Polyhedron Volume-Ratio-based Classification for Image Recognition

Size: px
Start display at page:

Download "Polyhedron Volume-Ratio-based Classification for Image Recognition"

Transcription

1 Polyhedron Volume-Ratio-based Classifiation for Image Reognition Qingxiang Feng, Jeng-Shyang Pan, Senior Member, IEEE, Jar-Ferr Yang, Fellow, IEEE and Yang-ing Chou Abstrat In this paper, a novel method, alled polyhedron volume ratio lassifiation (PVRC), is proposed for image reognition. he linear regression based lassifiation (LRC) and lass mean (CM) approahes aim to find the losest projetion to the subspae formed by the prototype data vetors in eah lass. Better than LRC and CM lassifiers, the PVRC lassifier omputes the ratio of two polyhedron s volumes, where the first inlusive polyhedron is enlosed by the test plus lass prototype data vetors while the seond exlusive polyhedron is enlosed by lass prototype data vetors. he PVRC tries to find the least ratio of lass-based inlusive and exlusive polyhedron volumes among all lasses. With mathematial optimization, the PVRC lassifier ahieves better reognition rate than the existing statistial methods for objet and fae reognition. With the proposed fast algorithm, it is noted that the omputational ost of the proposed lassifier is very small. A large number of experiments on Coil00 objet database, Eth80 objet database, Soil47 objet database, G fae database and UMIS fae database are used to evaluate the proposed algorithm. he experimental results demonstrate that the proposed method ahieves better reognition rate and less omputational ost than the existing methods, suh as LRC, CM, nearest neighbor (NN), sparse representation based lassifiation (SRC) and two phase test sample sparse representation (PSSR) lassifiers. Index erms Fae Reognition, Objet Reognition, Linear Regression Classifiation, Sparse Representation based Classifiation, Nearest Neighbor. F I. INRODUCION ACE and objet reognition systems are known to be ritially dependent on lassifiers. Nearest neighbor (NN) [] and lass mean (CM) [] methods, whih are designed to minimize the distane from the test vetor to the subspae spanned by all training vetors of eah objet lass, are the well-known approahes in pattern reognition area. he NN lassifies the test sample based on the best representation in terms of a single training sample, while the CM lassifies it based on the best mean representation in terms of all the training samples in eah lass. Samples from a speifi objet lass are known to lie on a linear subspae [3]-[4]. Borrowing the above subspae onept, the loally linear regression (LLR) [] is proposed speifially to solve the problem of pose identifiation. Linear regression-based lassifiation (LRC) [], whih is proposed for the problem of fae identifiation, uses the similar onept to develop lass-speifi models of the registered users and renovate the task of fae reognition as a problem of linear regression. For fae reognition, the LRC approahes, inluding kernel LRC, Improved PCA-LRC, LDA-LRC and Unitary-LRC [7]-[0], have been proposed to further improve the reognition performane under different situations suh as variable illumination and faial expression. Instead of the lass-model suggested in the LRC approahes, sparse representation based lassifiation (SRC) [-] uses all-lass-model to lassify the test sample. Based on the SRC lassifier, some improved methods are further presented for fae reognition. hese improved lassifiers ould be lassified into two ategories. he first ategory utilizes the novel representations of eah lass, in whih all use the first phase to hoose some prototype samples while the seond phase is different [3]-[6]. Chang et. al utilizes the ollaborative representation instead of the sparse representation [7]. he seond ategory uses the sparse representation for the subspae learning/disriminant analysis, whih employs the disriminant projetion tensor disriminant projetion [8], [9] or the sparse eigenfae [0] to selet the features. SRC lassifier gains better performane for fae reognition under variations of illumination, noise, and orruption. However, SRC lassifier has two drawbaks. Firstly, SRC lassifier ould not diretly apply for objet reognition [4]-[7] and fae gesture. Seondly, the SRC lassifier aquires large omputational omplexity. hus, we need a low omputation lassifier, whih an be used for fae and objet reognition, In this paper, the polyhedron volume ratio lassifiation (PVRC) method is proposed for image reognition [7]. With the minimum distane metri, the LRC and CM lassifiers test the distane between the test sample vetor and eah lass subspae, then the losest distane is identified as the best mathed lass. With the similar lass-based lassifiation, the PVRC lassifier tests the ratio of the polyhedron volume enlosed by the test sample plus the lass prototype samples over the polyhedron volume enlosed by the lass prototype samples. he objetive of LRC and PVRC are oneptually same, however, the differene between PVRC and LRC an be found in Setion V-B. With Cayley-Menger matrix [], in this paper, we first define the polyhedron volume and propose a new metri, alled polyhedron volume ratio for inlusive and exlusive test sample to a speifi lass-subspae, for image

2 lassifiation. Based on mathematial (linear algebra) derivations, the PVRC is separated into training phase and test phase (P&P), where the omputational ost in the test phase of PVRC is very small. In the test phase, the omputation of the PVRC is almost equal to that of NN lassifier and is muh less than those of SRC, CRC and LRC. he rest of the paper is organized as the follows. First, we review several well-known image lassifiers in Setion II. he omputation of polyhedron volume and the PVRC lassifier will be introdued in Setion III. o redue the omputation, the fast PVRC lassifiation proedure is desribed in Setion IV. he analyses of the proposed PVRC are disussed in Setion V. In Setion VI, a number of experiments to show the effetiveness of the proposed lassifier are present. he onlusions and future work are finally addressed in Setion VII. II. REVIEW OF CLASS-SPECIFIC CLASSIFICAIONS Let Y { yi, i,,..., N,,,..., M} R denote the prototype image set, where denotes the i th prototype image of the th lass, M denotes the number of lasses, and is the number of prototypes belonging to the th lass. y i II-A Linear Regression Classifiation Algorithm For linear regression, eah a b image is transformed to vetor ab q by olumn onatenation as y i R xi R, where q = a b. By using the onept that data vetors from the same lass lie on a linear subspae, the LRC develops a lass-speifi model by staking the q-dimensional image vetors as X X [ x x... x ] R qn N N. () Let y be an unlabeled a b test image and our problem is to lassify y as one of the lasses. Similarly, we an transform y into the vetor form x R q. If y belongs to the th lass, it shall be well represented as a linear ombination of the training images of the lass as ( x x X X X ) X x. () X X If X X is singular in (), will be replaed by X X 0.0I, where I is a unit matrix. In (), the predited vetor an be treated as the projetion of x onto the th lass subspae. he LRC now alulates the distane measure between the predited response vetor x and the original response vetor x as d( x) x x, (3) x where * means L -norm. he LRC lassifiation rule is in favor of the lass with the minimum distane of the lass as min * d ( x),,,..., M. (4) II-B. Collaborative representation based lassifiation (CRC) Suppose that we have M lasses of subjets, we an ollet all lass-speifi models X, =,,, M, defined in () to form the omplete data model as qmn X [ X X... X M ] R. (5) If the vetor of all-lass weighing parameters is denoted as R MN, it an be alulated as follows. ( X X ) X x. (6) For (6), if is singular, will be replaed by X X 0.0I. he regularized residual of the th lass, is given as x X r, (7) X X X X where orresponding to the oeffiient of the sample of lass is a setioned olumn of. And the CRC lassifiation rule in favor of the lass with the minimum distane beomes min * r,,,..., M. (8) II-C Sparse representation based lassifiation (SRC) For the sparse representation-based lassifiation (SRC), we first normalize the olumns of X stated into (5) to have unit L -norm. We an solve the L -norm minimization problem as: gˆ argmin g subjet to Xg x. (9) Compute the regularized residuals g r x X ˆ g, (0) the SRC lassifiation rule in favor of the lass with the minimum distane an be expressed by minr,,,..., M. () * r as, III. POLYHEDRON VOLUME RAIO CLASSIFICAION Before the introdution of the polyhedron volume ratio lassifiation (PVRC), we first review the onept of polyhedron volume of n sample points. In [], Cayley-Menger matrix onstruted by n (n ) sample points is given by: where 0 ( b ) ( b3)... ( b n) ( b) 0 ( b3)... ( bn ) ( ) ( ) 0... ( ) b3 b3 b3n Q n, () ( bn) ( bn ) ( bn3) b ij denotes Eulidian distane of ij b ji i j x i and x j as b x x. (3) he squared polyhedron volume of n sample points an be defined as the determinant of Cayley-Menger matrix as, with n n n v det( Q ), for n, (4) n ( ) n, (5) n (( n )!) r

3 where det(*) denotes the determinant of the argument and n represents the unifying fator in the omputation of polyhedron volume. o verify (4), we an hek the following interesting examples. For n =, the Cayley-Menger matrix in () for two sample points, i.e., x and x, is given as 0 ( b ) Q n ( b) 0. (6) 0 he determinant of Q is Q ( b) and = / with b x x. It is interested that the volume of -point polyhedron is the true distane between these two points. Aording to Heron s formula [], the true area of triangle of three points,,, and, an be expressed by S x x x [( b) ( b3) ( b3) ( b) ( b3) ( b3) 3 3 ( b ) ( b ) ( b ) ( b ) ( b ) ( b ) ]/6 (7) 3 3 Similarly, the Cayley-Menger matrix of three points, and is given as x 3 he determinant of Q x, x 0 ( b ) ( b3) ( b) 0 ( b3) Q 3. (8) ( b ) ( ) 0 3 b3 0 Q 3 stated in (8) is given as [( b) ( b3) ( b3) ( b) ( b3) ( b3) ( b) ( b3) ( b) ( b3) ( b3) ( b3) ]. (9) From (5), 3 = /6 and (9), the polyhedron volume defined in (4) is equal to the true area of triangle enlosed by three points, x, x, and x 3 stated in (7). Similarly, we an find 4-point Cayley-Menger matrix as 0 ( b ) ( b3) ( b4) ( b) 0 ( b3) ( b4) Q ( ) ( ) 0 ( ) 4 b3 b3 b. (0) 34 ( b4) ( b4) ( b43) 0 0 With 4=/88, the square volume of tetrahedron beomes 4 Q v det( 4 )/ 88. In [5], the omputation of n-point polyhedron volume an be interpreted as the iterative omputation of the shortest distane to the (n)-point surfae while we treat (n)-point polyhedron volume as the base volume. he multipliation of the shortest distane to the base (n)-point volume will be the n-point polyhedron volume. If the number of sample data points are less than the dimension of data vetors, it an be easily, inferred that the ratio of the n-point (with one test sample) polyhedron volume over the (n)-point polyhedron volume (with samples in a lass) will be the perpendiular distane of the test sample to the (n)-point polyhedron. Based on the above interpretation of n-point and (n)-point polyhedron volumes, we introdue a new lassifier, alled polyhedron volume ratio lassifiation (PVRC) method. For images with a b pixels, they are originally represented ab as yi R by labeling the th lass while y is an unlabeled test image. In the training phase, these (n) prototype images are transformed to prototype vetors by onatenation of olumns ab q as y i R xi R, with q=a b for i=, 3,, n. hus, eah lass ontains (n) prototype samples, whih is { x, x3,, x } q n R for =,,, M. he test image y is also transferred to a test vetor x and renamed as x =. hus, the metri of PVRC an be omputed as v n R q x / n )] / n vn [ n det( Q, () vn [ n det( Q )] where denotes the polyhedron volume of the test sample and (n) prototype samples of the -lass and represents the polyhedron volume of (n) prototype samples of the -lass. he deision rule of PVRC is simply to find the least ratio as vn *= min,,,..., M. () he above minimum polyhedron volume ratio means the least perpendiular distane between the test sample and the *-lass. However, the omputation of diret PVRC method involving the alulation of two determinates is too high for pratial appliations. hus, a fast omputation method for the PVRC is needed. IV. FAS COMPUAION MEHOD AND DEAILED CLASSIFICAION PROCEDURE OF PVRC he PVRC lassifier omputes the ratio of two polyhedron volumes, where the first polyhedron is enlosed by the test sample with (n) prototype images in eah lass and the seond polyhedron is enlosed by the (n) prototype images in eah lass. o redue the omputation of the PVRC method, we should first explore the matrix segmentation and its determinant omputation in heorem. heorem : Matrix Segmentation and Its Determinant Computation Let P be a square ( m n) ( m n) matrix to be segmented into four submatries, A, B, C and D are with the sizes of m m, m n, n m and n n, respetively. If D is a reversible square matrix, we an have the equality as, det( ) det A B P det( A BD C) det( D) C D, (3) A B where P C D. he detailed proof of heorem an be found in Appendix A.

