TOPOLOGY PRESERVING GRAPH MATCHING FOR PARTIAL FACE RECOGNITION

Size: px
Start display at page:

Download "TOPOLOGY PRESERVING GRAPH MATCHING FOR PARTIAL FACE RECOGNITION"

Transcription

1 TOPOLOGY PRESERVING GRAPH MATCHING FOR PARTIAL FACE RECOGNITION Yueq Duan 1,2,3, Jwen Lu 1,2,3,, Janjang Feng 1,2,3, Je Zhou 1,2,3 1 Department of Automaton, Tsnghua Unversty, Bejng, Chna 2 State Key Lab of Intellgent Technologes and Systems, Bejng, Chna 3 Tsnghua Natonal Laboratory for Informaton Scence and Technology (TNLst), Bejng, Chna duanyq14@mals.tsnghua.edu.cn; {lujwen,jfeng,jzhou}@tsnghua.edu.cn ABSTRACT In ths paper, we propose a topology preservng graph matchng (TPGM) method for partal face recognton. Most exstng face recognton methods extract features from holstc face mages, yet faces n real-world unconstraned envronments are usually occluded by objects or other faces, whch cannot provde the whole face mages for recognton. Latest keypont-based partal face recognton methods only match on the detected keyponts to remove the occluded regons. However, they smply measure the node-wse smlarty wthout hgher order geometrcal graph nformaton, thereby dependng heavly on descrptors whch are susceptble to noses. To address ths, our TPGM method estmates a non-rgd transformaton encodng the second order geometrc structure of the graph, so that more accurate and robust correspondence can be computed wth the topologcal nformaton. Expermental results on three wdely used face datasets show that the proposed TPGM outperforms most exstng state-of-the-art partal face recognton methods. Index Terms Partal Face Recognton, Graph Matchng, Keypont, Topology-Preservng 1. INTRODUCTION Face recognton s a longstandng computer vson problem and a number of face recognton approaches have been proposed over the past three decades [1 6]. Whle most methods have acheved mpressve performance under controlled condtons where frontal holstc face mages are prealgned and normalzed, there are stll challenges for unconstraned face recognton n many real-world applcatons. Typcal face recognton applcatons nclude smart survellance systems * Correspondng author. Ths work was supported n part by the Natonal Key Research and Development Program of Chna under Grant 2016YFB , the Natonal Natural Scence Foundaton of Chna under Grants , , , and , the Natonal 1000 Young Talents Plan Program, the Natonal Basc Research Program of Chna under Grant 2014CB349304, the Mnstry of Educaton of Chna under Grant , and the Tsnghua Unversty Intatve Scentfc Research Program. and handheld devces, where faces may be occluded by objects or other faces under crowded scenes. In such scenaros, only partal faces can be obtaned and face algnment may fal to work wth facal landmarks occluded. Most face recognton methods ncludng the state-of-theart CNN approaches [3 6] utlze the whole face mages for recognton, whch are not applcable to partal face recognton. On the one hand, they assume that each mage have the same content of the algned whole face and descrbe the holstc facal mages for representaton, yet the contents of mages may be dfferent even for the same person n partal face recognton, e.g. one mage wthout the left eye and another wthout the mouth. On the other hand, the occluded objects are ncluded n the representaton, whch harms the dscrmnatve power. Therefore, t s requred to desgn a partal face recognton method whch drectly recognzes partal faces and s robust to occlusons. In order to better recognze partal faces, the occluded facal parts should be removed when computng the smlartes between probe partal faces and gallery faces. Inspred by the motvatons above, a few keypont-based methods have been proposed by computng the smlartes over the detected feature keyponts on the face mages [7, 8], such as mult-keypont descrptor wth Gabor ternary pattern (MKD- GTP) [7] and robust pont set matchng (RPSM) [8], whch acheve the state-of-the-art performance on partal face recognton. However, they only explot node-wse smlarty, whch depend heavly on descrptors and are susceptble to llumnaton, deformaton and other noses. Graph matchng s an effectve manner to address such unstableness by explotng geometrc structure of the graph, whch enhances the robustness of feature matchng n varous of vsual analyss tasks such as object categorzaton [9], feature trackng [10] and acton recognton [11]. Inspred by the fact that graph matchng delvers hgher matchng accuracy and stronger stableness, we propose a topology preservng graph matchng (TPGM) method for partal face recognton by estmatng a non-rgd transformaton encodng the topologcal structure and measurng the correspondence between nodes and edges. Fg. 1 llustrates the ppelne of our proposed TPGM method /17/$31.00 c 2017 IEEE 1494

