Feature Extraction and Selection for Image Retrieval

Size: px
Start display at page:

Download "Feature Extraction and Selection for Image Retrieval"

Transcription

1 Feature Extracton and Selecton for Image Retreval Xang Sean Zhou, Ira Cohen, Q Tan, Thomas S. Huang Beckman Insttute for Advanced Scence and Technology Unversty of Illnos at Urbana Champagn Urbana, IL 680 {xzhou, racohen, qtan, huang}@fp.uuc.edu ABSTRACT In ths paper feature extracton process s analyzed and a new set of edge features s proposed. A revsed edge-based structural feature extracton approach s ntroduced. A prncple feature selecton algorthm s also proposed for new feature analyss and feature selecton. The results of the PFA s tested and compared to the orgnal feature set, random selectons, as well as those from Prncple Component Analyss and multvarate lnear dscrmnant analyss. The experments showed that the proposed features perform better than wavelet moment for mage retreval n a real world mage database; and the feature selected by the proposed algorthm yelds comparable results to orgnal feature set, and better results than random sets. Keywords Feature extracton, feature selecton, content-based mage retreval, prncple component analyss, dscrmnant analyss.. INTRODUCTION Image retreval problem, n some cases, can be regarded as a pattern classfcaton problem, where each mage s assumed as ground truth belongs to a specfc class. Then query-byexample s to fnd the class and return mages wthn that class. But usually there s no class labels avalable so ths can only be acheved by comparng the smlarty among mages. Applcatons nclude medcal mage database retreval, aeral/satellte mage analyss and retreval, etc., where the user s nterested n some pre-defned targets such as tumors or vehcles. However n other cases, the assumpton of fxed, statc class membershp assgnments does not hold. One example s the trademark database retreval. Even though trademarks are categorzed (by hand!) accordng to ther prmary component, a new trademark desgn wll be rejected f t s beleved to cause confuson wth an exstng one, regardless of the category assgnments. Another example can be llustrated by the mage n Fgure : as a query mage submtted by the user, t s possble that the user s lookng for cars, or the user s lookng for autumn, or house, etc. In general ths s the case for every mage n the database,.e., each mage has multple possble labels, or class membershps. And to complcate the ssue further, dfferent users can nterpret the same mage dfferently thus assgn dfferent labels n ther mnd, makng tranng or preclassfcaton much more dffcult. We refer the retreval process n ths case as dynamc matchng. Fgure An mage of multple objects In ths paper we wll dscuss the feature extracton and selecton process wth the aforementoned characterstcs of mage retreval problem n mnd. Especally for feature selecton, whch s defned as the process of selectng a subset from a set of orgnal features, dfferent cases wll lead to dfferent nputs and algorthms, whch n turn wll gve dfferent subsets of features. Wthout class labels, Prncple Component Analyss (PCA), or a novel feature selecton algorthm, Prncple Feature Analyss (PFA), can be appled to transform the features to a reduced space, or to select the subset of the features that contans the most nformaton, respectvely. PFA s ntroduced n secton 3. For the mage classfcaton case, when some class labels are avalable, multvarate lnear dscrmnant analyss (MLDA) s appled to transform the features nto the Most Dscrmnatng Feature space. The expermental results are compared n Secton 4. For dynamc matchng problem, contrbuton from each feature component would vary from tme to tme f dscrmnant analyss was conducted for dfferent labelng scheme for the mages. So the best way would be usng dynamc on-lne weghtng of the features through relevance feedback []. If a reduced feature space s desred, t can only be done based on the results of suffcent trals of the relevance feedback process, so that a feature s deleted or gnored to a great extent only f t does not contrbute n all or most of the dynamc weghtng schemes. (See Fgure 3) Feature extracton s the process of creatng a representaton for, or a transformaton from the orgnal data. There has been a phlosophcal dfference n attackng the problem of effectve and effcent feature extracton. One s human perceptoncentered approach,.e., based on human vsual percepton and