4 o further explore the relationship of the first and seond polyhedron volumes, we an first divide Q n into four sub matries as where and Q n 0 B Q B n, (4) 0 ( b 3) ( b n ) ( b3) 0 ( b3n ) Q n, (5) ( bn ) ( bn3) n B ( b ) ( b ) ( b ), (6) From (3), we know that the omputation of determinant of Q n beomes as det( Q n ) n det( Q )det( B ( Q ) B ). (7) n By substituting (7) into (), the omputation of the PVRC metri an be simply expressed by v / n ( n ) det( B ( Qn ) B ). (8) vn It is noted that B ( Qn ) B is a salar, the determinant operation will not hange it. One we pre-ompute ( ) Q n in advane, the omputation of ratio of two polyhedron volumes stated in (8) is less than any of the existing subspae-based lassifiations. Finally, the lassifiation proedure of the PVRC using fast omputation method an be addressed as follows. PVRC raining Phase: By using (5), in the PVRC training phase, we should use all (n) prototype samples, { x, x3,, xn} to ompute and store ( Q for =,,, M; n ) PVRC esting Phase: For any test sample, x, we should use all prototype samples { x, x3,, x } q n R for =,,, M to. ompute all distanes of the test sample to all prototype samples as: b i x xi, for i=, 3,, n and =,,, M;. onstrut B by using (6) for =,,, M; 3. ompute absolute value of square volume ratio as det( B ( Qn ) B ) ; (9) 4. pik *= min,,,..., M. * Atually, for true polyhedron volume, we need to inlude unifying fator n and square root operation as (5). In Step 3, the absolute value of square volume ratio with the fator related n is intentionally ignored due to fixed n. Besides, the square root operation is also skipped without affeting the minimization proess. When the numbers of samples in the lasses are various, we should further inlude a fator of 4( n ) 4 in (9) for orret lassifiation. It is noted that the prototype samples are very similar or the number of prototype samples is less than the dimension of sample vetors, the linear regression solutions stated in (), (6), (9) might not be stable due to the possible singular problem in matrix inversion. o solve the singular problems, we an add a small identity matrix for most linear regression problem. For the PVRC omputation, involving either (5) or (7), the singularity of matrix inversion only exists for similar prototype samples. Without sarifying the detetion performane, the addition of a small identify matrix to Qn V. ANALYSIS OF PVRC CLASSIFIER is also possible. In this setion, we analyze the lassifiation rule of PVRC lassifier firstly. hen, the interpretation among the PVRC, CM, NN, LRC, CRC, PSSR and SRC is introdued. In the last, the omputational ost of PVRC is desribed. V-A. Classifiation rule of PVRC From (8) and (9), we know that the deision rule of the PVRC is based on two matrixes and ( Qn ). he matrix ( Qn ) is unrelated to the test sample while the matrix is formed by the distanes between the test sample and the prototype samples of eah lass subspae. So, we learn that the lassifiation rule of the PVRC an be seen as the ombination of the distanes between the test sample and the samples of eah lass subspae. hat is to say, the PVRC lassifier an be treated as the extension of the NN lassifier. However, the ombination distane metri of PVRC lassifier is superior to the single distane metri of NN lassifier. V-B. Computation of PVRC he most omputational ost of the PVRC lassifier will be in the training phase by onstruting Qn and omputing its matrix inversion, ( Qn ) for all lasses. he omputational omplexity will be in the order of O(n 3 ) for eah lass. he omputational ost in the testing phase of the PVRC lassifier is very small. he detailed omputational omplexity of the testing phase of the PVRC lassifier is disussed as follows. From (9), the omputation in the testing phase mainly ontains two parts. he first part is the omputation ost in onstrution of the vetor, B stated in (6). If dimensionality of the sample vetor is with q, Its omputational omplexity is about in the order of O(nq) for eah lass, he total omputation omplexity will be in the order of O(nq)+O(n ). Sine the size of sample vetors, q is muh greater than the number of sample vetors, n, hus, we an say that the omputational omplexity of the PVRC lassifier is approximately equal to in the order of O(nq), i.e., the ost in onstrution of the vetor B. It is easy to know that the omputation in onstituting the matrix B is approximately equal to the ost of the NN lassifier. In summary, the omputation ost of the PVRC is approximately equal to the B B

5 ost of NN lassifier, whih is muh less than the osts of the SRC, CRC, LRC and PSSR lassifiers. he detailed advantages in various lassifiation performanes of the PVRC lassifier will be verified by experimental results. V-C Interpretation of well-known lassifiers Fig. also shows the onept of the other lassifiation methods based on lass linear regression and lass mean. he NN lassifier uses the distane between the test sample and eah prototype sample to lassify the test sample. he CM lassifier uses the distane between the test sample and the mean sample of eah lass to lassify the test sample. he other lassifiers suh as the CRC, PSSR and SRC lassifiers try to find the better approximations of the subspae of the lass to ahieve better reognition performanes. hough the NN, CM, LRC, CRC, PSSR and SRC lassifier are different, they have one thing in ommon, whih is that they all need to ompute the distane between the test sample and the approximated lass enter or the subspae, whih is derived from the prototype samples or the re-sampled prototype samples, for lassifiations. he performanes of the NN, CM, LRC, CRC, PSSR and SRC lassifiers only depend on the loseness of their approximated lass enters or subspaes. Approximated Subspae st lass Class Mean x Approximated Subspae x i m Approximated Subspae m x x x i nd lass Class Mean x est sample m x i th lass Class Mean : lass samples : lass mean : linear subspae Fig.. Coneptual relationships among the proposed PVRC and the projetion-based lassifiation methods V-D. Comparisons of PVRC and LRC lassifiers he ratio of polyhedron volumes with/without extra the test sample to the lass samples an be interpreted as the omputation of the perpendiular distane to the prototype samples. From Fig., the PVRC find the minimum distane from the polyhedron of the enlosed lass data vetors while the LRC find the least distane from spanned subspae formed by lass data vetors, he differenes between the LRC and the PVRC are addressed as follows. From omputation point of view, the PVRC is different from the LRC, where the former is derived through a mathematial (linear algebra) approah as stated in () while the latter is obtained from a statistial way to minimize the statistial preditin error as addressed in (). By simulations, Fig. shows the ratio of the omputed distanes obtained by the PVRC and the LRC if they use the same randomly-generated data points. o ompute the ratio between PVRC s distane and LRC s distane, the orresponding MALAB ode an be downloaded from [ he ratio between the PVRC s distane and LRC s distane is variation from 5 to 9. herefore, we learnt that the results of PVRC and LPC are different. Fig.. he ratio of PVRC s and LRC s distanes for 300 tests. For omputation analyses, if we separate the LRC into training phase and test phase (P&P) suh that we ould ahieve a faster lassifiation. In the test phase, the omputational omplexity of the LRC is O(Q ) while the PVRC and NN is about O(QN ). hus, the LRC aquires muh more omputation than the PVRC and NN lassifiers sine Q >> N. he similar results also verified from ode [ that the omputational time of PVRC is almost equal to that of the NN, whih is muh simpler than the original LRC and P&P LRC. Besides, the LRC needs about O(MQQ) storage spae to store pre-omputed results, while the PVRC needs about O(MN N ) storage spae. VI. EXPERIMENAL RESULS he lassifiation performanes of the PVRC lassifier are ompared to those of the LRC, SRC, CRC, PSSR, CM and NN lassifiers. wo test shemes are taken for omparisons. Firstly leave one out sheme: All images within prototype database are taken as the test samples. When an image is used as the test sample, it is not used as a prototype and it is removed from the prototype set during the lassifiations; Seondly First N sheme: the first N fae images of eah lass are used as the prototype set. he rest fae images of test database are used as test samples. he reognition rate (RR) is used to evaluate the performane of new algorithms. All experiments are implemented using the MALAB R009a under Intel (R) Core (M) i5 CPU 760 with a lok speed of.80ghz,.80ghz and 3GB RAM.

