Super-resolution with Nonlocal Regularized Sparse Representation

Size: px
Start display at page:

Download "Super-resolution with Nonlocal Regularized Sparse Representation"

Transcription

1 Super-resoluton wth Nonlocal Regularzed Sparse Representaton Wesheng Dong a, Guangmng Sh a, Le Zhang b, and Xaoln Wu c a Key Laboratory of Intellgent Percepton and Image Understandng (Chnese Mnstry of Educaton), School of Electronc Engneerng, Xdan Unversty, Chna b Dept. of Computng, The Hong Kong Polytechnc Unversty, Hong Kong c Dept. of Electrcal and Computer Engneerng, McMaster Unversty, Canada ABSTRACT The reconstructon of a hgh resoluton (HR) mage from ts low resoluton (LR) counterpart s a challengng problem. The recently developed sparse representaton (SR) technques provde new solutons to ths nverse problem by ntroducng the l 1 -norm sparsty pror nto the super-resoluton reconstructon process. In ths paper, we present a new SR based mage super-resoluton by optmzng the objectve functon under an adaptve sparse doman and wth the nonlocal regularzaton of the HR mages. The adaptve sparse doman s estmated by applyng prncpal component analyss to the grouped nonlocal smlar mage patches. The proposed objectve functon wth nonlocal regularzaton can be effcently solved by an teratve shrnkage algorthm. The experments on natural mages show that the proposed method can reconstruct HR mages wth sharp edges from degraded LR mages. Keywords: sparse representaton, nonlocal self-smlarty, super-resoluton, teratve shrnkage algorthm 1. INTRODUCTION Image super-resoluton ams to reconstruct a hgh resoluton (HR) mage from ts degraded and down-sampled low resoluton (LR) counterparts. It has wde applcatons n computer vson, mage enhancement, medcal magng and hgh defnton televsons. A typcal mage super-resoluton process conssts of LR mage denosng, nterpolaton and deblurrng, subjectng to a known magng model of the LR mage. The LR magng process can be modeled by: y=dh x+ n (1) where H s a blurrng operator that mmcs the pont spread functon of a camera, D s a down-samplng operator, n s the nose ntroduced n the LR mage generaton, x s the target HR mage, and y s the observed LR mage. The reconstructon of the orgnal HR mage s a typcal ll-posed nverse problem. Ths problem becomes even more underdetermned n the case that only one LR mage s avalable. In ths paper, we concentrate on the mage superresoluton from only a sngle mage. Conventonal lnear nterpolaton based methods, such as blnear, bcubc and cubc splne nterpolators, reconstruct HR mages wth jaggy and zppng artfacts. To mprove the lnear nterpolators, drectonal nterpolators [1]-[3] have been proposed to perform the nterpolaton along the edge drectons. Especally, n [3] Zhang and Wu optmze the nterpolator based on the local covarance of the mage sgnal and acheve state-of-the-art nterpolaton results. However, all these nterpolaton based approaches do not handle the blurrng and noses n the LR mages. Addtonal steps have to be carred out to remove the blurrng and noses n the LR mages. However, the separated denosng and deblurrng steps do not suffcently explot the nformaton hdden n the magng model n Eq. (1). Another classcal super-resoluton approach s teratve back-projecton (IBP) [4], whch s desgned based on the magng model n Eq. (1). However, the IBP process nvolves much uncertanty n recoverng the HR mages, and hence chessboard and zppng artfacts are commonly observed n the reconstructed HR mages by IBP algorthm. Ths manly because IBP tres to fnd the x that could mnmze y DHx ; however, there could be many possble canddates of x that could make y DHx very small. In other words, the soluton space of IBP s too bg so that the resultng soluton may not be the best one. To mprove the performance of the IBP method, mage pror knowledge that take nto account the local mage edge geometres and nonlocal mage redundances has been ncorporated nto the IBP process n [5] [6]. The regularzed IBP technques by blateral flters [5] and nonlocal means based flters [6] can remove many artfacts of the HR mages generated by the orgnal IBP method. However, ther performances are stll not very satsfyng n recoverng fne mage