2 psychologcal experments, compute the measurements of certan percepton-based features (e.g., smoothness for texture), and then select the best mathematcal representaton for t. The other s machne-centered approach, n whch case a unfed computng scheme s selected for extracton of certan ad hoc features (e.g., co-occurrence features, or wavelet moments for texture). The emphass s frst put on the computng aspect to provde an effcent algorthm to compute the numbers ; and then the effectveness, or the correlaton between the numbers and human percepton s establshed by experments. Examples of the former nclude Tamura texture features [], color quantzaton based on human perceptons, etc. A glance at the Appendx I of Haralck [] wll gve the taste of the latter approach. Or, feature extracton from nvsble bands for hyper-spectral mage analyss, where there s no nput to the human percepton system at all. The percepton-centered approach has the obvous advantages n servng the master or user of the mage retreval system human, gven that matchng human performance s the ultmate goal of the system. But n other cases where human percepton s not the target and performs not so well, whle computer sees somethng human doesn t, the latter approach s more approprate. The two approaches were both used for color and texture feature extracton n the lterature. But n terms of edge-based structure nformaton, much less attenton has been pad n both drectons. (We defne structural features as features nbetween texture and shape. Texture captures spatal dstrbuton of llumnance varatons n the specal form of repeatng patterns. Shape represents a specfc edge feature that s related to object contour. Nether of them pays attenton to nformaton represented n non-repeatng llumnance patterns, or more specfcally, edge patterns n general, whch s the category one can defne the edge-based structural features to represent.) There have been attempts to use percepton-based edge features for cty/buldng mage retreval,.e., use the features that human uses for recognzng buldngs: edge drecton; cotermnaton ( L or U junctons) and parallel lnes[7]; etc. We propose a machne-centered approach for general-purpose edge-based structural feature extracton. To llustrate the motvaton, consder the edge maps n Fgure. In all the three edge maps shape cannot be easly extracted. But a human subject can tell the top two as the Lena mage and a buldng mage. However ths s not the case for the fox edge map at the bottom. But what can a computer do? Nothng much f understandng of the content (.e., automatc segmentaton and recognton of the object) s the goal. But f the goal s for mage retreval and matchng only,.e., fnd a representaton whch gves smlar numbers whenever two mages have smlar content, and vce versa. Now the queston becomes how to effectve and effcently represent the nformaton carred n an edge map? In Secton we propose a revsed verson of our prevous proposed water-fllng algorthm[7] to extract more features from an edge map for the purpose of mage matchng and retreval. (Note that ths can work only to the extent where the nvarance property requred by the retreval task s satsfed n the correspondng edge maps.) We also used the proposed Prncple Feature Analyss algorthm to test whether the new features carry enough new nformaton. Fgure Edge maps The scope of ths paper s depcted n Fgure 3, where the modules wth a * are the emphass of ths paper. Orgnal Image Feature Extracton* Color Texture Edge Feature Transformaton Selecton* and Dynamc Weghtng Percepton-centered Percepton-based color quantzaton Tamura texture features Percepton-based edge features w/o labels PCA, PFA* w/ labels Dscrmnant analyss* Fgure 3 Scope of ths paper Machne-centered Hstogram, moments Wavelet moments, Co-Occurrences... Water-Fllng features* Dynamc labels Dynamc weghtng. EDGE-BASED STRUCTURE FEATURE EXTRACTION In ths paper we use water-fllng algorthm to extract edge/structural features. The advantages of ths algorthm nclude effcency t s a lnear-tme algorthm; and effectveness multple features correspondng to dfferent human perceptons can be extracted smultaneously.. The Water-Fllng Algorthm We propose an algorthm to extract features from the edge map drectly wthout edge lnkng or shape representaton. The dea