6 VI-A Computational Cost he omputational ost of the lassifiation proedure of several lassifier is depended on the number of samples, is not depended on the speifi value of samples. So, we only need provide the result of one database. he other databases are similar. In able I, the omputational ost of eah query of several lassifiers on oil00 objet database [3] is desribed, whih is orresponding to the run time of the seond experiments in the part B. From able I, after omputer MALAB simulations, we learn that the omputational osts (unit: seonds) of the PVRC, NN and CM lassifiers are similar, whih are less than those of the CRC and LRC lassifiers and muh less than that of the SRC lassifier. It is noted that the omputational ost of the first phase of the PSSR is similar to that of CRC, the omputational ost of the seond phase of the PSSR is aording to the number of samples hosen in the first phase. So, the total omputational ost of the PSSR is more than that of the CRC, whih is also more than that of the proposed PVRC. the prototypes. From Fig.4, we an know that the RR of the PVRC lassifier outperforms the best RR of the PSSR with 0.% (8.50%) improvement when the first 4 (8) samples of eah lass are used as the prototypes. he horizontal axis of Fig. 4 is the number of nearest neighbors of the PSSR. Fig. 3. Some images of the subset of Coil-00 database able I: COMPUAIONAL COS OF EACH QUERY ON COIL OBJEC DAABASE USING HE FIRS N SCHEME (UNI: SECONDS) Classifier RR (4) RR (8) NN CM LRC CRC SRC PVRC VI-B Objet reognition on oil objet database he Coil-00 data set [3] was widely used as an objet-reognition benhmark [4]-[7]. In this data set, there are 00 objets and eah objet has 7 different views (images) that are taken every 5 around an axis passing through the objet. Eah image is a 8 8 olor one with R, G, B hannels. We use only a limited number of views per objets for experiments. In our experiments, different views per objet (0, 30, 60, 90, 0, 50, 80, 0, 40, 70, 300 and 330 ) were used, shown in Fig. 3. So the subset of Coil-00 data set ontains 00 images, and all images in subset of Coil-00 database were manually ropped into a 3 3 olor one with R, G, B hannels. In the first experiment, we adopt the leave one out sheme on oil objet database. he results are listed in able II. In the seond experiment, we test the reognition rate on oil objet database using first N sheme. he results are desribed in Fig. 4 and able III. In able II, the reognition rate (RR) of PVRC lassifier outperforms the RRs of the SRC, LRC, CRC, NN and CM lassifiers with 7.08%,.50%, 3.5%, 6.75%, and 5.83% improvements, respetively. In able III, the RR of the PVRC lassifier outperforms the RRs of the SRC, LRC, CRC, NN and CM lassifiers, with 7.00%,.00%, 30.00%, 4.88%, and 4.00% improvements, respetively while the first 4 samples are used as prototype. he RR of the PVRC lassifier outperforms the RRs of them with 4.7%,.95%, 9.5%, 6.45%, and 3.5% improvements, respetively while the first 8 samples are used as Fig. 4. he RR of the PSSR lassifier on oil objet database using the first N sheme able II: RECOGNIION RAES (RRS) OF SEVERAL CLASSIFIERS ON COIL OBJEC DAABASE USING HE LEAVE ONE OU SCHEME Classifier NN CM LRC RR 86.5% 77.7% 9.50% Classifier CRC SRC PVRC RR 79.75% 85.9% 93.00% able III: RECOGNIION RAES (RRS) OF SEVERAL CLASSIFIERS ON COIL OBJEC DAABASE USING HE FIRS N SCHEME Classifier RR (4) RR (8) NN 69.87% 85.55% CM 70.75% 78.75% LRC 73.75% 90.05% CRC 65.00% 8.75% SRC 67.75% 87.5% PVRC 74.75% 9.00% VI-C Objet reognition on eth80 objet database In the eth80-ropped-lose8 objet database [8] [9], all images are ropped, so that they ontain only the objet without any border area. In addition, they are resaled to a size

7 of 8 8 pixels. Again, the sale is left the same for all images of the same objet. his dataset is useful when no derivatives are needed. In our experimental, all images are resized to 3 3 gray images. Some images of eth80 objet database are shown in Fig. 5. Fig. 5. Some sampled images of eth80 database In the third experiment, we adopt the leave one out sheme on eth80 objet database. he result is listed in able IV. In the fourth experiment, we test the reognition rate on eth80 objet database using first N sheme. he results are desribed in Fig. 6 and able V. In able IV, the reognition rate (RR) of the PVRC lassifier outperforms the RRs of the SRC, LRC, CRC, NN and CM lassifiers with.0%, 4.75%, 4.45%,.5%, and 9.05% improvements, respetively. In able V, the RR of the PVRC lassifier outperforms the RRs of the SRC, LRC, CRC, NN and CM lassifiers with 7.56%, 4.%, 9.73%, 3.07%, and 5.0% improvements, respetively while the first 4 samples are used as prototype. he RR of the PVRC lassifier outperforms the RRs of them with 6.5%, 4.40%,.4%, 3.5%, and 6.5% improvements, respetively while the first 6 samples are used as prototype. From Fig. 6, we learnt that the RR of PVRC lassifier outperforms the best RR of the PSSR.03% (3.86%) when the first 4 (6) samples of eah lass are used as prototype. he horizontal axis of Fig. 6 is the number of nearest neighbors of the PSSR. able IV: HE RR OF SEVERAL CLASSIFIERS ON EH80 OBJEC DAABASE USING HE LEAVE ONE OU SCHEME Classifier NN CM LRC RR 64.60% 46.80% 7.0% Classifier CRC SRC PVRC RR 5.40% 64.8% 75.85% Fig. 6. he RR of PSSR lassifier on eth80 objet database using the first N sheme VI-D Objet reognition on soil47 objet database he Soil-47 data set [30] was widely used as an objet-reognition benhmark. In the data set, there are 47 objets and eah objet has different views (images) that are taken every 9 around an axis passing through the objet. In the experimental, the subset of soil-47 inludes 966 images of 46 objets. Eah olor image is downsampled to a 4 3 gray image. Fig. 7 shows some sampled images of these objets. Fig. 7 Some sampled images of SOIL-47 data set In the fifth experiment, we adopt the leave one out sheme on eth80 objet database. he result is listed in able VI. In the sixth experiment, we test the reognition rate on eth80 objet database using first N sheme. he results are desribed in Fig. 8 and able VII. able V: RECOGNIION RAES (RRS) OF SEVERAL CLASSIFIERS ON EH80 OBJEC DAABASE USING HE FIRS N SCHEME Classifier RR (4) RR (6) NN 9.46%.86% CM 7.43% 9.% LRC 8.3% 0.96% CRC.80% 4.% SRC 4.97% 9.% PVRC.53% 5.36% Fig. 8. he RR of PSSR lassifier on Soil47 objet database using the first N sheme

8 able VI : HE RR OF SEVERAL CLASSIFIERS ON SOIL47 OBJEC DAABASE USING HE LEAVE ONE OU SCHEME Classifier NN CM LRC RR 64.60% 44.6% 7.67% Classifier CRC SRC PVRC RR 64.8% 7.0% 75.85% able VII : RECOGNIION RAE (RR) OF SEVERAL CLASSIFIERS ON SOIL47 OBJEC DAABASE USING HE FIRS N SCHEME Classifier RR (4) RR (6) NN 4.46% 39.57% CM 39.39% 35.07% LRC 49.74% 50.5% CRC 45.0% 43.04% SRC 48.47% 48.70% PVRC 50.6% 5.3% In able VI, the reognition rate (RR) of the PVRC lassifier outperforms the RRs of the SRC, LRC, CRC, NN and CM lassifiers with 4.84%, 3.8%,.67%,.5%, and 3.3% improvement, respetively. In able VII, the RR of the PVRC lassifier outperforms the RRs of the SRC, LRC, CRC, NN and CM lassifiers with.79%, 0.5%, 5.5%, 7.80%, and 0.87% improvements, respetively while the first 4 samples are used as prototype. he RR of the PVRC lassifier outperforms the RR of them with.6%,.6%, 8.7%,.74%, and 6.4% improvements, respetively while the first 6 samples are used as prototype. From Fig. 8, we learnt that the RR of PVRC lassifier outperforms the best RR of the PSSR with 7.45% (0.7%) improvements when the first 4 (6) samples of eah lass are used as the prototypes. he horizontal axis of Fig. 8 is the number of nearest neighbors of the PSSR. VI-E Fae reognition on G fae database Georgia eh fae database [3] ontains images of 50 people taken in two or three sessions between 06/0/99 and /5/99 at the Center for Signal and Image Proessing at Georgia Institute of ehnology. All people in the database are represented by 5 olor JPEG images with luttered bakground taken at resolution pixels. he average size of the faes in these images is pixels. he pitures show frontal and/or tilted faes with different faial expressions, lighting onditions and sale. Eah image is manually of the G fae database is manually ropped in a gray image. Fig. 9 shows some seleted fae images of G fae database. In the seventh experiment, we adopt the leave one out sheme on G fae database. he results are listed in able VIII. In the eighth experiment, we test the reognition rate on G fae database using first N sheme. he results are desribed in Fig. 0 and able VIIII. Fig. 9. Some seleted fae images of G fae database ABLE VIII : HE RR OF SEVERAL CLASSIFIERS ON G FACE DAABASE USING HE LEAVE ONE OU SCHEME Classifier NN CM LRC RR 8.64% 73.60% 84.67% Classifier CRC SRC PVRC RR 74.7% 83.87% 87.0% able VIIII : HE RECOGNIION RAE (RR) OF SEVERAL CLASSIFIERS ON G FACE DAABASE USING HE FIRS N SCHEME Classifier RR (3) RR (6) NN 49.83% 67.% CM 48.50% 54.% LRC 5.83% 68.44% CRC 46.00% 6.33% SRC 53.7% 68.89% PVRC 53.83% 7.89% In able VIII, the reognition rate (RR) of the PVRC lassifier outperforms the RR of the SRC, LRC, CRC, NN and CM lassifiers with 3.33%,.53%,.93%, 4.56%, and 3.60% improvements, respetively. In able VIIII, ompared with SRC, LRC, CRC, NN and CM lassifiers, the RR of PVRC lassifier outperforms the RRs of them with 0.66%,.8%, 7.83%, 6.36%, and 6.8% improvements, respetively, when the first 3 samples are used as the prototypes. he RR of the PVRC lassifier outperforms the RRs of them with 4.00%, 4.45%,.56%, 5.75%, and 8.67% improvements, respetively, when the first 6 samples are used as the prototypes. From Fig.0, we learn that the RR of the PVRC lassifier outperforms the best RR of the PSSR with 0.83% (6.45%) improvement when the first 3 (6) samples of eah lass are used as prototype. he horizontal axis of Fig. 0 is the number of nearest neighbors of the PSSR. Fig.0. he RR of PSSR lassifier on G fae database using the first N sheme VI-F Fae reognition on UMIS fae database he Sheffield (previously UMIS) Fae Database [3] onsists of 564 images of 0 individuals, whih are mixed with rae, gender, and appearane. Eah individual is shown in a