2 detals and suppressng noses. To reconstruct more vsually pleasng HR mages, more pror knowledge of natural mages should be used to reduce the uncertanty of the HR mages. Typcally, the regularzed super-resoluton can be formulated by solvng the followng mnmzaton problem: arg mn y DHx +λj ( x ) () x where J(x) s a regularzaton term specfyng the pror knowledge of the HR mage and λ s a scalar balancng between the quadratc fdelty term and the regularzaton term. A well-known regularzaton pror s the mnmum pxel ntensty total varatons (TV) [7]. The TV-regularzed approaches favor the mages wth pecewse smooth edge structures n the soluton space of x; however, the TV-based super-resoluton technques cannot recover mage fne detals and often have starcase artfacts. Other mage prors, such as edge smoothness [8] and gradent profle prors [9] have been proposed for mage super-resoluton, yet the resultng edges look unnatural. A recently proposed mage pror knowledge s sparsty of natural mages. It assumes that a natural mage can be sparsely represented n some specfc doman (e.g. wavelet doman, Fourer doman), or t has a sparse expanson over a dctonary of atoms,.e. x=ψα and most of the coeffcents n α are nearly zero. In general, the sparsty constran on α s mplemented by requrng that the l 1 -norm of α s small enough,.e. α 1 <t. The sparse representaton (SR) technques have been successfully appled to a seres of nverse problems, ncludng mage deblurrng [10], denosng [11] and compressve sensng [1]. In [13], the SR of the HR mage over a learned dctonary was proposed to regularze the mage super-resoluton process. As a learnng-based method, ts performance reles on the tranng set, and the learnng of a unversal dctonary for SR s very complcated. In ths paper, we present a new SR based model to reconstruct an HR mage from ts LR counterpart. The contrbutons of the proposed approach are twofold. Frst, we propose to adaptvely estmate the sparse doman of the HR mage patches usng adaptve prncpal component analyss (PCA) va non-local smlar patch groupng. Second, to further enhance the performance of the proposed approach, a nonlocal self-smlarty quadratc constran s also ntroduced to fully explot the nonlocal mage redundances. Snce both the two ponts use the non-local nformaton of the mage, we call the proposed method non-local regularzed SR for mage super-resoluton. In addton, the proposed mnmzaton problem wth non-local regularzatons can be effcently solved by usng a new famly of numercal algorthms, called teratve shrnkage algorthms [14] [15]. The rest of the paper s organzed as follows. Secton presents the proposed algorthm n detal. Secton 3 conducts experments to valdate the performance of the proposed method. Secton 4 concludes the paper.. THE PROPOSED IMAGE SUPER RESOLUTION APPROACH The sparsty pror of natural mages n a specfc transform doman s an effectve constran to refne the soluton space of mage super-resoluton. Wavelet transform, DCT, curvelet and contourlet transforms are commonly used transforms for sparse mage representaton. However, these transforms use a fxed set of bases, whch lack flexbltes n adaptng to varous complex local structures n natural mages. Therefore, dctonary learnng [16] technques have been proposed to learn a unversal over-complete dctonary of atoms so that the mage can be sparsely coded va l 1 -norm mnmzaton. However, for each mage local patch, there are too many rrelatve atoms n the unversal dctonary, whch degrades the effectveness and effcency of sparse representaton (SR) and hence degrades the mage reconstructon performance. It s of hgh demand that we could adaptvely determne the sparse doman of each local patch. In [17] [18], the prncpal component analyss (PCA) technque was used to adaptvely de-correlate the mage local structures for nose removal. For each mage block, a PCA transformaton matrx s locally computed. To ths end, a set of mage blocks that are smlar to the current mage block s grouped and a PCA transform matrx s computed over the tranng dataset. The PCA transformaton matrx actually defnes a type of mage local sparse doman because the mage local patch can be well reconstructed by usng only a few sgnfcant prncpal components. Dfferent from the wavelet transform, DCT, etc, such a PCA transform s sgnal adaptve. In [16-17], state-of-the-art mage denosng results have been obtaned by thresholdng n the adaptve PCA domans. Inspred by the work n [16-17], n ths paper we propose to use the locally adaptve PCA transform as the adaptve sparse doman. On the other hand, the nonlocal self-smlarty pror s used as regularzaton term n the SR based mage reconstructon. Fnally, an objectve functon wll be