2 Fg. 1. The flowchart of our proposed TPGM approach for partal face recognton. For each par of mages, we frst detect SIFT keypont descrptors on both faces marked as green dots. Then, the keyponts are coarsely matched and selected based on Lowe s method, where green lnes represent the correct matches whle the red ones lnk the wrong matchng keypont pars. We construct the graphs for both mages by deployng Delaunay trangulaton wth blue lnes. In TPGM procedure, a non-rgd transformaton s estmated encodng the structural graph teratvely, wth outlers removed durng the teraton. Fnally, the probe partal face s algned and recognzed wth the transformaton and the matchng result. Expermental results on LFW [12], AR [13] and PubFg [14] show the effectveness of the proposed approach. 2. RELATED WORK There have been extensve work on robust face recognton wth occlusons n recent years [15, 16]. These methods requre face algnment, whch has been proved to be crtcal by Ekenel and Stefelhagen [17]. However, these face algnment based methods fal to work well wth unknown mssng facal portons. A number of part-based representatons have also been proposed, whch can be manly classfed nto subregonbased methods [1, 2] and component-based methods [18, 19]. However, as occluded face regons n real-world applcatons are exceedngly unstructured, both subregon-based methods and component-based methods may fal wth ncorrect occluson detecton or occluded facal components. More recently, a few keypont-based approaches have been proposed [7, 8] whch remove the occluded facal regons by computng on the detected feature keyponts. Lao et al. [7] have proposed a MKD-GTP method whch presented a general formulaton of the keypont-based partal face recognton. Weng et al. [8] have appled pont set matchng method by consderng the frst-order compatblty between pont sets, whch acheved the state-of-the-art performance. However, these methods only explot node-wse unary smlarty wthout hgher order geometrcal graph nformaton, whch depend largely on descrptors and are susceptble to noses. 3. PROPOSED APPROACH In ths secton, we frst model the partal face recognton as a geometrc graph matchng problem. Then, we present our TPGM approach and the optmzaton detals. Fnally, we ntroduce how to use TPGM for partal face recognton Feature Extracton and Keypont Flter In order to present TPGM to match a par of partal faces, we frst extract features on both mages, and then flter the canddate matchng keyponts. We follow [8] to extract the SftSurfSILBP descrptor. More specfcally, we frst utlze Scale-Invarant Feature Transform (SIFT) to detect and descrbe the local keyponts [20], whch s the most successful keypont descrptor n the lterature. We then concatenate the SIFT descrptor wth the Speeded Up Robust Features (SURF) to enhance ts robustness aganst llumnaton varatons [21], and apply the Scale Invarant LBP (SILBP) features [22] to obtan more detaled facal textures wth four dfferent sets of {P,R} values ncludng {8,1}, {8,2}, {16,2} and {16,3} to obtan four SILBP hstograms. SftSurfSILBP combnes the dscrmnatve power of SIFT, SURF and SILBP, presentng robustness to rotaton, scale and llumnaton, whch s crtcal for partal face recognton. The number of keyponts detected by SIFT n a typcal face mage can be hundreds, whch suffer from heavy computatonal cost f we drectly apply graph matchng on the keyponts. Therefore, we apply Lowe s matchng method to flter out obvous outlers at frst, and then perform our TPGM method only on the canddate keypont pars Topology Preservng Graph Matchng In order to ncorporate the parwse nformaton, we utlze Delaunay trangulaton to construct a graph for each mage, and present each graph as a 4-tuple g = {P, Q, T, G}. We 1495

3 denote P =[p 1,...,p n ] R 2 n as the set of nodes and Q = [q 1,...,q m ] R 2 m as the set of edges, where edges are the coordnate dfferences between nodes, and n and m are the number of keyponts and edges, respectvely. Also, we denote the textural features of each node as T = [t 1,...,t n ] R dt n, and the node-edge ncdence matrx as G {0, 1} n m to encode the topology of the graph, where g c =1f the cth edge connects the th node. In partcular, g P = {P P, Q P, T P, G P } represents the 4-tuple from the probe set, and g G = {P G, Q G, T G, G G } s from the gallery. Let f p ( ) be the non-rgd transformaton functon of nodes, and f q ( ) be the transformaton functon of edges, where f q (q c )=f p (p ) f p (p j ) for q c = p p j. Let h p ( ) be the matchng map to connect each probe node to the correspondng gallery node, and X {0, 1} np ng be the bnary correspondence matrx between nodes, where X j =1 for the matched th probe keypont and jth gallery keypont. We can represent the mappng as h p (p P )=PG X T. Smlarly, we denote h q ( ) as the matchng functon of edges, and defne Y {0, 1} mp mg as the bnary correspondence matrx between edges. For q P = p P k pp l and q G j = pg k pg l, Y j =1f X kk =1and X ll =1: [ ] Y = 1 (G P ) T XG G > 1, (1) where 1(true) =1and 1(false) =0. The physcal meanng of (G P ) T XG G {0, 1, 2} mp mg s the number of matchng nodes connected wth the correspondng edges, and edges are matchng only f both the two pars of connected nodes are matchng. Wth the denotatons above, we formulate the objectve functon for the partal face matchng as follows: mn J = K t (t P,h t (t P )) + λ p K p (f p (p P ),h p (p P )) + λ q K q (f q (q P ),h q (q P )), (2) where the frst term s to mnmze the dfferences of textural features between the matched nodes, the second term ams to mnmze the dstances between the transformed matched nodes and the physcal meanng of the thrd term s dstances between the transformed matched edges. Textural Matchng Cost: The frst term n (2) measures the dfference n texture feature. We sum up the textural dfference of the correspondng feature descrptors to obtan the overall textural matchng cost: n p K t = T j X j = tr(t T X), (3) n g j where T j = ((t P t G j )T (t P t G j )). Node Geometrc Matchng Cost: The second term n (2) measures the postonal dfferences of transformed probe keyponts and correspondng gallery keyponts. In order to ncorporate the non-rgd transformaton, we apply the thn plate splne (TPS) model, whch s wdely used for representng flexble coordnate transformatons. The bendng energy of TPS s defned as follows: n g K p = f p (p P ) h p (p P ) 2 2 ) 2 + λ [( 2 f p ) 2 ( 2 f p λ p x 2 + x y ( 2 f p ) 2 ] + y 2 dxdy, (4) where f p (p P ) s computed as f p (p P )=A 2 2 p P + b WΦ() n 1, Φ() np 1 = p P p P log ( p P p P 1 2 2). p P p P n p 2 2 log ( p P p P n p 2 2). (5) In (5), we denote A and b as the affne transformaton matrx and the bas vector, W as the weght matrx and Φ as the TPS kernel. We substtute L1 norm for the L2 norm for lnearzaton and rewrte the node geometrc matchng cost as follows: K p = AP P 2 n p + b1 T n p + WΦ np n p P G 2 n g X T n g n p 1 + λ λ p WΦ 1. (6) Edge Geometrc Matchng Cost: The thrd term n (2) measures the geometrc dfferences of transformed probe edges and correspondng gallery edges. We calculate the dstance between edges by the squared L2 norm of vector dfference. Let q P = p P 1 pp 2 and qg = p G 1 pg 2, and we can obtan: m p K q = (f p (p P 1) f p (p P 2)) (h p (p P 1) h p (p P 2)) 2 2. (7) Reformulatng and lnearzng the above equaton nto a matrx form result n: K q = AQ P 2 m p + W 2 np (ΦP ΦP j ) np m p Q G 2 m g Y T m g m p 1. (8) 3.3. Optmzaton Detals As the objectve functon s NP-hard wth dscrete constrants whch cannot be effcently solved, we perform a lnearzaton and rewrte the objectve functon nto a lnear programmng (LP) form: mn J = tr(t T X)+λ p 1 T 2 U1 np + λ1 T 2 V1 np + λ q 1 T 2 M1 mp λ x 1 T n p X1 mp, (9) 1496