3 s to look for measures for the edge length and edge structure and complexty by a very effcent graph traverse algorthm. In ths paper, we use 4-connectvty for smplcty. The algorthm also assumes that thnnng operaton has been performed on the edge map so that all the edges are one pxel wde. To llustrate the algorthm, let s frst consder the smple case of an 8 by 8 edge map wth all the edge pxels connected (Fg. 4: shaded pxels are edge pxels). The algorthm wll do a frst raster scan on the edge map and start a traverse (thnk t as fllng n water) at the frst edge pxel encountered that has less than neghbors,.e., start at an end pont. In Fg.4 the pxel wth label s the frst end pont encountered. (In case all edge pxels have at least two neghbors,.e., no end ponts but all loops, e.g., Fg. 6, then start at the frst unflled edge pxel durng the second scan. So t s necessary to have two raster scans to avod possble mss). The waterfront then flows along the edges n the order ndcated by the numbers. Note that when there are more than one possble paths to go at one pont, the waterfront wll fork (e.g., at pxel 6 and 9 n Fg. ) Fgure 4 Fgure One can see that ths algorthm can be regarded as a smulaton of floodng of connected canal systems (.e., connected edges), hence the name water-fllng algorthm. The assumptons mpled by the algorthm nclude: ) we have unlmted water supply at the startng pxel; ) water flows at a constant speed n all drectons; ) water front wll stop only at a dead-end, or at a pont where all possble drectons have been flled. When there are more than one set of connected edges n the edge map (e.g., the case n Fg.), the algorthm wll fll all the sets ndependently n sequental or n parallel. As water flls the canals (edges), varous nformaton are extracted, whch are stored as the feature prmtves. Feature vectors can then be constructed based on these feature prmtves. The tme complexty of ths algorthm s lnear, proportonal to the number of edge ponts n the mage.. Feature Extracton Feature Prmtves We propose the concept of Feature prmtves, whch are defned as the quanttes assocated wth or calculated from an mage that can serve as bases for constructng feature vectors, often through usng ther statstcs or entropes. Feature prmtves can be used as feature vector drectly as well, but often they are not compact enough. For example, co-occurrence matrces are the feature prmtves for the co-occurrence texture features, most of whch are moments, correlatons, and entropes []; and wavelet transform coeffcents can be regarded as feature prmtves for wavelet based texture features such as wavelet moments. In our case, we propose the followng quanttes as structural feature prmtves: ). Fllng tme: Fllng tme s the tme for water to fll a set of connected edges. For Fg. 4 through 6, the fllng tmes are {}, {4,, }, and {}, respectvely. Usng dfferent startng pxels, the fllng tme can vary n a range of [t, t], where t s the mnmum fllng tme among all possble selecton of the startng pxels. Ths s easly proved as follows: denote the startng pxel that gves the mnmum fllng tme t as S. Snce we assume water runs n both drectons on an edge, water can reach S from any other startng pxel, say P, n tme t, and then the waterfront can go from S to reach any pxels left unflled wthn tme t. So the fllng tme from P s =t. To mnmze the varaton n fllng tme due to selecton of startng pxels, we can mpose addtonal constrants on the selecton of startng pxels (e.g., choose only the end ponts), or choose dfferent startng pxels and average the results. To acheve scalng nvarance, normalze the fllng tme accordng to the mage sze. For example, dvde fllng tme by (wdth + heght). ). Fork count: Fork count s the total number of branches the waterfront has forked durng the fllng of a set of edges. If we consder the ntal waterfront as one branch to start wth, then the fork count for Fg. 4 through 6 are {3}, {, 3, }, and {6}, respectvely. If we choose an end pxel as startng pxel whenever possble, fork count s almost nvarant to startng pxel selecton. The varaton s wthn ±, depend upon whether the water starts from the mddle of an edge or the end of the edge. Also f multple waterfronts collde at one ntersecton, even though the water does not actually fork, the potental forks should be counted to acheve the almost nvarance. E.g., an extra fork s counted both at the upper 9 and the lower 0 n Fg. 6; but none at, snce t s not a potental fork pont n any way). 3). Loop count: Loop count s the number of smple loops (or, smple cycles as defned n Corman et al., 997, p. 88) n a set of connected edges. For example, n Fg.4 through 6, the loop counts are {}, {0,, 0}, and {3}, respectvely. Loop count s nvarant to rotaton Fgure 6 Fgure 7 To get the loop count durng the water-fllng process, we make use of the followng Theorem of Splashes : S E N 3