3 constructed, whch s an l 1 -norm and l -norm compounded SR mnmzaton problem. Fortunately, an effcent numercal algorthm can be readly obtaned by usng the recently developed teratve shrnkage (IS) technques [14] [15]..1 Adaptve sparse doman determnaton by local PCA transform To adaptvely compute the sparse doman of each local patch, we model a local mage patch as a vector varable and then calculate ts statstcs by usng ts avalable samples. Specfcally, the adaptve PCA transformaton method proposed n [17] [18] s used here to determne the adaptve sparse doman of each local patch. Denote by v x = [ x1, x, L, x ] T m a vector varable contanng all pxel values wthn a w w mage patch. To compute the PCA transformaton matrx of x v, a tranng dataset of t s needed. Denote by x1,0 x1,1 L x1, n 1 x,0 x,1 x, n 1 X L = (3) M M M M xm,0 xm,1 L xm, n 1 a sample matrx of varable x v, where the th row of sample matrx X, denoted by X = x,0 x,1 L x, n 1, s the sample vector of varable x. The sample vector X s centralzed as X = X μ, where μ s the mean of X. Other rows of X can be centralzed analogously and we denote by X the centralzed matrx of X. Wth X, the covarance matrx of X s calculated by 1 T Ω = XX (4) n The goal of PCA s to fnd an orthonormal transformaton matrx P to decorrelate X, Y = PX, so that the covarance matrx of Y s dagonal. Snce Ω s symmetrcal, by apply the SVD (sngular value decomposton) to t, we have T Ω = ΦΛΦ (5) where Φ = φ1 φ L φm s the m m orthonormal egenvector matrx and Λ = dag{ λ1, λ,..., λm} s the dagonal egenvalue matrx wth λ 1 λ... λm. The terms φ 1, φ,..., φ m and λ1, λ,..., λ m are the egenvectors and egenvalues T 1 T of Ω. By settng Ρ = Φ, X can be decorrelated,.e. Y = PX and Λ = YY. An mportant property of PCA s that t n fully de-correlates the orgnal dataset X. Most of the energy of a sgnal wll concentrate on a small subset of the PCA transformed dataset. To apply the block-based adaptve PCA transform for the gven mage, we need a set of tranng sample for each mage patch x v so that the PCA transformaton matrx P can be computed. To ths end, we can collect every possble mage patches n a large wndow centered on x v. However, there can be many dfferent mage patches from the current one n the search area, whch wll deterorate the estmaton of the covarance and hence the PCA matrx. To exclude these rrelatve tranng samples, a block matchng method s used to select the mage patches wth smlar spatal structures. The mage patches whose Eucld dstances to x v are smaller than a predefned threshold are selected: v v x x < τ. By such constrant, nonlocal mage nformaton can be used to compute the PCA transformaton matrxes. The orderng property of PCA bases allows a good reconstructon of the mage by usng only a few prncpal components, whch naturally provde an adaptve sparse doman of each local patch. By allowng the overlappng of the mage patches, pxels n overlapped regon wll be transformed nto dfferent PCA domans. Ths forms a redundant representaton of mage sgnals, whch s very helpful n suppressng noses.. Sparse representaton wth adaptve sparse doman Snce the problem of mage super-resoluton nvolves severe uncertantes, pror knowledge s requred to regularze the soluton. The proposed SR based mage super-resoluton usng adaptve PCA transformatons can be formulated by arg mn y DHx + λ P x v 1 (6) x

4 where x v s the vector of the th mage patch and P s the computed PCA transformaton matrx of x v. λ s a scalar that balances the l 1 -norm and l -norm terms. A problem of (6) s that the computaton of P requres that the mage sgnal x should be estmated frst. However, the estmaton of x wth Eq. (6) requres that P s avalable. Ths s a chcken and egg dlemma. To solve ths problem, we ntally reconstruct x by settng P as a wavelet transformaton matrx. Once x s ntalzed, P can be estmated, and subsequently x can be updated by solvng Eq. (6). Wth the updated x, the PCA transformaton matrxes P can be further updated. These procedures can be terated to alternatvely optmze the reconstructed HR mage and the transformaton matrxes..3 Regularzaton wth non-local self-smlarty constrant The adaptve block-based PCA transformaton can better characterze the varous local mage structures and hence an adaptve sparse representaton can be acheved, whch s very helpful to mprovng the mage reconstructon performance. However, the estmated PCA transformaton matrxes may not be very accurate due to the lmted nformaton of the orgnal HR mage. Ths wll deterorate the performance of the proposed approach. To mprove the qualty of the reconstructed HR mages, more pror nformaton should be ncorporated. One mportant mage pror s that natural mages often contan repettve patterns and structures throughout the mage. Such non-local statstcal redundances can be very useful n enhancng the qualty of reconstructed mages. Actually, n the PCA-based adaptve sparse doman determnaton process, the non-local redundancy nformaton has been already used for tranng dataset constructon. Inspred by the success of the non-local means flterng for mage denosng [19], we further ntroduce a nonlocal self-smlarty quadratc constrant nto the super-resoluton process to fully explot the nonlocal redundances. Wth ths quadratc constrant as another regularzaton term, the mage superresoluton model n Eq. (6) s lfted to: arg mn x v v v y DHx + η x γ x + λ P x j j 1 j where η s a scalar parameter to balance the non-local regularzaton term; x v j s the j th smlar (vectorzed) block to x v n a nonlocal neghborhood; γ j s the weght assgned to x v j. We use the block matchng method to locate the smlar blocks to x v v v n a large enough wndow: x xj < T, where T s a predefned threshold. The weghts γ j depend on the smlarty between x v and x v j, whch can be calculated by: v v x xj h 1 γ j = e (8) c where h s a scalar controllng the smlarty and c s the normalzaton factor. As n secton., n Eq. (7) the selecton of smlar nonlocal neghbors and the computaton of the weghts requre that an ntal estmate of the orgnal HR mage, and then Eq. (7) can be teratvely solved. (7).4 Numercal algorthm Eq. (7) s a mnmzaton problem wth compound regularzaton terms. For the convenence of expresson, we can rewrte Eq. (7) n a matrx form as follows arg mn y DHx + η ( I A) x + λ Px (9) x 1 where P s form by all the P ; A s a matrx of dmenson of N N wth N beng the dmenson of the target HR mage. A s set as follows vm th v v vm γ j, f xn s the j smlar neghbor to xm, and j are the locatons of xm and xn Amn (, ) = (10) 0, otherwse Eq. (9) can be expressed as