9 range of poses from profile to frontal views - eah in a separate diretory labelled a, b,... t and images are numbered onseutively as they were taken. he files are all in PGM format, approximately 0 0 pixels with 56-bit grey-sale. In the experiment, a subset of UMIS fae database ontains 480 fae images of 0 individuals. Eah individual has 4 fae images. All images of the subset are manually ropped in gray images. Fig. exhibits some sampled fae images of UMIS fae database. Fig.. Some sampled fae images of UMIS fae database able VV: HE RR OF SEVERAL CLASSIFIERS ON UMIS FACE DAABASE USING HE LEAVE ONE OU SCHEME Classifier NN CM LRC RR 99.5% 95.75% 99.75% Classifier CRC SRC PVRC RR 98.5% 99.75% 99.75% Fig.: he RR of the PSSR lassifier on UMIS fae database using the first N sheme In the ninth experiment, we adopt the leave one out sheme on UMIS fae database. he results are listed in able VV. In the tenth experiment, we test the reognition rate on G fae database using first N sheme. he results are desribed in Fig. and able VVI. able VVI : RECOGNIION RAE (RR) OF SEVERAL CLASSIFIERS ON UMIS FACE DAABASE USING HE FIRS N SCHEME Classifier RR (3) RR (5) NN 90.00% 89.67% CM 89.% 88.00% LRC 88.4% 87.33% CRC 84.4% 83.67% SRC 86.47% 88.33% PVRC 90.00% 90.00% In able VV, the reognition rate (RR) of the PVRC lassifier outperforms the RRs of the SRC, LRC, CRC, NN and CM lassifiers, with 0.00%, 0.00%,.50%, 0.50%, 4.00% improvements, respetively. In able VVI, the RR of PVRC lassifier outperforms the RRs of the SRC, LRC, CRC, NN and CM lassifiers with 3.53%,.76%, 5.59%, 0.00%, and 0.88%, improvements, respetively, when the first 3 samples are used as prototype. he RR of PVRC lassifier outperforms the RRs of the SRC, LRC, CRC, NN and CM lassifiers with.67%,.67%, 5.33%, 0.33%, and.00% improvements, respetively, when the first 5 samples are used as the prototypes. From Fig., we learnt that the RR of the PVRC lassifier outperforms the best RR of the PSSR 3.4% (0.00%) when the first 3 (5) samples of eah lass are used as prototype. he horizontal axis of Fig. is the number of nearest neighbors of the PSSR. VII. CONCLUSION In this paper, a novel lassifier alled polyhedron volume ratio lassifiation (PVRC) is proposed for image reognition. he PVRC lassifier uses the test sample vetor and the original lass prototype samples to alulate the distane between the test sample and the orresponding lass prototype samples. he lass-based distane omputation is oneptually similar to the LRC and CM lassifiers. However, the PVRC omputes the distane between the test sample and the lass subspae by the ratio of two polyhedron volumes instead of alulating the distane between the test sample and the simplified lass features, suh as the lass means or approximated lass subspaes. he proposed PVRC lassifier, whih omputes the perpendiular distane of the test sample to the lass polyhedron, ahieves the better reognition rate than the NN, CM, CRC, SRC, PSSR and LRC lassifiers. he omputation of the PVRC in the testing phase is similar to those of the NN and CM lassifiers. hus, the PRVR takes less omputation than the CRC, LRC and SRC lassifiers. All experimental results onfirm the effetiveness and analyses of the proposed lassifiation algorithm. Appendix A: Matrix Segmentation and Its Determinant Computation A B Let P C D be a square matrix ( m n ) ( m n ) to be segmented into four submatries, A, B, C and D are with sizes of m m, m n, n m and n n, respetively. It is noted that D should be a reversible square matrix. hen, we an prove:

10 A C B D det( P) det det( A BD C) det( D) Proof: Im O Give a square matrix, D C I, where are m m, n and n n identify matries, respetively, and O is zeros matrix. hen, it is easy to verify I, I A B Im O A BD C B C D D C I n O D m n. (A) By using the existing properties of determinant, det(ab) = E F det(a)det(b) and det det( E)det( G) O G, we an ompute the determinant of both sides of (A) to obtain, A B Im O det( )det( ) C D D C I n A BD C B det( ) O D det( D)det( A BD C) (A) In (A), it is also straightforward to prove that Im O det det( )det( ) I m I D C I n. End of Proof. n REFERENCES []. M. Cover and P. E. Hart, Nearest Neighbor Pattern Classifiation, IEEE rans. on Inform. heory, vol. 3, no., pp. 7, Jan [] R. C. Gonzalez, and R. E. Woods, Digital image proessing (Addison Wesley, 997). [3] P. N. Belhumeur, J. P. Hespanha, and D. J. Kriegman, Eigenfaes vs. Fisherfaes: Reognition Using Class Speifi Linear Projetion, IEEE rans. on Pattern Analysis and Mahine Intelligene, vol. 9, no. 7, pp. 7-70, July 997. [4] R. Barsi and D. Jaobs, Lambertian Refletion and Linear Subspaes, IEEE rans. on Pattern Analysis and Mahine Intelligene, vol. 5, no., pp. 8-33, Feb [5] X. Chai, S. Shan, X. Chen, and W. Gao, Loally Linear Regression for Pose-Invariant Fae Reognition, IEEE rans. on Image Proessing, vol. 6, no. 7, pp , July 007. [6] I. Naseem, R. ogneri and M. Bennamoun, Linear Regression for Fae Reognition, IEEE rans.on Pattern Analysis and Mahine Intelligene, vol. 3, no., pp. 06-, Nov. 00. [7] S.-M. Huang and J.-F. Yang, Kernel Linear Regression For Low Resolution Fae Reognition Under Variable Illumination, Pro. IEEE International Conf. on Aoustis, Speeh and Signal Proessing (ICASSP), pp , 0. [8] S.-M. Huang and J.-F. Yang, Improved Prinipal Component Regression for Fae Reognition Under Illumination Variations, IEEE Signal Proessing Letters, vol. 9, no. 4, pp. 79-8, 0. [9] S.-M. Huang and J.-F. Yang, Linear Disriminant Regression Classifiation for Fae Reognition, IEEE Signal Proessing Letters, vol. 0, no., pp. 9-94, 03. [0] S.-M. Huang and J.-F. Yang, Unitary Regression Classifiation with otal Minimum Projetion Error for Fae Reognition, IEEE Signal Proessing Letters, vol. 0, no. 5, pp , 03. [] J. Wright, Y. Ma, J. Mairal, G. Sapiro,. S. Huang, and S. Yan, Sparse representation for omputer vision and pattern reognition, Pro. of IEEE, vol. 98, no. 6, pp , Jun. 00. [] J. Wright, A. Y. Yang, A. Ganesh, S. S. Sastry, and Y. Ma, Robust fae reognition via sparse representation, IEEE rans. on Pattern Anal. Mah. Intell., vol. 3, no., pp. 0 7, Feb [3] Y. Xu, D. Zhang, J. Yang, J.-Y. Yang, A wo-phase est Sample Sparse Representation Method for Use with Fae Reognition, IEEE rans. on Ciruits and Systems for Video ehnology, vol., no. 9, pp.55-6, 0 [4] Y. Xu, Q. Zhu, Z. Fan, D. Zhang, J. Mi, Z.Lai, Using the Idea of the Sparse Representation to Perform Coarse to Fine Fae Reognition, Information Sienes, vol.38, pp.38-48, 03 [5] Y. Xu, Q. Zhu, Z. Fan, M. Qiu, Y. Chen, H. Liu, Coarse to Fine K Nearest neighbor lassifier, Pattern Reognition Letters, vol.34, pp , 03 [6] Y. Xu, X. Zhu, Z. Li, G. Liu, Y. Lu, H. Liu, Using the original and symmetrial fae training samples to perform representation based two-step fae reognition, Pattern Reognition, vol.46, pp.5-58, 03 [7] L. Zhang, M. Yang, and X. Feng, Sparse representation or ollaborative representation: Whih helps fae reognition?, Pro. of IEEE Int l Conf. Computer Vision, pp , Nov. 0. [8] J. Yang, D. Chu, L. Zhang, Y. Xu and J. Yang, Sparse Representation Classifier Steered Disriminative Projetion With Appliations to Fae Reognition, IEEE rans. on Neural Networks and Learning Systems, vol. 4, no.7, pp ,July 03. [9] Z. Lai, Y. Xu, J. Yang, J. ang, and D. Zhang, Sparse ensor Disriminant Analysis, IEEE rans. on Image Proessing, vol., no.0, pp , 03. [0] Z. Lai, W. K. Wong, Z. Jin, J. Yang, and Y. Xu, Sparse Approximation to the Eigensubspae for Disrimination, IEEE ransations on Neural Networks and Learning Systems, vol. 3, no., pp , 0. [] M. J. Sippl andh. A. Sheraga, Cayley-Menger Coordinates, Pro. Natl. Aad. Si., Applied Physial and Mathematial Sienes, vol. 83, pp.83-87, April 986. [] E.W. Weisstein, Heron s Formula. MathWorld A Wolfram Web Resoure,00. < [3] S. A. Nene, S. K. Nayar, H. Murase, Columbia objet image library (COIL-00), ehnial Report CUCS , Columbia University, 996. [4] H. Murase, S.K. Nayar, Visual Learning and Reognition of 3D Objets from Appearane, Int. J. Comput. Vision, vol. 4, no., 995, pp [5] D. Roobaert, M.M. van Hulle, View-based 3D Objet Reognition with Support Vetor Mahines, Pro. of 999 IEEE International Workshop on Neural Networks for Signal Proessing, Madison, WI, USA, 999, pp [6] D. Roth, M.-H. Yang, N. Ahuja, Learning to Reognize hree-dimensional Objets, Neural Comput. vol. 4, no. 5, pp , May 00. [7] X.-Y. Jing, S. Li, D. Zhang, J. Yang, and J.-Y. Yang, Supervised and Unsupervised Parallel Subspae Learning for Large-sale Image Reognition, IEEE ransations on Ciruits and Systems for Video ehnology, vol., no.0, pp [8] [9] B. Leibe and B. Shiele, Analyzing Appearane and Contour Based Methods for Objet Categorization. Pro. of International Conferene on Computer Vision and Pattern Reognition (CVPR'03), Madison, Wisonsin, June 003 [30] D. Koubaroulis, J. Matas, J. Kittler, Evaluating olour-based objet reognition algorithms using the SOIL-47 database, Pro. of the Fifth Asian Conferene on Computer Vision, Melbourne, Australia, 00, pp [3] [3] D. B. Graham, N. M. Allinson, Charaterising Virtual Eigensignatures for General Purpose Fae Reognition, Fae Reognition: from heory to Appliations, vol.63, pp , 998.