4 subject to U AP P + b1 T n p + WΦ P G X T U, V WΦ V, M AQ P + W(ΦP ΦP j ) Q G Y T M, U 0, V 0, M 0, X j 1, X j 1, X j 0, j A 1,1 = A 2,2, A 1,2 = A 2,1. (10) In (9), the bnary constrant of X j {0, 1} np ng s relaxed to a contnuous doman 0 X j 1, and U R 2 np, V R 2 np and M R 2 mp are auxlary matrces representng upper bounds of K p1, K p2 and K q respectvely. As there are the degenerate cases when the optmal soluton of {A, W, X, b} are all zeros, we add a penalty term of X n the formulaton. Regon Shrnkage: We apply the successve trust regon shrnkage [23, 24] to solve the LP model. For each keypont p P n the probe mage, we set a trust regon D n the gallery mage, whch s a crcle wth center pont f p (p P ) and a radus of r. Only keyponts nsde the correspondng trust regon are consdered as the matchng canddates, whle others are excluded n the optmzaton process by settng X j =0. The trust regon cover the entre gallery mage at frst, and then gradually shrnks to refne the matchng canddates durng the teraton, whch s formulated as: r (n+1) = α 1 r (n), where 0 <α 1 < 1. Parameter Shrnkage: As face mages are globally rgd and locally non-rgd, we confne the transformaton to be rgd n the global exploraton perod, and then gradually convert to non-affne transformaton n the local explotaton perod. As λ controls the energy of non-affne transformaton, we ntalze λ to a relatvely large value, whch s gradually reduced t durng teratons. Specfcally, we apply λ (n+1) = α 2 λ (n+1), where 0 <α 2 < 1. Outler Removal: We remove the outler matchng pars durng the teraton by calculatng the summaton of each row of X. For an outler probe p P j, all elements n the jth row of X are close to 0. Gven a threshold 0 < τ < 1, f np k=1 X jk <τ, we consder the p P j as an outler and remove t from the canddate keyponts. Algorthm 1 summarzes the detaled procedure of the proposed TPGM method. As the physcal meanng of the frst three lambdas are smlar wth [8], we drectly followed the same parameters by settng λ p =0.01, λ x = mn(c) and λ =5for smplcty and a far comparson. λ q s fnally fxed to 0.05 for all the experments Partal Face Recognton Usng TPGM We obtan the average matchng cost d = J mn /(,j X j) smply dvdng the matchng cost by the number of matchng Algorthm 1: TPGM Input: Probe mage P, Gallery mage G, teraton number T and parameters λ p, λ x, λ, λ q, τ, α 1 and α 2 Output: A, b, W, X, Y 1: Detect keyponts and extract SftSurfSILBP features. 2: Usng Lowe s method to flter the keyponts and construct the graph g P and g G. 3: Intalze r = r nt and set the constrant set Φ =. 4: for t =1, 2,,T do 5: Construct Φ. 6: Obtan A, b, W, X, Y usng (9). 7: Clear the constrant set Φ to. 8: Fnd the outsders of the trust regon and construct the constrant set Φ. 9: Shrnkage: r (n+1) = α 1 r (n) and λ (n+1) 2 = α 2 λ (n+1) 2. 10: Remove p P, X :,f j Xj <τ. 11: Remove p G j, X :j,f 12: end for 13: Bnarze X. 14: return A, b, W, X, Y Xj <τ. keyponts. We further defne the dstance between the probe and the gallery as follows: d J d = mn,j X = j (,j X j) 2 = K t + λ p K p + λ q K q (,j X j) 2. (11) In (11), We separate the dstance nto two parts, the texture matchng dstance and the graph matchng dstance. It s proportonal to the average matchng cost whch ndcates the affnty of the matchng dfference, and t s nversely proportonal to the number of matchng pars whch ndcates the area of smlar parts n two mages. 4. EXPERIMENTS We evaluated the performance of our proposed TPGM on LFW [12], AR [13] and PubFg [14], where LFW and Pub- Fg are holstc face datasets and AR s a partal face dataset. We compared the proposed TPGM to several state-of-the-art methods on partal face recognton tasks, whch llustrated the effectveness and robustness of the proposed approach. 1 For LFW and PubFg, we added random transformaton (e.g. random occluson, rotaton) based on the orgnal face mages to evaluate the proposed method on partal face recognton tasks. 2 For AR, we tested on the orgnal AR datasets wth varant mage szes and dsalgnment. In the experments, we compared aganst 5 baselne algorthms, where three of them are algnment-free matchng methods and others are state-of-the-art partal face recognton approaches, ncludng Locally Affne Invarant Robust Pont 1 As the key contrbuton of the proposed method s for partal face recognton, we only conduct experments on partal face settngs. 2 We wll provde the modfed data along wth the codes for reproducton and evaluaton. 1497