5 Let arg mn X y DH + λ P η( I A) x x 0 1 (11) Snce P s orthogonal, we have y y% = 0, DH K = η( I A), u = Px (1) T x= P u. Then (11) can be rewrtten as follows arg mn u y% u λ u (13) T KP + Eq. (13) s a challengng large-scale compound l 1 -norm and l -norm mnmzaton problem. Conventonal optmzaton technques, such as steepest-descent, conjugate gradent and nteror-pont algorthms, are neffcent n solvng ths mnmzaton problem. In ths paper, we adopt the recently proposed teratve shrnkage algorthm [14] [15] to solve ths mnmzaton problem. We summarze the teratve algorthm for solvng (13) n Algorthm 1. 1 Algorthm 1 for solvng Eq. (13) 1 Intalzaton: ntalze x by settng P as the wavelet transformaton matrx; then A can be ntalzed, and set (0) u = 0. Iterate on k untl convergence ( k+ 1/) ( k) T T ( k) (a) u = u + K ( y% KP u ); ( k+ 1) ( k+ 1/) (b) u = soft( u, μ), where soft(, μ) s a soft thresholdng functon wth threshold μ. (c) If mod( km, ) = 0, update the PCA transformaton matrxes P and A n (13) usng the mproved estmate of the k + orgnal HR mage $ ( 1) T ( k+ 1) x = P u. In Algorthm 1, we update P and A n every M teratons to reduce the computatonal complexty. 3. EXPERIMENTAL RESULTS In ths secton, we conduct experments to verfy the effcency of the proposed technque for mage super resoluton. The degraded LR mages are generated by frst applyng a blur kernel and then down-samplng. The blurrng kernel n the smulatons s a 7 7 Gaussan flter wth standard devaton of 1.6. We magnfy the LR mages by a factor of 3, whch s common n the lterature of super-resoluton. We use 5 5 HR mage patches wth overlap of 1 pxel between adjacent patches when transformng HR mages nto adaptve PCA domans. These 5 5 patches are also used to locate the nonlocal smlar neghbors. For color mages, the proposed approach s only appled to the lumnance component and bcubc nterpolator s used for the chromatc components. We compare the proposed approach wth some state-of-the-art mage super-resoluton approaches, ncludng the teratve back-projecton (IBP) [4], the softcuts based method n [8], and the SR based method n [13] 1. The vsual results by these competng approaches are presented through Fg.1 ~ Fg. 3. From these fgures we can see that the IBP method reconstructs the HR mages wth jaggy and chessboard artfacts. The HR mages reconstructed by the SoftCuts based method remove most of such artfacts but they are over-smoothed and many mage detals are elmnated. The approach n [13] s compettve n vsual qualty. However, the reconstructed edges and textures by t are not smooth. Chessboard artfacts and noses can be observed n the edge regons. The reason s that t heavly reles on the tranng data and tends to generate nconsstences between adjacent mage patches. Wthout any exceptons, the proposed approach reconstructs the most vsually pleasant HR mages. The edges and textures reconstructed by our approach are much sharper and cleaner than others. Also, more mage detals are recovered by our approach. 1 We thank the authors of [8] and [13] for provdng ther code or expermental results.

6 In practcal LR magng process, nose s often ntroduced. To demonstrate the robustness of the proposed method to nose, we add the Gaussan noses wth standard devaton of 5 to the smulated LR mages. The HR mages produced by the competng approaches are shown n Fg. 4 and Fg. 5. We can see that the IBP method magnfy the nose snce the back-projecton process s very senstve to nose. The nose s mostly removed by the Softcuts and the SR-based methods n [8] and [13]. However, the mage detals are also removed n ther results. The proposed approach can well handle the super-resoluton and denosng smultaneously. As shown n Fg. 4 and Fg. 5, not only the noses are well suppressed, but also sharp edges and textures are well preserved. The PSNR results of the reconstructed mages n Fg. 1~Fg. 5 are shown n Table 1, from whch we can see that the PSNR values by the proposed method are much hgher that others. 4. CONCLUSION Ths paper presented a sparse representaton (SR) based mage super-resoluton approach by maxmzng the sparsty of the HR mages n adaptve sparse domans. The sparse doman of an mage patch was locally determned by applyng prncple component analyss (PCA) of the nonlocal smlar mage patches. The sparsty of the HR mages was enforced by an l 1 -norm regularzaton term that penalzes the prncple components of the mage sgnals. In addton, a nonlocal smlarty pror of natural mages was also ncorporated to explot the nonlocal mage redundances. The proposed objectve functon wth compound l 1 -norm and l -norm regularzaton terms can be effcently solved by an teratve shrnkage algorthm. Expermental results demonstrated that the proposed approach can reconstruct sharp edges and fne mage detals and s robust to noses. 5. ACKNOWLEDGEMENTS Ths work s supported by the Program for New Century Excellent Talents (No. NCET ), the NSF Chna (No , , and ), and the Ph.D. Program Foundaton of Mnstry of Educaton of Chna (Nos , ). (a) Orgnal (b) Input LR mage (c) Back-projecton [4] (d) SoftCuts [8] (e) Sparse representaton [13] (f) Proposed Fg. 1 Results on the Grl mage wth scalng factor 3.