The Minimum Redundancy Maximum Relevance Approach to Building Sparse Support Vector Machines

The Minimum Redundancy Maximum Relevance Approach to Building Sparse Support Vector Machines The Minimum Redundany Maximum Relevane Approah to Building Sparse Support Vetor Mahines Xiaoxing Yang, Ke Tang, and Xin Yao, Nature Inspired Computation and Appliations Laboratory (NICAL), Shool of Computer

More information

the data. Structured Principal Component Analysis (SPCA)

the data. Structured Principal Component Analysis (SPCA) Strutured Prinipal Component Analysis Kristin M. Branson and Sameer Agarwal Department of Computer Siene and Engineering University of California, San Diego La Jolla, CA 9193-114 Abstrat Many tasks involving

More information

KERNEL SPARSE REPRESENTATION WITH LOCAL PATTERNS FOR FACE RECOGNITION

KERNEL SPARSE REPRESENTATION WITH LOCAL PATTERNS FOR FACE RECOGNITION KERNEL SPARSE REPRESENTATION WITH LOCAL PATTERNS FOR FACE RECOGNITION Cuiui Kang 1, Shengai Liao, Shiming Xiang 1, Chunhong Pan 1 1 National Laboratory of Pattern Reognition, Institute of Automation, Chinese

More information

Abstract. Key Words: Image Filters, Fuzzy Filters, Order Statistics Filters, Rank Ordered Mean Filters, Channel Noise. 1.

Abstract. Key Words: Image Filters, Fuzzy Filters, Order Statistics Filters, Rank Ordered Mean Filters, Channel Noise. 1. Fuzzy Weighted Rank Ordered Mean (FWROM) Filters for Mixed Noise Suppression from Images S. Meher, G. Panda, B. Majhi 3, M.R. Meher 4,,4 Department of Eletronis and I.E., National Institute of Tehnology,

More information

A DYNAMIC ACCESS CONTROL WITH BINARY KEY-PAIR

A DYNAMIC ACCESS CONTROL WITH BINARY KEY-PAIR Malaysian Journal of Computer Siene, Vol 10 No 1, June 1997, pp 36-41 A DYNAMIC ACCESS CONTROL WITH BINARY KEY-PAIR Md Rafiqul Islam, Harihodin Selamat and Mohd Noor Md Sap Faulty of Computer Siene and

More information

Boosted Random Forest

Boosted Random Forest Boosted Random Forest Yohei Mishina, Masamitsu suhiya and Hironobu Fujiyoshi Department of Computer Siene, Chubu University, 1200 Matsumoto-ho, Kasugai, Aihi, Japan {mishi, mtdoll}@vision.s.hubu.a.jp,

More information

A Coarse-to-Fine Classification Scheme for Facial Expression Recognition

A Coarse-to-Fine Classification Scheme for Facial Expression Recognition A Coarse-to-Fine Classifiation Sheme for Faial Expression Reognition Xiaoyi Feng 1,, Abdenour Hadid 1 and Matti Pietikäinen 1 1 Mahine Vision Group Infoteh Oulu and Dept. of Eletrial and Information Engineering

More information

A Novel Validity Index for Determination of the Optimal Number of Clusters

A Novel Validity Index for Determination of the Optimal Number of Clusters IEICE TRANS. INF. & SYST., VOL.E84 D, NO.2 FEBRUARY 2001 281 LETTER A Novel Validity Index for Determination of the Optimal Number of Clusters Do-Jong KIM, Yong-Woon PARK, and Dong-Jo PARK, Nonmembers

More information

Performance of Histogram-Based Skin Colour Segmentation for Arms Detection in Human Motion Analysis Application

Performance of Histogram-Based Skin Colour Segmentation for Arms Detection in Human Motion Analysis Application World Aademy of Siene, Engineering and Tehnology 8 009 Performane of Histogram-Based Skin Colour Segmentation for Arms Detetion in Human Motion Analysis Appliation Rosalyn R. Porle, Ali Chekima, Farrah

More information

arxiv: v1 [cs.db] 13 Sep 2017

arxiv: v1 [cs.db] 13 Sep 2017 An effiient lustering algorithm from the measure of loal Gaussian distribution Yuan-Yen Tai (Dated: May 27, 2018) In this paper, I will introdue a fast and novel lustering algorithm based on Gaussian distribution

More information

System-Level Parallelism and Throughput Optimization in Designing Reconfigurable Computing Applications

System-Level Parallelism and Throughput Optimization in Designing Reconfigurable Computing Applications System-Level Parallelism and hroughput Optimization in Designing Reonfigurable Computing Appliations Esam El-Araby 1, Mohamed aher 1, Kris Gaj 2, arek El-Ghazawi 1, David Caliga 3, and Nikitas Alexandridis

More information

FUZZY WATERSHED FOR IMAGE SEGMENTATION

FUZZY WATERSHED FOR IMAGE SEGMENTATION FUZZY WATERSHED FOR IMAGE SEGMENTATION Ramón Moreno, Manuel Graña Computational Intelligene Group, Universidad del País Vaso, Spain http://www.ehu.es/winto; {ramon.moreno,manuel.grana}@ehu.es Abstrat The

More information

A {k, n}-secret Sharing Scheme for Color Images

A {k, n}-secret Sharing Scheme for Color Images A {k, n}-seret Sharing Sheme for Color Images Rastislav Luka, Konstantinos N. Plataniotis, and Anastasios N. Venetsanopoulos The Edward S. Rogers Sr. Dept. of Eletrial and Computer Engineering, University

More information

Pipelined Multipliers for Reconfigurable Hardware

Pipelined Multipliers for Reconfigurable Hardware Pipelined Multipliers for Reonfigurable Hardware Mithell J. Myjak and José G. Delgado-Frias Shool of Eletrial Engineering and Computer Siene, Washington State University Pullman, WA 99164-2752 USA {mmyjak,

More information

Capturing Large Intra-class Variations of Biometric Data by Template Co-updating

Capturing Large Intra-class Variations of Biometric Data by Template Co-updating Capturing Large Intra-lass Variations of Biometri Data by Template Co-updating Ajita Rattani University of Cagliari Piazza d'armi, Cagliari, Italy ajita.rattani@diee.unia.it Gian Lua Marialis University

More information

Comparing Images Under Variable Illumination

Comparing Images Under Variable Illumination ( This paper appeared in CVPR 8. IEEE ) Comparing Images Under Variable Illumination David W. Jaobs Peter N. Belhumeur Ronen Basri NEC Researh Institute Center for Computational Vision and Control The

More information

Virtual Training Samples and CRC based Test Sample Reconstruction and Face Recognition Experiments Wei HUANG and Li-ming MIAO

Virtual Training Samples and CRC based Test Sample Reconstruction and Face Recognition Experiments Wei HUANG and Li-ming MIAO 7 nd International Conference on Computational Modeling, Simulation and Applied Mathematics (CMSAM 7) ISBN: 978--6595-499-8 Virtual raining Samples and CRC based est Sample Reconstruction and Face Recognition

More information

3-D IMAGE MODELS AND COMPRESSION - SYNTHETIC HYBRID OR NATURAL FIT?

3-D IMAGE MODELS AND COMPRESSION - SYNTHETIC HYBRID OR NATURAL FIT? 3-D IMAGE MODELS AND COMPRESSION - SYNTHETIC HYBRID OR NATURAL FIT? Bernd Girod, Peter Eisert, Marus Magnor, Ekehard Steinbah, Thomas Wiegand Te {girod eommuniations Laboratory, University of Erlangen-Nuremberg

More information

On - Line Path Delay Fault Testing of Omega MINs M. Bellos 1, E. Kalligeros 1, D. Nikolos 1,2 & H. T. Vergos 1,2

On - Line Path Delay Fault Testing of Omega MINs M. Bellos 1, E. Kalligeros 1, D. Nikolos 1,2 & H. T. Vergos 1,2 On - Line Path Delay Fault Testing of Omega MINs M. Bellos, E. Kalligeros, D. Nikolos,2 & H. T. Vergos,2 Dept. of Computer Engineering and Informatis 2 Computer Tehnology Institute University of Patras,

More information

Cluster-Based Cumulative Ensembles

Cluster-Based Cumulative Ensembles Cluster-Based Cumulative Ensembles Hanan G. Ayad and Mohamed S. Kamel Pattern Analysis and Mahine Intelligene Lab, Eletrial and Computer Engineering, University of Waterloo, Waterloo, Ontario N2L 3G1,

More information

Chromaticity-matched Superimposition of Foreground Objects in Different Environments

Chromaticity-matched Superimposition of Foreground Objects in Different Environments FCV216, the 22nd Korea-Japan Joint Workshop on Frontiers of Computer Vision Chromatiity-mathed Superimposition of Foreground Objets in Different Environments Yohei Ogura Graduate Shool of Siene and Tehnology

More information

An Optimized Approach on Applying Genetic Algorithm to Adaptive Cluster Validity Index

An Optimized Approach on Applying Genetic Algorithm to Adaptive Cluster Validity Index IJCSES International Journal of Computer Sienes and Engineering Systems, ol., No.4, Otober 2007 CSES International 2007 ISSN 0973-4406 253 An Optimized Approah on Applying Geneti Algorithm to Adaptive

More information

Multi-Piece Mold Design Based on Linear Mixed-Integer Program Toward Guaranteed Optimality

Multi-Piece Mold Design Based on Linear Mixed-Integer Program Toward Guaranteed Optimality INTERNATIONAL CONFERENCE ON MANUFACTURING AUTOMATION (ICMA200) Multi-Piee Mold Design Based on Linear Mixed-Integer Program Toward Guaranteed Optimality Stephen Stoyan, Yong Chen* Epstein Department of

More information

-z c = c T - c T B B-1 A 1 - c T B B-1 b. x B B -1 A 0 B -1 b. (a) (b) Figure 1. Simplex Tableau in Matrix Form

-z c = c T - c T B B-1 A 1 - c T B B-1 b. x B B -1 A 0 B -1 b. (a) (b) Figure 1. Simplex Tableau in Matrix Form 3. he Revised Simple Method he LP min, s.t. A = b ( ),, an be represented by Figure (a) below. At any Simple step, with known and -, the Simple tableau an be represented by Figure (b) below. he minimum

More information

Improved Vehicle Classification in Long Traffic Video by Cooperating Tracker and Classifier Modules