5 Table 1. Mean verfcaton rate (VR) and correspondng standard devatons (%) comparson on LFW. Method VR ± S E HDLBP ± 1.09 CNN ± 1.38 CPD-SftSurfSILBP ± 1.19 MKD-SRC-GTP ± 1.77 MLERPM-SftSurf ± 1.53 MLERPM-SftSurfLBP ± 1.83 LAIRPM-SftSurf ± 1.02 LAIRPM-SftSurfSILBP ± 1.68 RPSM-SftSurf ± 1.46 RPSM-SftSurfSILBP ± 1.57 TPGM-SftSurfSILBP ± 1.12 set Matchng (LAIRPM) [8], Robust Pont Set Matchng (RPSM) [8], Metrc Learned Extended Robust Pont Matchng (MLERPM) [25], Mult-Keypont Descrptors-Sparse Representaton-based Classfcaton-Gabor Ternary Pattern (MKD-SRC-GTP) [7] and Coherent pont drft (CPD) [26] Results on LFW The Labeled Face n the Wld (LFW) database [12] conssts of labeled faces of 5749 subjects. The mages were obtaned under unconstraned crcumstances, causng large varatons n scale, rotaton, llumnaton and occluson. We evaluated the proposed method on the Vew 2 of the LFW dataset, whch ncludes 6000 pars n total wth half of them matched and the other msmatched. They are separated nto 10 folds wth 600 pars for each fold. In order to obtan partal face mages, we frst appled the Vola-Jones face detector to detect and crop the facal regons of all mages. Then, we randomly transformed those detected holstc facal mages to obtan the arbtrary face patches, where both gallery and probe mages were partal faces wth some facal components cropped out, makng t extremely dffcult to match. We added hgh dmensonal LBP (HDLBP) method [27] wth face algnment by CFAN facal landmark detector and CNN [5] feature wth VGG-16 network nto comparson, whch acheve the state-of-the-art performances on holstc face recognton tasks. Table 1 shows the mean verfcaton rates of dfferent methods. Our TPGM outperforms other baselne methods whch proves ts effectveness, whle HDLBP presents the poorest performance on the partal face recognton task. HDLBP employs CFAN to detect 25 landmarks, yet t does not work on partal faces especally when some facal components are occluded. CNN explots holstc facal mages for representaton through learnng from large amount of outsde data. However, the contents of each partal face mage are dfferent, whch leads to a confusng descrpton. The proposed TPGM drectly matches on the mage pars wthout tranng Table 2. Recognton accuracy (%) of dfferent methods on the AR dataset. Method S1-G S1-S S2-G S2-S CPD MLERPM LAIRPM MKD-SRC-GTP RPSM TPGM procedure and acheves the hghest accuracy on partal face recognton, whch proves ts effectveness and robustness. Iteraton Number and Computatonal Tme: We tested the average teraton number and computatonal tme of the proposed TPGM. Our hardware confguraton comprses of a 2.8-GHz CPU and a 15G RAM. In each experment, the tme complexty vares wth the number of ntal matchng keyponts, where the mean teraton number s 1.4 and the computatonal tme s 0.94s on LFW. The proposed TPGM converges much faster than RPSM, because the geometrc nformaton accelerates the convergence Results on AR The AR database [13] conssts of 126 denttes ncludng 70 males and 56 females. It contans two sessons and there are 13 face mages for each subject n a sesson, where three of them are taken under varous llumnaton condtons, four wth dfferent expressons, three wearng sunglasses and the other three wearng scarves. We evaluated the performance of our proposed approach on the face recognton task where the orgnal mages from the AR dataset were drectly used as the probe set wthout crop or algnment. It s ubqutous n real-lfe applcatons that probe mages and gallery mages are of dfferent sze and are not properly algned. Table 2 shows the recognton accuracy of dfferent methods. We see that our method obtaned the best result, demonstratng the robustness aganst varous scales. MKD-SRC-GTP utlzes Harrs-Laplacan keypont detector [7] whch s senstve to corners, so that a great number of mproper keyponts were detected n har regons, leadng to unstable and dscrmnatve descrptors Results on PubFg PubFg dataset [14] contans mages of 200 people from unconstraned condtons, whch vary n poses, llumnatons, expressons and background. The evaluaton set of 140 people are used for the probe set and the remanng 60 people are regarded as the gallery set. Followng the settngs n [8], we select 5 mages for each subject. The selected dataset well preserves the varaton n poses, llumnatons, expressons and background. 1498

6 Table 3. Average recognton rates (%) of dfferent methods wth dfferent ranks on PubFg. Method rank = 1 rank = 10 rank = 20 SftSurfSILBP CPD MLERPM MKD-SRC-GTP LAIRPM RPSM TPGM We randomly transformed the orgnal mages to obtan partal face patches. We frst rotated each mage randomly from 20 to 20. Then, we cropped the rotated mage to the 0.8 tmes of the orgnal mage sze. Fnally, the cropped mage was randomly scaled rangng from 0.8 to 1.2. We conducted fve-fold testng scheme by randomly splttng the cropped mages nto fve subsets. Besdes the baselne approaches, we also tested on the orgnal SftSurfSILBP where the Lowe s matchng scheme was utlzed for partal face recognton. Table 3 shows that TPGM outperforms other state-of-theart partal face recognton methods on PubFg. Compared wth SftSurfSILBP whch smply apples Lowe s matchng scheme, the proposed TPGM performs a graph matchng procedure to encode the graphcal nformaton and obtans about 18% ncreases n recognton rates, whch demonstrates the effectveness of the proposed TPGM approach. 5. CONCLUSION In ths paper, we have proposed a topology preservng graph matchng (TPGM) method for partal face recognton, where a non-rgd transformaton s estmated encodng the geometrc structure of the graph. Wth the geometrc graph structure, TPGM obtans more accurate and robust correspondence, whch leads to a stronger recognton ablty. Moreover, the proposed TPGM drectly matches on the partal mage pars, whch does not requre large amount of tranng data. The expermental results show the effectveness of TPGM. 6. REFERENCES [1] Tmo Ahonen, Abdenour Hadd, and Matt Petkanen, Face descrpton wth local bnary patterns: Applcaton to face recognton, TPAMI, vol. 28, no. 12, pp , [2] Jwen Lu, Vence Ern Long, Xuzhuang Zhou, and Je Zhou, Learnng compact bnary face descrptor for face recognton, TPAMI, vol. 37, no. 10, pp , [3] Y Sun, Xaogang Wang, and Xaoou Tang, Deep learnng face representaton from predctng 10,000 classes, n CVPR, 2014, pp [4] Yanv Tagman, Mng Yang, Marc Aurelo Ranzato, and Lor Wolf, Deepface: Closng the gap to human-level performance n face verfcaton, n CVPR, 2014, pp [5] Omkar M Parkh, Andrea Vedald, and Andrew Zsserman, Deep face recognton, n BMVC, 2015, vol. 1, pp [6] Floran Schroff, Dmtry Kalenchenko, and James Phlbn, Facenet: A unfed embeddng for face recognton and clusterng, n CVPR, 2015, pp [7] Shengca Lao, Anubhav K Jan, and Stan Z L, Partal face recognton: Algnment-free approach, TPAMI, vol. 35, no. 5, pp , [8] Renlang Weng, Jwen Lu, and Yap-Peng Tan, Robust pont set matchng for partal face recognton, TIP, vol. 25, no. 3, pp , [9] Olver Duchenne, Armand Jouln, and Jean Ponce, A graph-matchng kernel for object categorzaton, n ICCV, 2011, pp [10] Hao Jang, Stella X Yu, and Davd R Martn, Lnear scale and rotaton nvarant matchng, TPAMI, vol. 33, no. 7, pp , [11] Wllam Brendel and Snsa Todorovc, Learnng spatotemporal graphs of human actvtes, n ICCV, 2011, pp [12] Gary B Huang, Manu Ramesh, Tamara Berg, and Erk Learned-Mller, Labeled faces n the wld: A database for studyng face recognton n unconstraned envronments, Tech. Rep., Techncal Report 07-49, Unversty of Massachusetts, Amherst, [13] Alex M Martnez, The AR face database, CVC Techncal Report, vol. 24, [14] Neeraj Kumar, Alexander C Berg, Peter N Belhumeur, and Shree K Nayar, Attrbute and smle classfers for face verfcaton, n ICCV, 2009, pp [15] John Wrght, Allen Y Yang, Arvnd Ganesh, Shankar S Sastry, and Y Ma, Robust face recognton va sparse representaton, TPAMI, vol. 31, no. 2, pp , [16] Andrew Wagner, John Wrght, Arvnd Ganesh, Zhan Zhou, Hossen Mobah, and Y Ma, Toward a practcal face recognton system: Robust algnment and llumnaton by sparse representaton, TPAMI, vol. 34, no. 2, pp , [17] Hazım Kemal Ekenel and Raner Stefelhagen, Why s facal occluson a challengng problem?, n ICB, 2009, pp [18] Alex Pentland, Baback Moghaddam, and Thad Starner, Vew-based and modular egenspaces for face recognton, n CVPR, 1994, pp [19] Bernd Hesele, Purdy Ho, Jane Wu, and Tomaso Poggo, Face recognton: component-based versus global approaches, CVIU, vol. 91, no. 1, pp. 6 21, [20] Davd G Lowe, Dstnctve mage features from scale-nvarant keyponts, IJCV, vol. 60, no. 2, pp , [21] Luo Juan and Oubong Gwun, A comparson of sft, pca-sft and surf, IJIP, vol. 3, no. 4, pp , [22] Zh L, Guzhong Lu, Yang Yang, and Junyong You, Scale-and rotaton-nvarant local bnary pattern usng scale-adaptve texton and subunform-based crcular shft, TIP, vol. 21, no. 4, pp , [23] Hongsheng L, Xaole Huang, and Le He, Object matchng usng a locally affne nvarant and lnear programmng technques, TPAMI, vol. 35, no. 2, pp , [24] Hao Jang, Mark S Drew, and Ze-Nan L, Matchng by lnear programmng and successve convexfcaton, TPAMI, vol. 29, no. 6, pp , [25] Renlang Weng, Jwen Lu, Junln Hu, Gao Yang, and Yap-Peng Tan, Robust feature set matchng for partal face recognton, n ICCV, 2013, pp [26] Andry Myronenko and Xubo Song, Pont set regstraton: Coherent pont drft, TPAMI, vol. 32, no. 12, pp , [27] Dong Chen, Xudong Cao, Fang Wen, and Jan Sun, Blessng of dmensonalty: Hgh-dmensonal feature and ts effcent compresson for face verfcaton, n CVPR, 2013, pp