7 (a) Orgnal (b) Nearest neghbor (c) Back-projecton [4] (d) SoftCuts [8] (e) Sparse representaton [13] (f) Proposed Fg. Results on the Parrot mage wth scalng factor 3. Table 1 The PSNR (db) results of the lumnance components reconstructed by dfferent methods. Images IBP [4] SoftCuts [8] [13] Proposed Grl Parrot Butterfly Nosy Parthenon Nosy Parrot

8 (a) Orgnal (d) SoftCuts [8] (b) Input LR mage (c) Back-projecton [4] (e) Sparse representaton [13] (f) Proposed Fg. 3 Results on the Butterfly mage wth scalng factor 3. (a) Orgnal (d) SoftCuts [8] (b) Input LR mage (e) Sparse representaton [13] (c) Back-projecton [4] (f) Proposed Fg. 4 Results on the nosy Parthenon mage wth scalng factor 3. The standard devaton of Gaussan nose s 5.

9 (a) Orgnal (b) Input LR mage (c) Back-projecton [4] (d) SoftCuts [8] (e) Sparse representaton [13] (f) Proposed Fg. 5 Results on the nosy Parrot mage wth scalng factor 3. The standard devaton of Gaussan nose s 5. REFERENCES 1. X. L and M. T. Orchard, New edge-drected nterpolaton IEEE Trans. Image Process., vol. 10, no. 10, pp , Oct L. Zhang and X. Wu, An edge-guded mage nterpolaton algorthm va drectonal flterng and data fuson, IEEE Trans. Image Process., vol. 15, no. 8, pp. 6-38, Aug X. Zhang and X. Wu, Image nterpolaton by adaptve D autoregressve modelng and soft-decson estmaton, IEEE Trans. Image Process., vol. 17, no. 6, pp , Jun M. Iran and S. Peleg, Moton analyss for mage enhancement: resoluton, occluson and transparency, JVCI., S. Da, M. Han, Y. Wu, and Y. Gong, Blateral back-projecton for sngle mage super resoltuon n Proc. Int. Conf. on Multmeda and Expo, 007, pp , July W. Dong, L. Zhang, G. Sh, and X. Wu, Nonlocal back-projecton for adaptve mage enlargement, n Proc. Int. Conf. Image Process.,009, Oct S. Farsu, M. D. Robnson, M. Elad, and P. Mlanfar, Fast and robust multframe super-resoluton, IEEE Trans. Image Process., vol. 15, no. 1, pp , Jan S. Da, M. Han, W. Xu, Y. Wu, Y. Gong, and A. K. Katsaggelos, SoftCuts: a soft edge smoothness pror for color mage super-resoluton, IEEE Trans. Image Process., vol. 18, no. 5, pp , May J. Sun, J. Sun, Z. Xu, and H. Shum, Image super-resoluton usng gradent profle pror, n Proc. IEEE Computer Vson and Pattern Recognton, pp. 1-8, Jun J. Ca, H. J, C. Lu and Z. Shen, Blnd moton deblurrng from a sngle mage usng sparse approxmaton, n Proc. IEEE Computer Vson and Pattern Recognton, Jun. 009.

10 11. M. Elad and M. Aharon, Image denosng va sparse and redundant representatons over learned dctonares, IEEE Trans. Image Process., vol. 15, no. 1, pp , Dec E. J. Candes and M. B. Wakn, An ntroducton to compressve samplng, IEEE Sgnal Process. Mag., vol. 5, no., pp. 1-30, Mar J. Yang, J. Wrght, Y. Ma, and T. Huang, Image super-resoluton as sparse representaton of raw mage patches, IEEE Computer Vson and Pattern Recognton, vol. 1, pp. 1-8, Jun I. Daubeches, M. Defrse, and C. De. Mol, an teratve thresholdng algorthm for lnear nverse problems wth sparsty constrant, Commu. On Pure and Appled Math., vol. 57, no. 11, pp , Aug P. Combettes, and V. Wajs, Sgnal recovery by proxmal forward-backward splttng, SIAM J. Multscale Model Smul., vol. 4, pp , M. Aharon, M. Elad, and A. Brucksten, K-SVD: an algorthm for desgnng overcompleter dctonares for sparse representaton, IEEE Trans. Sgnal Process., vol. 54, no. 11, pp , Nov L. Zhang, R. Lukac, X. Wu, and D. Zhang, PCA-based spatally adaptve denosng of CFA mages for snglesensor dgtal cameras, IEEE Trans. Image Process., vol. 18, no. 4, pp , Apr L. Zhang, W. Dong, D. Zhang, and G. Sh, Two-stage mage denosng by prncpal component analyss wth local pxel groupng, to appear n Pattern Recognton. 19. A. Buades, B. Coll, and J. M. Morel, A revew of mage denosng algorthms, wth a new one, Multsc. Model. Smulat., vol. 4, no., pp , 005.