Improved Vehicle Classification in Long Traffic Video by Cooperating Tracker and Classifier Modules Improved Vehile Classifiation in Long Traffi Video by Cooperating Traker and Classifier Modules Brendan Morris and Mohan Trivedi University of California, San Diego San Diego, CA 92093 {b1morris, trivedi}@usd.edu

More information

Machine Vision. Laboratory Exercise Name: Student ID: S

Machine Vision. Laboratory Exercise Name: Student ID: S Mahine Vision 521466S Laoratory Eerise 2011 Name: Student D: General nformation To pass these laoratory works, you should answer all questions (Q.y) with an understandale handwriting either in English

More information

Time delay estimation of reverberant meeting speech: on the use of multichannel linear prediction

Time delay estimation of reverberant meeting speech: on the use of multichannel linear prediction University of Wollongong Researh Online Faulty of Informatis - apers (Arhive) Faulty of Engineering and Information Sienes 7 Time delay estimation of reverberant meeting speeh: on the use of multihannel

More information

Dr.Hazeem Al-Khafaji Dept. of Computer Science, Thi-Qar University, College of Science, Iraq

Dr.Hazeem Al-Khafaji Dept. of Computer Science, Thi-Qar University, College of Science, Iraq Volume 4 Issue 6 June 014 ISSN: 77 18X International Journal of Advaned Researh in Computer Siene and Software Engineering Researh Paper Available online at: www.ijarsse.om Medial Image Compression using

More information

Outline: Software Design

Outline: Software Design Outline: Software Design. Goals History of software design ideas Design priniples Design methods Life belt or leg iron? (Budgen) Copyright Nany Leveson, Sept. 1999 A Little History... At first, struggling

More information

Figure 1. LBP in the field of texture analysis operators.

Figure 1. LBP in the field of texture analysis operators. L MEHODOLOGY he loal inary pattern (L) texture analysis operator is defined as a gray-sale invariant texture measure, derived from a general definition of texture in a loal neighorhood. he urrent form

More information

A scheme for racquet sports video analysis with the combination of audio-visual information

A scheme for racquet sports video analysis with the combination of audio-visual information A sheme for raquet sports video analysis with the ombination of audio-visual information Liyuan Xing a*, Qixiang Ye b, Weigang Zhang, Qingming Huang a and Hua Yu a a Graduate Shool of the Chinese Aadamy

More information

Evolutionary Feature Synthesis for Image Databases

Evolutionary Feature Synthesis for Image Databases Evolutionary Feature Synthesis for Image Databases Anlei Dong, Bir Bhanu, Yingqiang Lin Center for Researh in Intelligent Systems University of California, Riverside, California 92521, USA {adong, bhanu,

More information

Unsupervised Stereoscopic Video Object Segmentation Based on Active Contours and Retrainable Neural Networks

Unsupervised Stereoscopic Video Object Segmentation Based on Active Contours and Retrainable Neural Networks Unsupervised Stereosopi Video Objet Segmentation Based on Ative Contours and Retrainable Neural Networks KLIMIS NTALIANIS, ANASTASIOS DOULAMIS, and NIKOLAOS DOULAMIS National Tehnial University of Athens

More information

Spatial-Aware Collaborative Representation for Hyperspectral Remote Sensing Image Classification

Spatial-Aware Collaborative Representation for Hyperspectral Remote Sensing Image Classification Spatial-Aware Collaborative Representation for Hyperspetral Remote Sensing Image ifiation Junjun Jiang, Member, IEEE, Chen Chen, Member, IEEE, Yi Yu, Xinwei Jiang, and Jiayi Ma Member, IEEE Representation-residual

More information

Discrete sequential models and CRFs. 1 Case Study: Supervised Part-of-Speech Tagging

Discrete sequential models and CRFs. 1 Case Study: Supervised Part-of-Speech Tagging 0-708: Probabilisti Graphial Models 0-708, Spring 204 Disrete sequential models and CRFs Leturer: Eri P. Xing Sribes: Pankesh Bamotra, Xuanhong Li Case Study: Supervised Part-of-Speeh Tagging The supervised

More information

Type of document: Usebility Checklist

Type of document: Usebility Checklist Projet: JEGraph Type of doument: Usebility Cheklist Author: Max Bryan Version: 1.30 2011 Envidate GmbH Type of Doumet Developer guidelines User guidelines Dutybook Speifiation Programming and testing Test

More information

Detection and Recognition of Non-Occluded Objects using Signature Map

Detection and Recognition of Non-Occluded Objects using Signature Map 6th WSEAS International Conferene on CIRCUITS, SYSTEMS, ELECTRONICS,CONTROL & SIGNAL PROCESSING, Cairo, Egypt, De 9-31, 007 65 Detetion and Reognition of Non-Oluded Objets using Signature Map Sangbum Park,

More information

Using Augmented Measurements to Improve the Convergence of ICP

Using Augmented Measurements to Improve the Convergence of ICP Using Augmented Measurements to Improve the onvergene of IP Jaopo Serafin, Giorgio Grisetti Dept. of omputer, ontrol and Management Engineering, Sapienza University of Rome, Via Ariosto 25, I-0085, Rome,

More information

Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors Eurographis Symposium on Geometry Proessing (003) L. Kobbelt, P. Shröder, H. Hoppe (Editors) Rotation Invariant Spherial Harmoni Representation of 3D Shape Desriptors Mihael Kazhdan, Thomas Funkhouser,

More information

Plot-to-track correlation in A-SMGCS using the target images from a Surface Movement Radar

Plot-to-track correlation in A-SMGCS using the target images from a Surface Movement Radar Plot-to-trak orrelation in A-SMGCS using the target images from a Surfae Movement Radar G. Golino Radar & ehnology Division AMS, Italy ggolino@amsjv.it Abstrat he main topi of this paper is the formulation

More information

Video Data and Sonar Data: Real World Data Fusion Example

Video Data and Sonar Data: Real World Data Fusion Example 14th International Conferene on Information Fusion Chiago, Illinois, USA, July 5-8, 2011 Video Data and Sonar Data: Real World Data Fusion Example David W. Krout Applied Physis Lab dkrout@apl.washington.edu

More information

Learning Convention Propagation in BeerAdvocate Reviews from a etwork Perspective. Abstract

Learning Convention Propagation in BeerAdvocate Reviews from a etwork Perspective. Abstract CS 9 Projet Final Report: Learning Convention Propagation in BeerAdvoate Reviews from a etwork Perspetive Abstrat We look at the way onventions propagate between reviews on the BeerAdvoate dataset, and

More information

Exploiting Enriched Contextual Information for Mobile App Classification

Exploiting Enriched Contextual Information for Mobile App Classification Exploiting Enrihed Contextual Information for Mobile App Classifiation Hengshu Zhu 1 Huanhuan Cao 2 Enhong Chen 1 Hui Xiong 3 Jilei Tian 2 1 University of Siene and Tehnology of China 2 Nokia Researh Center

More information

1. Inversions. A geometric construction relating points O, A and B looks as follows.

1. Inversions. A geometric construction relating points O, A and B looks as follows. 1. Inversions. 1.1. Definitions of inversion. Inversion is a kind of symmetry about a irle. It is defined as follows. he inversion of degree R 2 entered at a point maps a point to the point on the ray

More information

NONLINEAR BACK PROJECTION FOR TOMOGRAPHIC IMAGE RECONSTRUCTION. Ken Sauer and Charles A. Bouman

NONLINEAR BACK PROJECTION FOR TOMOGRAPHIC IMAGE RECONSTRUCTION. Ken Sauer and Charles A. Bouman NONLINEAR BACK PROJECTION FOR TOMOGRAPHIC IMAGE RECONSTRUCTION Ken Sauer and Charles A. Bouman Department of Eletrial Engineering, University of Notre Dame Notre Dame, IN 46556, (219) 631-6999 Shool of

More information

Accepted Manuscript. Domain Class Consistency based Transfer Learning for Image Classification Across Domains. Lei Zhang, Jian Yang, David Zhang

Accepted Manuscript. Domain Class Consistency based Transfer Learning for Image Classification Across Domains. Lei Zhang, Jian Yang, David Zhang Aepted Manusript Domain lass onsisteny based ransfer Learning for Image lassifiation Aross Domains Lei Zhang Jian Yang David Zhang PII: 000-055(6)335-9 DOI: 0.06/j.ins.07.08.034 Referene: IN 3038 o appear

More information

特集 Road Border Recognition Using FIR Images and LIDAR Signal Processing

特集 Road Border Recognition Using FIR Images and LIDAR Signal Processing デンソーテクニカルレビュー Vol. 15 2010 特集 Road Border Reognition Using FIR Images and LIDAR Signal Proessing 高木聖和 バーゼル ファルディ Kiyokazu TAKAGI Basel Fardi ヘンドリック ヴァイゲル Hendrik Weigel ゲルド ヴァニーリック Gerd Wanielik This paper

More information

Graph-Based vs Depth-Based Data Representation for Multiview Images

Graph-Based vs Depth-Based Data Representation for Multiview Images Graph-Based vs Depth-Based Data Representation for Multiview Images Thomas Maugey, Antonio Ortega, Pasal Frossard Signal Proessing Laboratory (LTS), Eole Polytehnique Fédérale de Lausanne (EPFL) Email:

More information

Restricted Nearest Feature Line with Ellipse for Face Recognition

Restricted Nearest Feature Line with Ellipse for Face Recognition Journal of Information Hiding and Multimedia Signal Processing c 2012 ISSN 2073-4212 Ubiquitous International Volume 3, Number 3, July 2012 Restricted Nearest Feature Line with Ellipse for Face Recognition

More information

A Partial Sorting Algorithm in Multi-Hop Wireless Sensor Networks

A Partial Sorting Algorithm in Multi-Hop Wireless Sensor Networks A Partial Sorting Algorithm in Multi-Hop Wireless Sensor Networks Abouberine Ould Cheikhna Department of Computer Siene University of Piardie Jules Verne 80039 Amiens Frane Ould.heikhna.abouberine @u-piardie.fr

More information

2017 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media,

2017 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, 2017 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any urrent or future media, inluding reprinting/republishing this material for advertising

More information

Gray Codes for Reflectable Languages

Gray Codes for Reflectable Languages Gray Codes for Refletable Languages Yue Li Joe Sawada Marh 8, 2008 Abstrat We lassify a type of language alled a refletable language. We then develop a generi algorithm that an be used to list all strings