Discriminative Dictionary Learning with Pairwise Constraints

Discriminative Dictionary Learning with Pairwise Constraints Dscrmnatve Dctonary Learnng wth Parwse Constrants Humn Guo Zhuoln Jang LARRY S. DAVIS UNIVERSITY OF MARYLAND Nov. 6 th, Outlne Introducton/motvaton Dctonary Learnng Dscrmnatve Dctonary Learnng wth Parwse

More information

Outline. Discriminative classifiers for image recognition. Where in the World? A nearest neighbor recognition example 4/14/2011. CS 376 Lecture 22 1

Outline. Discriminative classifiers for image recognition. Where in the World? A nearest neighbor recognition example 4/14/2011. CS 376 Lecture 22 1 4/14/011 Outlne Dscrmnatve classfers for mage recognton Wednesday, Aprl 13 Krsten Grauman UT-Austn Last tme: wndow-based generc obect detecton basc ppelne face detecton wth boostng as case study Today:

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

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

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

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

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

A Fast Content-Based Multimedia Retrieval Technique Using Compressed Data

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

More information

A 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

Scale Selective Extended Local Binary Pattern For Texture Classification

Scale Selective Extended Local Binary Pattern For Texture Classification Scale Selectve Extended Local Bnary Pattern For Texture Classfcaton Yutng Hu, Zhlng Long, and Ghassan AlRegb Multmeda & Sensors Lab (MSL) Georga Insttute of Technology 03/09/017 Outlne Texture Representaton

More information

Gender Classification using Interlaced Derivative Patterns

Gender Classification using Interlaced Derivative Patterns Gender Classfcaton usng Interlaced Dervatve Patterns Author Shobernejad, Ameneh, Gao, Yongsheng Publshed 2 Conference Ttle Proceedngs of the 2th Internatonal Conference on Pattern Recognton (ICPR 2) DOI

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

Multi-View Face Alignment Using 3D Shape Model for View Estimation

Multi-View Face Alignment Using 3D Shape Model for View Estimation Mult-Vew Face Algnment Usng 3D Shape Model for Vew Estmaton Yanchao Su 1, Hazhou A 1, Shhong Lao 1 Computer Scence and Technology Department, Tsnghua Unversty Core Technology Center, Omron Corporaton ahz@mal.tsnghua.edu.cn

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

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

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

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

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

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

Recognizing Faces. Outline

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

More information

Face Recognition by Fusing Binary Edge Feature and Second-order Mutual Information

Face Recognition by Fusing Binary Edge Feature and Second-order Mutual Information Face Recognton by Fusng Bnary Edge Feature and Second-order Mutual Informaton Jatao Song, Bejng Chen, We Wang, Xaobo Ren School of Electronc and Informaton Engneerng, Nngbo Unversty of Technology Nngbo,

More information

EYE CENTER LOCALIZATION ON A FACIAL IMAGE BASED ON MULTI-BLOCK LOCAL BINARY PATTERNS

EYE CENTER LOCALIZATION ON A FACIAL IMAGE BASED ON MULTI-BLOCK LOCAL BINARY PATTERNS P.G. Demdov Yaroslavl State Unversty Anatoly Ntn, Vladmr Khryashchev, Olga Stepanova, Igor Kostern EYE CENTER LOCALIZATION ON A FACIAL IMAGE BASED ON MULTI-BLOCK LOCAL BINARY PATTERNS Yaroslavl, 2015 Eye

