensive Optimization

Problem Definitions and Evaluation Criteria for Computational Expensive Optimization Problem efntons and Evaluaton Crtera for Computatonal Expensve Optmzaton B. Lu 1, Q. Chen and Q. Zhang 3, J. J. Lang 4, P. N. Suganthan, B. Y. Qu 6 1 epartment of Computng, Glyndwr Unversty, UK Faclty

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

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

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

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

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

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

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

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

A mathematical programming approach to the analysis, design and scheduling of offshore oilfields

A mathematical programming approach to the analysis, design and scheduling of offshore oilfields 17 th European Symposum on Computer Aded Process Engneerng ESCAPE17 V. Plesu and P.S. Agach (Edtors) 2007 Elsever B.V. All rghts reserved. 1 A mathematcal programmng approach to the analyss, desgn 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

Collaboratively Regularized Nearest Points for Set Based Recognition

Collaboratively Regularized Nearest Points for Set Based Recognition Academc Center for Computng and Meda Studes, Kyoto Unversty Collaboratvely Regularzed Nearest Ponts for Set Based Recognton Yang Wu, Mchhko Mnoh, Masayuk Mukunok Kyoto Unversty 9/1/013 BMVC 013 @ Brstol,

More information

Amnon Shashua Shai Avidan Michael Werman. The Hebrew University, objects.

Amnon Shashua Shai Avidan Michael Werman. The Hebrew University,   objects. Trajectory Trangulaton over Conc Sectons Amnon Shashua Sha Avdan Mchael Werman Insttute of Computer Scence, The Hebrew Unversty, Jerusalem 91904, Israel e-mal: fshashua,avdan,wermang@cs.huj.ac.l Abstract

More information

A Background Subtraction for a Vision-based User Interface *

A Background Subtraction for a Vision-based User Interface * A Background Subtracton for a Vson-based User Interface * Dongpyo Hong and Woontack Woo KJIST U-VR Lab. {dhon wwoo}@kjst.ac.kr Abstract In ths paper, we propose a robust and effcent background subtracton

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

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

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 high precision collaborative vision measurement of gear chamfering profile

A high precision collaborative vision measurement of gear chamfering profile Internatonal Conference on Advances n Mechancal Engneerng and Industral Informatcs (AMEII 05) A hgh precson collaboratve vson measurement of gear chamferng profle Conglng Zhou, a, Zengpu Xu, b, Chunmng

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

Backpropagation: In Search of Performance Parameters

Backpropagation: In Search of Performance Parameters Bacpropagaton: In Search of Performance Parameters ANIL KUMAR ENUMULAPALLY, LINGGUO BU, and KHOSROW KAIKHAH, Ph.D. Computer Scence Department Texas State Unversty-San Marcos San Marcos, TX-78666 USA ae049@txstate.edu,

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

An inverse problem solution for post-processing of PIV data

An inverse problem solution for post-processing of PIV data An nverse problem soluton for post-processng of PIV data Wt Strycznewcz 1,* 1 Appled Aerodynamcs Laboratory, Insttute of Avaton, Warsaw, Poland *correspondng author: wt.strycznewcz@lot.edu.pl Abstract

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

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

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

Optimal Combination of Stereo Camera Calibration from Arbitrary Stereo Images.

Optimal Combination of Stereo Camera Calibration from Arbitrary Stereo Images. Tna Memo No. 1991-002 Image and Vson Computng, 9(1), 27-32, 1990. Optmal Combnaton of Stereo Camera Calbraton from Arbtrary Stereo Images. N.A.Thacker and J.E.W.Mayhew. Last updated 6 / 9 / 2005 Imagng

More information

Steps for Computing the Dissimilarity, Entropy, Herfindahl-Hirschman and. Accessibility (Gravity with Competition) Indices

Steps for Computing the Dissimilarity, Entropy, Herfindahl-Hirschman and. Accessibility (Gravity with Competition) Indices Steps for Computng the Dssmlarty, Entropy, Herfndahl-Hrschman and Accessblty (Gravty wth Competton) Indces I. Dssmlarty Index Measurement: The followng formula can be used to measure the evenness between

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

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

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

Multiblock method for database generation in finite element programs

Multiblock method for database generation in finite element programs Proc. of the 9th WSEAS Int. Conf. on Mathematcal Methods and Computatonal Technques n Electrcal Engneerng, Arcachon, October 13-15, 2007 53 Multblock method for database generaton n fnte element programs

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

Compiler Design. Spring Register Allocation. Sample Exercises and Solutions. Prof. Pedro C. Diniz

Compiler Design. Spring Register Allocation. Sample Exercises and Solutions. Prof. Pedro C. Diniz Compler Desgn Sprng 2014 Regster Allocaton Sample Exercses and Solutons Prof. Pedro C. Dnz USC / Informaton Scences Insttute 4676 Admralty Way, Sute 1001 Marna del Rey, Calforna 90292 pedro@s.edu Regster

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