More information

Face and Facial Feature Tracking for Natural Human-Computer Interface

Face and Facial Feature Tracking for Natural Human-Computer Interface Fae and Faial Feature Traking for Natural Human-Computer Interfae Vladimir Vezhnevets Graphis & Media Laboratory, Dept. of Applied Mathematis and Computer Siene of Mosow State University Mosow, Russia

More information

Extracting Partition Statistics from Semistructured Data

Extracting Partition Statistics from Semistructured Data Extrating Partition Statistis from Semistrutured Data John N. Wilson Rihard Gourlay Robert Japp Mathias Neumüller Department of Computer and Information Sienes University of Strathlyde, Glasgow, UK {jnw,rsg,rpj,mathias}@is.strath.a.uk

More information

Micro-Doppler Based Human-Robot Classification Using Ensemble and Deep Learning Approaches

Micro-Doppler Based Human-Robot Classification Using Ensemble and Deep Learning Approaches Miro-Doppler Based Human-Robot Classifiation Using Ensemble and Deep Learning Approahes Sherif Abdulatif, Qian Wei, Fady Aziz, Bernhard Kleiner, Urs Shneider Department of Biomehatroni Systems, Fraunhofer

More information

Gradient based progressive probabilistic Hough transform

Gradient based progressive probabilistic Hough transform Gradient based progressive probabilisti Hough transform C.Galambos, J.Kittler and J.Matas Abstrat: The authors look at the benefits of exploiting gradient information to enhane the progressive probabilisti

More information

Multiple-Criteria Decision Analysis: A Novel Rank Aggregation Method

Multiple-Criteria Decision Analysis: A Novel Rank Aggregation Method 3537 Multiple-Criteria Deision Analysis: A Novel Rank Aggregation Method Derya Yiltas-Kaplan Department of Computer Engineering, Istanbul University, 34320, Avilar, Istanbul, Turkey Email: dyiltas@ istanbul.edu.tr

More information

Supplementary Material: Geometric Calibration of Micro-Lens-Based Light-Field Cameras using Line Features

Supplementary Material: Geometric Calibration of Micro-Lens-Based Light-Field Cameras using Line Features Supplementary Material: Geometri Calibration of Miro-Lens-Based Light-Field Cameras using Line Features Yunsu Bok, Hae-Gon Jeon and In So Kweon KAIST, Korea As the supplementary material, we provide detailed

More information

Naïve Bayesian Rough Sets Under Fuzziness

Naïve Bayesian Rough Sets Under Fuzziness IJMSA: Vol. 6, No. 1-2, January-June 2012, pp. 19 25 Serials Publiations ISSN: 0973-6786 Naïve ayesian Rough Sets Under Fuzziness G. GANSAN 1,. KRISHNAVNI 2 T. HYMAVATHI 3 1,2,3 Department of Mathematis,

More information

Hyperspectral Images Classification Using Energy Profiles of Spatial and Spectral Features

Hyperspectral Images Classification Using Energy Profiles of Spatial and Spectral Features journal homepage: www.elsevier.om Hyperspetral Images Classifiation Using Energy Profiles of Spatial and Spetral Features Hamid Reza Shahdoosti a a Hamedan University of ehnology, Department of Eletrial

More information

A Novel Bit Level Time Series Representation with Implication of Similarity Search and Clustering

A Novel Bit Level Time Series Representation with Implication of Similarity Search and Clustering A Novel Bit Level Time Series Representation with Impliation of Similarity Searh and lustering hotirat Ratanamahatana, Eamonn Keogh, Anthony J. Bagnall 2, and Stefano Lonardi Dept. of omputer Siene & Engineering,

More information

13.1 Numerical Evaluation of Integrals Over One Dimension

13.1 Numerical Evaluation of Integrals Over One Dimension 13.1 Numerial Evaluation of Integrals Over One Dimension A. Purpose This olletion of subprograms estimates the value of the integral b a f(x) dx where the integrand f(x) and the limits a and b are supplied

More information

Cross-layer Resource Allocation on Broadband Power Line Based on Novel QoS-priority Scheduling Function in MAC Layer

Cross-layer Resource Allocation on Broadband Power Line Based on Novel QoS-priority Scheduling Function in MAC Layer Communiations and Networ, 2013, 5, 69-73 http://dx.doi.org/10.4236/n.2013.53b2014 Published Online September 2013 (http://www.sirp.org/journal/n) Cross-layer Resoure Alloation on Broadband Power Line Based

More information

The Mathematics of Simple Ultrasonic 2-Dimensional Sensing

The Mathematics of Simple Ultrasonic 2-Dimensional Sensing The Mathematis of Simple Ultrasoni -Dimensional Sensing President, Bitstream Tehnology The Mathematis of Simple Ultrasoni -Dimensional Sensing Introdution Our ompany, Bitstream Tehnology, has been developing

More information

Boundary Correct Real-Time Soft Shadows

Boundary Correct Real-Time Soft Shadows Boundary Corret Real-Time Soft Shadows Bjarke Jakobsen Niels J. Christensen Bent D. Larsen Kim S. Petersen Informatis and Mathematial Modelling Tehnial University of Denmark {bj, nj, bdl}@imm.dtu.dk, kim@deadline.dk

More information

The Implementation of RRTs for a Remote-Controlled Mobile Robot

The Implementation of RRTs for a Remote-Controlled Mobile Robot ICCAS5 June -5, KINEX, Gyeonggi-Do, Korea he Implementation of RRs for a Remote-Controlled Mobile Robot Chi-Won Roh*, Woo-Sub Lee **, Sung-Chul Kang *** and Kwang-Won Lee **** * Intelligent Robotis Researh

More information

Algorithms, Mechanisms and Procedures for the Computer-aided Project Generation System

Algorithms, Mechanisms and Procedures for the Computer-aided Project Generation System Algorithms, Mehanisms and Proedures for the Computer-aided Projet Generation System Anton O. Butko 1*, Aleksandr P. Briukhovetskii 2, Dmitry E. Grigoriev 2# and Konstantin S. Kalashnikov 3 1 Department

More information

Towards Optimal Naive Bayes Nearest Neighbor

Towards Optimal Naive Bayes Nearest Neighbor Towards Optimal Naive Bayes Nearest Neighbor Régis Behmo 1, Paul Marombes 1,2, Arnak Dalalyan 2,andVéronique Prinet 1 1 NLPR / LIAMA, Institute of Automation, Chinese Aademy of Sienes 2 IMAGINE, LIGM,

More information

Semi-Supervised Affinity Propagation with Instance-Level Constraints

Semi-Supervised Affinity Propagation with Instance-Level Constraints Semi-Supervised Affinity Propagation with Instane-Level Constraints Inmar E. Givoni, Brendan J. Frey Probabilisti and Statistial Inferene Group University of Toronto 10 King s College Road, Toronto, Ontario,

More information

Self-Adaptive Parent to Mean-Centric Recombination for Real-Parameter Optimization

Self-Adaptive Parent to Mean-Centric Recombination for Real-Parameter Optimization Self-Adaptive Parent to Mean-Centri Reombination for Real-Parameter Optimization Kalyanmoy Deb and Himanshu Jain Department of Mehanial Engineering Indian Institute of Tehnology Kanpur Kanpur, PIN 86 {deb,hjain}@iitk.a.in

More information

Contour Box: Rejecting Object Proposals Without Explicit Closed Contours

Contour Box: Rejecting Object Proposals Without Explicit Closed Contours Contour Box: Rejeting Objet Proposals Without Expliit Closed Contours Cewu Lu, Shu Liu Jiaya Jia Chi-Keung Tang The Hong Kong University of Siene and Tehnology Stanford University The Chinese University

More information

timestamp, if silhouette(x, y) 0 0 if silhouette(x, y) = 0, mhi(x, y) = and mhi(x, y) < timestamp - duration mhi(x, y), else

timestamp, if silhouette(x, y) 0 0 if silhouette(x, y) = 0, mhi(x, y) = and mhi(x, y) < timestamp - duration mhi(x, y), else 3rd International Conferene on Multimedia Tehnolog(ICMT 013) An Effiient Moving Target Traking Strateg Based on OpenCV and CAMShift Theor Dongu Li 1 Abstrat Image movement involved bakground movement and

More information

Approximate logic synthesis for error tolerant applications

Approximate logic synthesis for error tolerant applications Approximate logi synthesis for error tolerant appliations Doohul Shin and Sandeep K. Gupta Eletrial Engineering Department, University of Southern California, Los Angeles, CA 989 {doohuls, sandeep}@us.edu

More information

One Against One or One Against All : Which One is Better for Handwriting Recognition with SVMs?

One Against One or One Against All : Which One is Better for Handwriting Recognition with SVMs? One Against One or One Against All : Whih One is Better for Handwriting Reognition with SVMs? Jonathan Milgram, Mohamed Cheriet, Robert Sabourin To ite this version: Jonathan Milgram, Mohamed Cheriet,

More information

Developing Dually Optimal LCA Features in Sensory and Action Spaces for Classification

Developing Dually Optimal LCA Features in Sensory and Action Spaces for Classification Developing Dually Optimal LCA Features in Sensory and Ation Spaes for Classifiation Nikita Wagle and Juyang Weng Department of Computer Siene & Engineering Mihigan State University East Lansing, Mihigan

More information

Shape Outlier Detection Using Pose Preserving Dynamic Shape Models

Shape Outlier Detection Using Pose Preserving Dynamic Shape Models Shape Outlier Detetion Using Pose Preserving Dynami Shape Models Chan-Su Lee Ahmed Elgammal Department of Computer Siene, Rutgers University, Pisataway, NJ 8854 USA CHANSU@CS.RUTGERS.EDU ELGAMMAL@CS.RUTGERS.EDU

More information

Acoustic Links. Maximizing Channel Utilization for Underwater

Acoustic Links. Maximizing Channel Utilization for Underwater Maximizing Channel Utilization for Underwater Aousti Links Albert F Hairris III Davide G. B. Meneghetti Adihele Zorzi Department of Information Engineering University of Padova, Italy Email: {harris,davide.meneghetti,zorzi}@dei.unipd.it

More information

Learning Discriminative and Shareable Features. Scene Classificsion

Learning Discriminative and Shareable Features. Scene Classificsion Learning Disriminative and Shareable Features for Sene Classifiation Zhen Zuo, Gang Wang, Bing Shuai, Lifan Zhao, Qingxiong Yang, and Xudong Jiang Nanyang Tehnologial University, Singapore, Advaned Digital

More information

Particle Swarm Optimization for the Design of High Diffraction Efficient Holographic Grating