More information

Face Recognition University at Buffalo CSE666 Lecture Slides Resources:

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

More information

Face Detection with Deep Learning

Face Detection with Deep Learning Face Detecton wth Deep Learnng Yu Shen Yus122@ucsd.edu A13227146 Kuan-We Chen kuc010@ucsd.edu A99045121 Yzhou Hao y3hao@ucsd.edu A98017773 Mn Hsuan Wu mhwu@ucsd.edu A92424998 Abstract The project here

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

Computer Aided Drafting, Design and Manufacturing Volume 25, Number 2, June 2015, Page 14

Computer Aided Drafting, Design and Manufacturing Volume 25, Number 2, June 2015, Page 14 Computer Aded Draftng, Desgn and Manufacturng Volume 5, Number, June 015, Page 14 CADDM Face Recognton Algorthm Fusng Monogenc Bnary Codng and Collaboratve Representaton FU Yu-xan, PENG Lang-yu College

More information

Robust Kernel Representation with Statistical Local Features. for Face Recognition

Robust Kernel Representation with Statistical Local Features. for Face Recognition Robust Kernel Representaton wth Statstcal Local Features for Face Recognton Meng Yang, Student Member, IEEE, Le Zhang 1, Member, IEEE Smon C. K. Shu, Member, IEEE, and Davd Zhang, Fellow, IEEE Dept. of

More information

An efficient method to build panoramic image mosaics

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

More information

Combination of Local Multiple Patterns and Exponential Discriminant Analysis for Facial Recognition

Combination of Local Multiple Patterns and Exponential Discriminant Analysis for Facial Recognition Sensors & ransducers 203 by IFSA http://.sensorsportal.com Combnaton of Local Multple Patterns and Exponental Dscrmnant Analyss for Facal Recognton, 2 Lfang Zhou, 2 Bn Fang, 3 Wesheng L, 3 Ldou Wang College

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

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

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

Integrated Expression-Invariant Face Recognition with Constrained Optical Flow

Integrated Expression-Invariant Face Recognition with Constrained Optical Flow Integrated Expresson-Invarant Face Recognton wth Constraned Optcal Flow Chao-Kue Hseh, Shang-Hong La 2, and Yung-Chang Chen Department of Electrcal Engneerng, Natonal Tsng Hua Unversty, Tawan 2 Department

More information

Competitive Sparse Representation Classification for Face Recognition

Competitive Sparse Representation Classification for Face Recognition Vol. 6, No. 8, 05 Compettve Sparse Representaton Classfcaton for Face Recognton Yng Lu Chongqng Key Laboratory of Computatonal Intellgence Chongqng Unversty of Posts and elecommuncatons Chongqng, Chna

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

3D Face Recognition Fusing Spherical Depth Map and Spherical Texture Map

3D Face Recognition Fusing Spherical Depth Map and Spherical Texture Map Journal of Computer and Communcatons, 14, *, ** Publshed Onlne **** 14 n ScRes. http://www.scrp.org/journal/jcc http://dx.do.org/1.436/jcc.14.***** 3D Face Recognton Fusng Sphercal Depth Map and Sphercal

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

Histogram of Template for Pedestrian Detection

Histogram of Template for Pedestrian Detection PAPER IEICE TRANS. FUNDAMENTALS/COMMUN./ELECTRON./INF. & SYST., VOL. E85-A/B/C/D, No. xx JANUARY 20xx Hstogram of Template for Pedestran Detecton Shaopeng Tang, Non Member, Satosh Goto Fellow Summary In

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

Support Vector Machines

Support Vector Machines Support Vector Machnes Decson surface s a hyperplane (lne n 2D) n feature space (smlar to the Perceptron) Arguably, the most mportant recent dscovery n machne learnng In a nutshell: map the data to a predetermned

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

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

Data Mining: Model Evaluation

Data Mining: Model Evaluation Data Mnng: Model Evaluaton Aprl 16, 2013 1 Issues: Evaluatng Classfcaton Methods Accurac classfer accurac: predctng class label predctor accurac: guessng value of predcted attrbutes Speed tme to construct

More information

Object-Based Techniques for Image Retrieval

Object-Based Techniques for Image Retrieval 54 Zhang, Gao, & Luo Chapter VII Object-Based Technques for Image Retreval Y. J. Zhang, Tsnghua Unversty, Chna Y. Y. Gao, Tsnghua Unversty, Chna Y. Luo, Tsnghua Unversty, Chna ABSTRACT To overcome the

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

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

Machine Learning 9. week

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

More information

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

Tone-Aware Sparse Representation for Face Recognition

Tone-Aware Sparse Representation for Face Recognition Tone-Aware Sparse Representaton for Face Recognton Lngfeng Wang, Huayu Wu and Chunhong Pan Abstract It s stll a very challengng task to recognze a face n a real world scenaro, snce the face may be corrupted

More information

Fitting and Alignment

Fitting and Alignment Fttng and Algnment Computer Vson Ja-Bn Huang, Vrgna Tech Many sldes from S. Lazebnk and D. Hoem Admnstratve Stuffs HW 1 Competton: Edge Detecton Submsson lnk HW 2 wll be posted tonght Due Oct 09 (Mon)

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

Feature-based image registration using the shape context

Feature-based image registration using the shape context Feature-based mage regstraton usng the shape context LEI HUANG *, ZHEN LI Center for Earth Observaton and Dgtal Earth, Chnese Academy of Scences, Bejng, 100012, Chna Graduate Unversty of Chnese Academy

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

Positive Semi-definite Programming Localization in Wireless Sensor Networks

Positive Semi-definite Programming Localization in Wireless Sensor Networks Postve Sem-defnte Programmng Localzaton n Wreless Sensor etworks Shengdong Xe 1,, Jn Wang, Aqun Hu 1, Yunl Gu, Jang Xu, 1 School of Informaton Scence and Engneerng, Southeast Unversty, 10096, anjng Computer

More information

Extraction of Texture Information from Fuzzy Run Length Matrix

Extraction of Texture Information from Fuzzy Run Length Matrix Internatonal Journal of Computer Applcatons (0975 8887) Volume 55 o.1, October 01 Extracton of Texture Informaton from Fuzzy Run Length Matrx Y. Venkateswarlu Head Dept. of CSE&IT Chatanya Insttuteof Engg.