Image Deblurring and Super-resolution by Adaptive Sparse Domain Selection and Adaptive Regularization

Image Deblurring and Super-resolution by Adaptive Sparse Domain Selection and Adaptive Regularization Image Deblurrng and Super-resoluton by Adaptve Sparse Doman Selecton and Adaptve Regularzaton Wesheng Dong a,b, Le Zhang b,1, Member, IEEE, Guangmng Sh a, Senor Member, IEEE, and Xaoln Wu c, Senor Member,

More information

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

Nonlocally Centralized Sparse Representation for Image Restoration

Nonlocally Centralized Sparse Representation for Image Restoration Nonlocally Centralzed Sparse Representaton for Image Restoraton Wesheng Dong a, Le Zhang b,1, Member, IEEE, Guangmng Sh a, Senor Member, IEEE, and Xn L c, Senor Member, IEEE a Key Laboratory of Intellgent

More information

RECONSTRUCTING a high quality image from one or

RECONSTRUCTING a high quality image from one or 1620 IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 22, NO. 4, APRIL 2013 Nonlocally Centralzed Sparse Representaton for Image Restoraton Wesheng Dong, Le Zhang, Member, IEEE, Guangmng Sh, Senor Member, IEEE,

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

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

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

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

Joint Example-based Depth Map Super-Resolution

Joint Example-based Depth Map Super-Resolution Jont Example-based Depth Map Super-Resoluton Yanje L 1, Tanfan Xue,3, Lfeng Sun 1, Janzhuang Lu,3,4 1 Informaton Scence and Technology Department, Tsnghua Unversty, Bejng, Chna Department of Informaton

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

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

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

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

A Fast Visual Tracking Algorithm Based on Circle Pixels Matching

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

More information

SLAM Summer School 2006 Practical 2: SLAM using Monocular Vision

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

More information

A Study on Clustering for Clustering Based Image De-Noising

A Study on Clustering for Clustering Based Image De-Noising Journal of Informaton Systems and Telecommuncaton, Vol. 2, No. 4, October-December 2014 196 A Study on Clusterng for Clusterng Based Image De-Nosng Hossen Bakhsh Golestan* Department of Electrcal Engneerng,

More information

AS a classical problem in low level vision, image denoising. Group Sparsity Residual Constraint for Image Denoising

AS a classical problem in low level vision, image denoising. Group Sparsity Residual Constraint for Image Denoising 1 Group Sparsty Resdual Constrant for Image Denosng Zhyuan Zha, Xnggan Zhang, Qong Wang, Lan Tang and Xn Lu arxv:1703.00297v6 [cs.cv] 31 Jul 2017 Abstract Group-based sparse representaton has shown great

More information

Wavefront Reconstructor

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

More information

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

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

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

IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. *, NO. *, Dictionary Pair Learning on Grassmann Manifolds for Image Denoising

IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. *, NO. *, Dictionary Pair Learning on Grassmann Manifolds for Image Denoising IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. *, NO. *, 2015 1 Dctonary Par Learnng on Grassmann Manfolds for Image Denosng Xanhua Zeng, We Ban, We Lu, Jale Shen, Dacheng Tao, Fellow, IEEE Abstract Image

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

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

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

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

Novel Fuzzy logic Based Edge Detection Technique

Novel Fuzzy logic Based Edge Detection Technique Novel Fuzzy logc Based Edge Detecton Technque Aborsade, D.O Department of Electroncs Engneerng, adoke Akntola Unversty of Tech., Ogbomoso. Oyo-state. doaborsade@yahoo.com Abstract Ths paper s based on

More information

Texture Enhanced Image Denoising via Gradient Histogram Preservation

Texture Enhanced Image Denoising via Gradient Histogram Preservation 203 IEEE Conference on Computer Vson and Pattern Recognton Texture Enhanced Image Denosng va Gradent Hstogram Preservaton Wangmeng Zuo,2 Le Zhang 2 Chunwe Song Davd Zhang 2 Harbn Insttute of Technology,

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

S INGLE1 image super resolution is the process of reconstructing

S INGLE1 image super resolution is the process of reconstructing Sngle Image Interpolaton va Adaptve Non-Local Sparsty-ased Modelng Yanv Romano, Matan Protter, and Mchael Elad, Fellow, IEEE Abstract Sngle mage nterpolaton s a central and extensvely studed problem n

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

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

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

More information

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

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

PCA Based Gait Segmentation

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

More information

Review of approximation techniques

Review of approximation techniques CHAPTER 2 Revew of appromaton technques 2. Introducton Optmzaton problems n engneerng desgn are characterzed by the followng assocated features: the objectve functon and constrants are mplct functons evaluated

More information

IMAGE super-resolution (SR) [2] is the problem of recovering

IMAGE super-resolution (SR) [2] is the problem of recovering 1 SoftCuts: A Soft Edge Smoothness Pror for Color Image Super-Resoluton Shengyang Da*, Student Member, IEEE, Me Han, We Xu, Yng Wu, Senor Member, IEEE, Yhong Gong, Aggelos K. Katsaggelos, Fellow, IEEE.