Particle Swarm Optimization for the Design of High Diffraction Efficient Holographic Grating Original Artile Partile Swarm Optimization for the Design of High Diffration Effiient Holographi Grating A.K. Tripathy 1, S.K. Das, M. Sundaray 3 and S.K. Tripathy* 4 1, Department of Computer Siene, Berhampur

More information

Directed Rectangle-Visibility Graphs have. Abstract. Visibility representations of graphs map vertices to sets in Euclidean space and

Directed Rectangle-Visibility Graphs have. Abstract. Visibility representations of graphs map vertices to sets in Euclidean space and Direted Retangle-Visibility Graphs have Unbounded Dimension Kathleen Romanik DIMACS Center for Disrete Mathematis and Theoretial Computer Siene Rutgers, The State University of New Jersey P.O. Box 1179,

More information

Exploring the Commonality in Feature Modeling Notations

Exploring the Commonality in Feature Modeling Notations Exploring the Commonality in Feature Modeling Notations Miloslav ŠÍPKA Slovak University of Tehnology Faulty of Informatis and Information Tehnologies Ilkovičova 3, 842 16 Bratislava, Slovakia miloslav.sipka@gmail.om

More information

Folding. Hardware Mapped vs. Time multiplexed. Folding by N (N=folding factor) Node A. Unfolding by J A 1 A J-1. Time multiplexed/microcoded

Folding. Hardware Mapped vs. Time multiplexed. Folding by N (N=folding factor) Node A. Unfolding by J A 1 A J-1. Time multiplexed/microcoded Folding is verse of Unfolding Node A A Folding by N (N=folding fator) Folding A Unfolding by J A A J- Hardware Mapped vs. Time multiplexed l Hardware Mapped vs. Time multiplexed/mirooded FI : y x(n) h

More information

Stable Road Lane Model Based on Clothoids

Stable Road Lane Model Based on Clothoids Stable Road Lane Model Based on Clothoids C Gakstatter*, S Thomas**, Dr P Heinemann*, Prof Gudrun Klinker*** *Audi Eletronis Venture GmbH, **Leibniz Universität Hannover, ***Tehnishe Universität Münhen

More information

PARAMETRIC SAR IMAGE FORMATION - A PROMISING APPROACH TO RESOLUTION-UNLIMITED IMAGING. Yesheng Gao, Kaizhi Wang, Xingzhao Liu

PARAMETRIC SAR IMAGE FORMATION - A PROMISING APPROACH TO RESOLUTION-UNLIMITED IMAGING. Yesheng Gao, Kaizhi Wang, Xingzhao Liu 20th European Signal Proessing Conferene EUSIPCO 2012) Buharest, Romania, August 27-31, 2012 PARAMETRIC SAR IMAGE FORMATION - A PROMISING APPROACH TO RESOLUTION-UNLIMITED IMAGING Yesheng Gao, Kaizhi Wang,

More information

arxiv: v1 [cs.gr] 10 Apr 2015

arxiv: v1 [cs.gr] 10 Apr 2015 REAL-TIME TOOL FOR AFFINE TRANSFORMATIONS OF TWO DIMENSIONAL IFS FRACTALS ELENA HADZIEVA AND MARIJA SHUMINOSKA arxiv:1504.02744v1 s.gr 10 Apr 2015 Abstrat. This work introdues a novel tool for interative,

More information

An Efficient and Scalable Approach to CNN Queries in a Road Network

An Efficient and Scalable Approach to CNN Queries in a Road Network An Effiient and Salable Approah to CNN Queries in a Road Network Hyung-Ju Cho Chin-Wan Chung Dept. of Eletrial Engineering & Computer Siene Korea Advaned Institute of Siene and Tehnology 373- Kusong-dong,

More information

Fast Rigid Motion Segmentation via Incrementally-Complex Local Models

Fast Rigid Motion Segmentation via Incrementally-Complex Local Models Fast Rigid Motion Segmentation via Inrementally-Complex Loal Models Fernando Flores-Mangas Allan D. Jepson Department of Computer Siene, University of Toronto {mangas,jepson}@s.toronto.edu Abstrat The

More information

Relevance for Computer Vision

Relevance for Computer Vision The Geometry of ROC Spae: Understanding Mahine Learning Metris through ROC Isometris, by Peter A. Flah International Conferene on Mahine Learning (ICML-23) http://www.s.bris.a.uk/publiations/papers/74.pdf

More information

Automatic Physical Design Tuning: Workload as a Sequence Sanjay Agrawal Microsoft Research One Microsoft Way Redmond, WA, USA +1-(425)

Automatic Physical Design Tuning: Workload as a Sequence Sanjay Agrawal Microsoft Research One Microsoft Way Redmond, WA, USA +1-(425) Automati Physial Design Tuning: Workload as a Sequene Sanjay Agrawal Mirosoft Researh One Mirosoft Way Redmond, WA, USA +1-(425) 75-357 sagrawal@mirosoft.om Eri Chu * Computer Sienes Department University

More information

Naïve Bayes Slides are adapted from Sebastian Thrun (Udacity ), Ke Chen Jonathan Huang and H. Witten s and E. Frank s Data Mining and Jeremy Wyatt,

Naïve Bayes Slides are adapted from Sebastian Thrun (Udacity ), Ke Chen Jonathan Huang and H. Witten s and E. Frank s Data Mining and Jeremy Wyatt, Naïve Bayes Slides are adapted from Sebastian Thrun (Udaity ), Ke Chen Jonathan Huang and H. Witten s and E. Frank s Data Mining and Jeremy Wyatt, Bakground There are three methods to establish a lassifier

More information

BioTechnology. An Indian Journal FULL PAPER. Trade Science Inc. Improvement of low illumination image enhancement algorithm based on physical mode

BioTechnology. An Indian Journal FULL PAPER. Trade Science Inc. Improvement of low illumination image enhancement algorithm based on physical mode [Type text] [Type text] [Type text] ISSN : 0974-7435 Volume 10 Issue 22 BioTehnology 2014 An Indian Journal FULL PAPER BTAIJ, 10(22), 2014 [13995-14001] Improvement of low illumination image enhanement

More information

What are Cycle-Stealing Systems Good For? A Detailed Performance Model Case Study

What are Cycle-Stealing Systems Good For? A Detailed Performance Model Case Study What are Cyle-Stealing Systems Good For? A Detailed Performane Model Case Study Wayne Kelly and Jiro Sumitomo Queensland University of Tehnology, Australia {w.kelly, j.sumitomo}@qut.edu.au Abstrat The

More information

A Dual-Hamiltonian-Path-Based Multicasting Strategy for Wormhole-Routed Star Graph Interconnection Networks

A Dual-Hamiltonian-Path-Based Multicasting Strategy for Wormhole-Routed Star Graph Interconnection Networks A Dual-Hamiltonian-Path-Based Multiasting Strategy for Wormhole-Routed Star Graph Interonnetion Networks Nen-Chung Wang Department of Information and Communiation Engineering Chaoyang University of Tehnology,

More information

A New RBFNDDA-KNN Network and Its Application to Medical Pattern Classification

A New RBFNDDA-KNN Network and Its Application to Medical Pattern Classification A New RBFNDDA-KNN Network and Its Appliation to Medial Pattern Classifiation Shing Chiang Tan 1*, Chee Peng Lim 2, Robert F. Harrison 3, R. Lee Kennedy 4 1 Faulty of Information Siene and Tehnology, Multimedia

More information

Cluster Centric Fuzzy Modeling

Cluster Centric Fuzzy Modeling 10.1109/TFUZZ.014.300134, IEEE Transations on Fuzzy Systems TFS-013-0379.R1 1 Cluster Centri Fuzzy Modeling Witold Pedryz, Fellow, IEEE, and Hesam Izakian, Student Member, IEEE Abstrat In this study, we

More information

Department of Electrical and Computer Engineering University of Wisconsin Madison. Fall

Department of Electrical and Computer Engineering University of Wisconsin Madison. Fall Department of Eletrial and Computer Engineering University of Wisonsin Madison ECE 553: Testing and Testable Design of Digital Systems Fall 2014-2015 Assignment #2 Date Tuesday, September 25, 2014 Due

More information

TUMOR DETECTION IN MRI BRAIN IMAGE SEGMENTATION USING PHASE CONGRUENCY MODIFIED FUZZY C MEAN ALGORITHM

TUMOR DETECTION IN MRI BRAIN IMAGE SEGMENTATION USING PHASE CONGRUENCY MODIFIED FUZZY C MEAN ALGORITHM TUMOR DETECTION IN MRI BRAIN IMAGE SEGMENTATION USING PHASE CONGRUENCY MODIFIED FUZZY C MEAN ALGORITHM M. Murugeswari 1, M.Gayathri 2 1 Assoiate Professor, 2 PG Sholar 1,2 K.L.N College of Information

More information

DOMAIN ADAPTATION BY ITERATIVE IMPROVEMENT OF SOFT-LABELING AND MAXIMIZATION OF NON-PARAMETRIC MUTUAL INFORMATION. M.N.A. Khan, Douglas R.

DOMAIN ADAPTATION BY ITERATIVE IMPROVEMENT OF SOFT-LABELING AND MAXIMIZATION OF NON-PARAMETRIC MUTUAL INFORMATION. M.N.A. Khan, Douglas R. DOMAIN ADAPTATION BY ITERATIVE IMPROVEMENT OF SOFT-LABELING AND MAXIMIZATION OF NON-PARAMETRIC MUTUAL INFORMATION M.N.A. Khan, Douglas R. Heisterkamp Department of Computer Siene Oklahoma State University,

More information

New Fuzzy Object Segmentation Algorithm for Video Sequences *

New Fuzzy Object Segmentation Algorithm for Video Sequences * JOURNAL OF INFORMATION SCIENCE AND ENGINEERING 24, 521-537 (2008) New Fuzzy Obet Segmentation Algorithm for Video Sequenes * KUO-LIANG CHUNG, SHIH-WEI YU, HSUEH-JU YEH, YONG-HUAI HUANG AND TA-JEN YAO Department

More information

EASY TRANSFER LEARNING BY EXPLOITING INTRA-DOMAIN STRUCTURES

EASY TRANSFER LEARNING BY EXPLOITING INTRA-DOMAIN STRUCTURES EASY TRANSFER LEARNING BY EXPLOITING INTRA-DOMAIN STRUCTURES Jindong Wang 1, Yiqiang Chen 1,, Han Yu 2, Meiyu Huang 3, Qiang Yang 4 1 Beiing Key Lab. of Mobile Computing and Pervasive Devie, Inst. of Computing

More information