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

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

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

More information

A Unified Framework for Semantics and Feature Based Relevance Feedback in Image Retrieval Systems

A Unified Framework for Semantics and Feature Based Relevance Feedback in Image Retrieval Systems A Unfed Framework for Semantcs and Feature Based Relevance Feedback n Image Retreval Systems Ye Lu *, Chunhu Hu 2, Xngquan Zhu 3*, HongJang Zhang 2, Qang Yang * School of Computng Scence Smon Fraser Unversty

More information

Correlative features for the classification of textural images

Correlative features for the classification of textural images Correlatve features for the classfcaton of textural mages M A Turkova 1 and A V Gadel 1, 1 Samara Natonal Research Unversty, Moskovskoe Shosse 34, Samara, Russa, 443086 Image Processng Systems Insttute

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

Kernel Collaborative Representation Classification Based on Adaptive Dictionary Learning

Kernel Collaborative Representation Classification Based on Adaptive Dictionary Learning Internatonal Journal of Intellgent Informaton Systems 2018; 7(2): 15-22 http://www.scencepublshnggroup.com/j/js do: 10.11648/j.js.20180702.11 ISSN: 2328-7675 (Prnt); ISSN: 2328-7683 (Onlne) Kernel Collaboratve

More information

Fast High Dimensional Vector Multiplication Face Recognition

Fast High Dimensional Vector Multiplication Face Recognition Fast Hgh Dmensonal Vector Multplcaton Face Recognton Oren Barkan Tel Avv Unversty orenbarkan@post.tau.ac.l Jonathan Well Tel Avv Unversty yonathanw@post.tau.ac.l Lor Wolf Tel Avv Unversty wolf@cs.tau.ac.l

More information

Modular PCA Face Recognition Based on Weighted Average

Modular PCA Face Recognition Based on Weighted Average odern Appled Scence odular PCA Face Recognton Based on Weghted Average Chengmao Han (Correspondng author) Department of athematcs, Lny Normal Unversty Lny 76005, Chna E-mal: hanchengmao@163.com Abstract

More information

PERFORMANCE EVALUATION FOR SCENE MATCHING ALGORITHMS BY SVM

PERFORMANCE EVALUATION FOR SCENE MATCHING ALGORITHMS BY SVM PERFORMACE EVALUAIO FOR SCEE MACHIG ALGORIHMS BY SVM Zhaohu Yang a, b, *, Yngyng Chen a, Shaomng Zhang a a he Research Center of Remote Sensng and Geomatc, ongj Unversty, Shangha 200092, Chna - yzhac@63.com

More information

IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY 1. SSDH: Semi-supervised Deep Hashing for Large Scale Image Retrieval

IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY 1. SSDH: Semi-supervised Deep Hashing for Large Scale Image Retrieval IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY SSDH: Sem-supervsed Deep Hashng for Large Scale Image Retreval Jan Zhang, and Yuxn Peng arxv:607.08477v2 [cs.cv] 8 Jun 207 Abstract Hashng

More information

Palmprint Feature Extraction Using 2-D Gabor Filters

Palmprint Feature Extraction Using 2-D Gabor Filters Palmprnt Feature Extracton Usng 2-D Gabor Flters Wa Kn Kong Davd Zhang and Wenxn L Bometrcs Research Centre Department of Computng The Hong Kong Polytechnc Unversty Kowloon Hong Kong Correspondng author:

More information

Face Recognition using 3D Directional Corner Points

Face Recognition using 3D Directional Corner Points 2014 22nd Internatonal Conference on Pattern Recognton Face Recognton usng 3D Drectonal Corner Ponts Xun Yu, Yongsheng Gao School of Engneerng Grffth Unversty Nathan, QLD, Australa xun.yu@grffthun.edu.au,

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

Robust Inlier Feature Tracking Method for Multiple Pedestrian Tracking

Robust Inlier Feature Tracking Method for Multiple Pedestrian Tracking 2011 Internatonal Conference on Informaton and Intellgent Computng IPCSIT vol.18 (2011) (2011) IACSIT Press, Sngapore Robust Inler Feature Trackng Method for Multple Pedestran Trackng Young-Chul Lm a*

More information

Robust Low-Rank Regularized Regression for Face Recognition with Occlusion

Robust Low-Rank Regularized Regression for Face Recognition with Occlusion Robust Low-Rank Regularzed Regresson for ace Recognton wth Occluson Janjun Qan, Jan Yang, anlong Zhang and Zhouchen Ln School of Computer Scence and ngneerng, Nanjng Unversty of Scence and echnology Key

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

Robust Dictionary Learning with Capped l 1 -Norm

Robust Dictionary Learning with Capped l 1 -Norm Proceedngs of the Twenty-Fourth Internatonal Jont Conference on Artfcal Intellgence (IJCAI 205) Robust Dctonary Learnng wth Capped l -Norm Wenhao Jang, Fepng Ne, Heng Huang Unversty of Texas at Arlngton

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

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

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

More information

Mercer Kernels for Object Recognition with Local Features

Mercer Kernels for Object Recognition with Local Features TR004-50, October 004, Department of Computer Scence, Dartmouth College Mercer Kernels for Object Recognton wth Local Features Swe Lyu Department of Computer Scence Dartmouth College Hanover NH 03755 A

More information

Fast Feature Value Searching for Face Detection

Fast Feature Value Searching for Face Detection Vol., No. 2 Computer and Informaton Scence Fast Feature Value Searchng for Face Detecton Yunyang Yan Department of Computer Engneerng Huayn Insttute of Technology Hua an 22300, Chna E-mal: areyyyke@63.com

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

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

A Novel Fingerprint Matching Method Combining Geometric and Texture Features

A Novel Fingerprint Matching Method Combining Geometric and Texture Features A Novel ngerprnt Matchng Method Combnng Geometrc and Texture eatures Me Xe, Chengpu Yu and Jn Q Unversty of Electronc Scence and Technology of Chna. Chengdu,P.R.Chna xeme@ee.uestc.edu.cn Post Code:6154