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

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

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

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

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

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

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

Lecture 4: Principal components

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

More information

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

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

Nonlocal Mumford-Shah Model for Image Segmentation

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

More information

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

Classifying Acoustic Transient Signals Using Artificial Intelligence

Classifying Acoustic Transient Signals Using Artificial Intelligence Classfyng Acoustc Transent Sgnals Usng Artfcal Intellgence Steve Sutton, Unversty of North Carolna At Wlmngton (suttons@charter.net) Greg Huff, Unversty of North Carolna At Wlmngton (jgh7476@uncwl.edu)

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

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

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

More information

Fuzzy Filtering Algorithms for Image Processing: Performance Evaluation of Various Approaches

Fuzzy Filtering Algorithms for Image Processing: Performance Evaluation of Various Approaches Proceedngs of the Internatonal Conference on Cognton and Recognton Fuzzy Flterng Algorthms for Image Processng: Performance Evaluaton of Varous Approaches Rajoo Pandey and Umesh Ghanekar Department of

More information

A Computer Vision System for Automated Container Code Recognition

A Computer Vision System for Automated Container Code Recognition A Computer Vson System for Automated Contaner Code Recognton Hsn-Chen Chen, Chh-Ka Chen, Fu-Yu Hsu, Yu-San Ln, Yu-Te Wu, Yung-Nen Sun * Abstract Contaner code examnaton s an essental step n the contaner

More information

An Application of the Dulmage-Mendelsohn Decomposition to Sparse Null Space Bases of Full Row Rank Matrices

An Application of the Dulmage-Mendelsohn Decomposition to Sparse Null Space Bases of Full Row Rank Matrices Internatonal Mathematcal Forum, Vol 7, 2012, no 52, 2549-2554 An Applcaton of the Dulmage-Mendelsohn Decomposton to Sparse Null Space Bases of Full Row Rank Matrces Mostafa Khorramzadeh Department of Mathematcal

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

CS246: Mining Massive Datasets Jure Leskovec, Stanford University

CS246: Mining Massive Datasets Jure Leskovec, Stanford University CS46: Mnng Massve Datasets Jure Leskovec, Stanford Unversty http://cs46.stanford.edu /19/013 Jure Leskovec, Stanford CS46: Mnng Massve Datasets, http://cs46.stanford.edu Perceptron: y = sgn( x Ho to fnd

More information

SVM-based Learning for Multiple Model Estimation

SVM-based Learning for Multiple Model Estimation SVM-based Learnng for Multple Model Estmaton Vladmr Cherkassky and Yunqan Ma Department of Electrcal and Computer Engneerng Unversty of Mnnesota Mnneapols, MN 55455 {cherkass,myq}@ece.umn.edu Abstract:

More information

INTER-BLOCK CONSISTENT SOFT DECODING OF JPEG IMAGES WITH SPARSITY AND GRAPH-SIGNAL SMOOTHNESS PRIORS

INTER-BLOCK CONSISTENT SOFT DECODING OF JPEG IMAGES WITH SPARSITY AND GRAPH-SIGNAL SMOOTHNESS PRIORS INTER-BLOCK CONSISTENT SOFT DECODING OF IMAGES WITH SPARSITY AND GRAPH-SIGNAL SMOOTHNESS PRIORS Xanmng Lu 1,2, Gene Cheung 2, Xaoln Wu 3, Debn Zhao 1 1 School of Computer Scence and Technology, Harbn Insttute

More information

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

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

More information

Hybrid Non-Blind Color Image Watermarking

Hybrid Non-Blind Color Image Watermarking Hybrd Non-Blnd Color Image Watermarkng Ms C.N.Sujatha 1, Dr. P. Satyanarayana 2 1 Assocate Professor, Dept. of ECE, SNIST, Yamnampet, Ghatkesar Hyderabad-501301, Telangana 2 Professor, Dept. of ECE, AITS,

More information

Edge Detection in Noisy Images Using the Support Vector Machines

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

More information

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

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

More information

Fitting: Deformable contours April 26 th, 2018

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

More information

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

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

More information

NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS

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

More information

RESOLUTION ENHANCEMENT OF SATELLITE IMAGES USING DUAL-TREE COMPLEX WAVELET AND CURVELET TRANSFORM

RESOLUTION ENHANCEMENT OF SATELLITE IMAGES USING DUAL-TREE COMPLEX WAVELET AND CURVELET TRANSFORM Avalable Onlne at www.csmc.com Internatonal Journal of Computer Scence and Moble Computng A Monthly Journal of Computer Scence and Informaton Technology IJCSMC, Vol. 3, Issue. 4, Aprl 2014, pg.1315 1320

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

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

Classification / Regression Support Vector Machines

Classification / Regression Support Vector Machines Classfcaton / Regresson Support Vector Machnes Jeff Howbert Introducton to Machne Learnng Wnter 04 Topcs SVM classfers for lnearly separable classes SVM classfers for non-lnearly separable classes SVM

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

L 1 -FUSION: ROBUST LINEAR-TIME IMAGE RECOVERY FROM FEW SEVERELY CORRUPTED COPIES. Panos P. Markopoulos, Sandipan Kundu, and Dimitris A.