More information

Feature-Area Optimization: A Novel SAR Image Registration Method

Feature-Area Optimization: A Novel SAR Image Registration Method Feature-Area Optmzaton: A Novel SAR Image Regstraton Method Fuqang Lu, Fukun B, Lang Chen, Hao Sh and We Lu Abstract Ths letter proposes a synthetc aperture radar (SAR) mage regstraton method named Feature-Area

More information

SIFT Match Verification by Geometric Coding for Large-Scale Partial-Duplicate Web Image Search

SIFT Match Verification by Geometric Coding for Large-Scale Partial-Duplicate Web Image Search SIFT Match Verfcaton by Geometrc Codng for Large-Scale Partal-Duplcate Web Image Search WENGANG ZHOU and HOUQIANG LI, Unversty of Scence and Technology of Chna YIJUAN LU, Texas State Unversty QI TIAN,

More information

Augmented Distinctive Features for Efficient Image Matching

Augmented Distinctive Features for Efficient Image Matching Augmented Dstnctve Features for Effcent Image Matchng Quan Wang, We Guan and Suya You CGIT/IMSC USC Los Angeles, CA 90089 quanwang@usc.edu, wguan@usc.edu and suyay@graphcs.usc.edu Abstract Fndng correspondng

More information

Deformable Part-based Robust Face Detection under Occlusion by Using Face Decomposition into Face Components

Deformable Part-based Robust Face Detection under Occlusion by Using Face Decomposition into Face Components Deformable Part-based Robust Face Detecton under Occluson by Usng Face Decomposton nto Face Components Darjan Marčetć, Slobodan Rbarć Unversty of Zagreb, Faculty of Electrcal Engneerng and Computng, Croata

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

Discriminative classifiers for object classification. Last time

Discriminative classifiers for object classification. Last time Dscrmnatve classfers for object classfcaton Thursday, Nov 12 Krsten Grauman UT Austn Last tme Supervsed classfcaton Loss and rsk, kbayes rule Skn color detecton example Sldng ndo detecton Classfers, boostng

More information

Range Data Registration Using Photometric Features

Range Data Registration Using Photometric Features Range Data Regstraton Usng Photometrc Features Joon Kyu Seo, Gregory C. Sharp, and Sang Wook Lee Dept. of Meda Technology, Sogang Unversty, Seoul, Korea Dept. of Radaton Oncology, Massachusetts General

More information

Signature and Lexicon Pruning Techniques

Signature and Lexicon Pruning Techniques Sgnature and Lexcon Prunng Technques Srnvas Palla, Hansheng Le, Venu Govndaraju Centre for Unfed Bometrcs and Sensors Unversty at Buffalo {spalla2, hle, govnd}@cedar.buffalo.edu Abstract Handwrtten word

More information

Learning a Class-Specific Dictionary for Facial Expression Recognition

Learning a Class-Specific Dictionary for Facial Expression Recognition BULGARIAN ACADEMY OF SCIENCES CYBERNETICS AND INFORMATION TECHNOLOGIES Volume 16, No 4 Sofa 016 Prnt ISSN: 1311-970; Onlne ISSN: 1314-4081 DOI: 10.1515/cat-016-0067 Learnng a Class-Specfc Dctonary for

More information

TN348: Openlab Module - Colocalization

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

More information

Using the Visual Words based on Affine-SIFT Descriptors for Face Recognition

Using the Visual Words based on Affine-SIFT Descriptors for Face Recognition Usng the Vsual Words based on Affne-SIFT Descrptors for Face Recognton Yu-Shan Wu, Heng-Sung Lu, Gwo-Hwa Ju, Tng-We Lee, Yen-Ln Chu Busness Customer Solutons Lab., Chunghwa Telecommuncaton Laboratores

More information

Robust Mean Shift Tracking with Corrected Background-Weighted Histogram

Robust Mean Shift Tracking with Corrected Background-Weighted Histogram Robust Mean Shft Trackng wth Corrected Background-Weghted Hstogram Jfeng Nng, Le Zhang, Davd Zhang and Chengke Wu Abstract: The background-weghted hstogram (BWH) algorthm proposed n [] attempts to reduce

More information

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

Categorizing objects: of appearance

Categorizing objects: of appearance Categorzng objects: global and part-based models of appearance UT Austn Generc categorzaton problem 1 Challenges: robustness Realstc scenes are crowded, cluttered, have overlappng objects. Generc category

More information

2 ZHENG et al.: ASSOCIATING GROUPS OF PEOPLE (a) Ambgutes from person re dentfcaton n solaton (b) Assocatng groups of people may reduce ambgutes n mat

2 ZHENG et al.: ASSOCIATING GROUPS OF PEOPLE (a) Ambgutes from person re dentfcaton n solaton (b) Assocatng groups of people may reduce ambgutes n mat ZHENG et al.: ASSOCIATING GROUPS OF PEOPLE 1 Assocatng Groups of People We-Sh Zheng jason@dcs.qmul.ac.uk Shaogang Gong sgg@dcs.qmul.ac.uk Tao Xang txang@dcs.qmul.ac.uk School of EECS, Queen Mary Unversty

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

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

A COMBINED APPROACH USING TEXTURAL AND GEOMETRICAL FEATURES FOR FACE RECOGNITION

A COMBINED APPROACH USING TEXTURAL AND GEOMETRICAL FEATURES FOR FACE RECOGNITION ISSN: 0976-910(ONLINE) ICTACT JOURNAL ON IMAGE AND VIDEO PROCESSING, MAY 013, VOLUME: 03, ISSUE: 04 A COMBINED APPROACH USING TEXTURAL AND GEOMETRICAL FEATURES FOR FACE RECOGNITION A. Suruland 1, R. Reena

More information

Improved SIFT-Features Matching for Object Recognition

Improved SIFT-Features Matching for Object Recognition Improved SIFT-Features Matchng for Obect Recognton Fara Alhwarn, Chao Wang, Danela Rstć-Durrant, Axel Gräser Insttute of Automaton, Unversty of Bremen, FB / NW Otto-Hahn-Allee D-8359 Bremen Emals: {alhwarn,wang,rstc,ag}@at.un-bremen.de

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

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