L 1 -FUSION: ROBUST LINEAR-TIME IMAGE RECOVERY FROM FEW SEVERELY CORRUPTED COPIES. Panos P. Markopoulos, Sandipan Kundu, and Dimitris A. L -FUSION: ROBUST LINEAR-TIME IMAGE RECOVERY FROM FEW SEVERELY CORRUPTED COPIES Panos P. Markopoulos, Sandpan Kundu, and Dmtrs A. Pados Department of Electrcal Engneerng, The State Unversty of New York

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

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

IMAGE FUSION TECHNIQUES

IMAGE FUSION TECHNIQUES Int. J. Chem. Sc.: 14(S3), 2016, 812-816 ISSN 0972-768X www.sadgurupublcatons.com IMAGE FUSION TECHNIQUES A Short Note P. SUBRAMANIAN *, M. SOWNDARIYA, S. SWATHI and SAINTA MONICA ECE Department, Aarupada

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

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

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

Contourlet-Based Image Fusion using Information Measures

Contourlet-Based Image Fusion using Information Measures Proceedngs of the 2nd WSEAS Internatonal Symposum on WAVELETS THEORY & APPLICATIONS n Appled Mathematcs, Sgnal Processng & Modern Scence (WAV '08), Istanbul, Turkey, May 2730, 2008 ContourletBased Image

More information

Classifier Selection Based on Data Complexity Measures *

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

More information

A PATTERN RECOGNITION APPROACH TO IMAGE SEGMENTATION

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

More information

Towards Semantic Knowledge Propagation from Text to Web Images

Towards Semantic Knowledge Propagation from Text to Web Images Guoun Q (Unversty of Illnos at Urbana-Champagn) Charu C. Aggarwal (IBM T. J. Watson Research Center) Thomas Huang (Unversty of Illnos at Urbana-Champagn) Towards Semantc Knowledge Propagaton from Text

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

Lecture 13: High-dimensional Images

Lecture 13: High-dimensional Images Lec : Hgh-dmensonal Images Grayscale Images Lecture : Hgh-dmensonal Images Math 90 Prof. Todd Wttman The Ctadel A grayscale mage s an nteger-valued D matrx. An 8-bt mage takes on values between 0 and 55.

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

Lecture #15 Lecture Notes

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

More information

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

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

4580 IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 25, NO. 10, OCTOBER 2016

4580 IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 25, NO. 10, OCTOBER 2016 4580 IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 25, NO. 0, OCTOBER 206 Sparse Representaton Wth Spato-Temporal Onlne Dctonary Learnng for Promsng Vdeo Codng Wenru Da, Member, IEEE, Yangme Shen, Xn Tang,

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

Brushlet Features for Texture Image Retrieval

Brushlet Features for Texture Image Retrieval DICTA00: Dgtal Image Computng Technques and Applcatons, 1 January 00, Melbourne, Australa 1 Brushlet Features for Texture Image Retreval Chbao Chen and Kap Luk Chan Informaton System Research Lab, School

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

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

Categories and Subject Descriptors B.7.2 [Integrated Circuits]: Design Aids Verification. General Terms Algorithms

Categories and Subject Descriptors B.7.2 [Integrated Circuits]: Design Aids Verification. General Terms Algorithms 3. Fndng Determnstc Soluton from Underdetermned Equaton: Large-Scale Performance Modelng by Least Angle Regresson Xn L ECE Department, Carnege Mellon Unversty Forbs Avenue, Pttsburgh, PA 3 xnl@ece.cmu.edu

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

Image Deblurring Using Adaptive Sparse Domain Selection and Adaptive Regularization

Image Deblurring Using Adaptive Sparse Domain Selection and Adaptive Regularization Volume 3, No. 3, May-June 2012 International Journal of Advanced Research in Computer Science RESEARCH PAPER Available Online at www.ijarcs.info ISSN No. 0976-5697 Image Deblurring Using Adaptive Sparse

More information

A Newton-Type Method for Constrained Least-Squares Data-Fitting with Easy-to-Control Rational Curves

A Newton-Type Method for Constrained Least-Squares Data-Fitting with Easy-to-Control Rational Curves A Newton-Type Method for Constraned Least-Squares Data-Fttng wth Easy-to-Control Ratonal Curves G. Cascola a, L. Roman b, a Department of Mathematcs, Unversty of Bologna, P.zza d Porta San Donato 5, 4017

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

Generalized Video Deblurring for Dynamic Scenes

Generalized Video Deblurring for Dynamic Scenes Generalzed Vdeo Deblurrng for Dynamc Scenes Tae Hyun Km and Kyoung Mu Lee Department of ECE, ASRI, Seoul Natonal Unversty, 151-742, Seoul, Korea {llger9, kyoungmu}@snu.ac.kr, http://cv.snu.ac.kr However,

More information

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

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

More information