4 If we assume that when two waterfronts collde, we see one splash ; or more generally, when n waterfronts collde at one ntersecton, we see n- splashes (thnk t as n- waterfronts collde wth the frst waterfront sequentally). Then the number of splashes = the number of smple loops. For example, n Fg. 6, three splashes are recorded at tme 9, 0, and. Hence the loop count s 3. Proof: Treat the set of connected edges as a graph; and regard the startng pont, the fork ponts, and the collson/splash ponts as the nodes of the graph; and the water branches as the edges. For example, the correspondng graph for Fg. 6 s shown n Fg. 7. Then the water-fllng process s a traverse of the graph, and the total number of splashes s the total number of revsts to some of the nodes, whch s the number of smple cycles/loops n the graph. The above theorem provdes a way of recordng loop counts wthn the water-fllng process wth very lttle overhead computaton. 4). Water amount: Water amount s the total amount of water used to fll up the set of edges n terms of number of pxels. So t s the edge pxel count. In Fg. 4 through 6, the water amounts are {8}, {4, 8, }, and {9}, respectvely. ). Horzontal (vertcal) cover: Horzontal (vertcal) cover s the wdth (heght) of the rectangular boundng box of the set of edges. In Fg. 4 through 6, the horzontal covers are {}, {6,, }, and {7}, respectvely. 6). Longest horzontal (vertcal) flow: Longest horzontal (vertcal) flow s the longest horzontal (vertcal) edge n the set of connected edges. For Fg. 4 through 6, the longest vertcal flows are {6}, {8, 6, }, and {6}, respectvely. Note that there exst many other possbltes on selectng the feature prmtves. However, the fnal selecton should depend upon the specfc applcaton,.e., what nformaton s mportant and most dscrmnatve toward the classfcaton... Edge/structural feature formaton Based on the feature prmtves n 4., we can then construct edge/structural features from ther statstcs. For example: moments (e.g., average fllng tme); order statstcs (e.g., maxmum loop count); dstrbutons (e.g., water amount hstogram); etc. In the followng we dscuss some examples wth the emphass on ther meanngs from a human percepton pont of vew. ). MaxFllngTme and the assocated ForkCount MaxFllngTme(MFC) s defned as max{fllng tmes}. For Fg. 4 through 6, the MFC s, 4, and, respectvely. And the assocated ForkCount(FC) s 3,, and 6 respectvely. So the MFT&FC vectors for Fg. 4 through 6 are (, 3), (4, ), and (, 6), respectvely. MFT&FC are features most probably assocated wth a salent object (wth the longest edge) n the mage. The MFT conveys a rough measure of the sze (edge length) of ths object, whle the assocated FC gves measure of complexty of the structure of the object (complexty of the edges). ). MaxForkCount and the assocated FllngTme Smlarly defned as MFT&FC, these are also features most probably assocated wth a salent object n the mage. The MFT conveys a rough measure of the complexty of the object. Ths object may or may NOT be the same object as the prevous one. For Fg. 4 and 6, the MFC&FT s the same as the MFT&FC. But for Fg., the MFC&FT vector s (3, ). 3). (GLC&MLC) GlobalLoopCount and MaxLoopCount GlobalLoopCount s defned as sum{loop counts}. MaxLoopCount s max{loop counts}. Ths feature vector can capture structural nformaton such as the wndows n the buld mages. Or can be used toward character detect and recognton applcatons. 4). (FTH&FC) FllngTme Hstogram and the assocated averaged ForkCount wthn each bn Ths s a global feature on all sets of connected edges n the edge map. It represents the edge map by the dstrbuton of edge length. Nose or changng background wth short edges may only affect part of the hstogram, leavng the porton depctng the salent objects unchanged. Thus by proper weghtng of the components (e.g. by relevance feedback[]), we could acheve robust retreval. ). (WAH) WaterAmount Hstogram Ths s also a global feature wth multple components. It s another measure of the dstrbuton n edge length or densty. Note agan that there can be many other possble ways to construct feature vectors, such as the moments of fllng tmes, fork counts, loop counts, or water amounts, etc., 3. FEATURE TRANSFORMATION AND SELECTION Fgure 8 st PC Most dscrmnant projecton drecton For any handcrafted feature set, data-dependent analyss can always be carred out to select an optmal projecton or transformaton to best represent the nformaton carred n the orgnal feature set, or reduce the dmensonalty wth mnmum nformaton loss. When there s no label for the data, prncple components analyss [9] gves the best lnear transform from the orgnal feature set n terms of reconstructon errors. If class labels are (partally) avalable, 4

5 then multvarate lnear dscrmnant analyss (MLDA) [] gves the best lnear transformaton n terms of maxmzng the rato of nter-class scatter over n-class scatter,.e., the MLDA vector has the most dscrmnatng power among all lnear transformatons. The resultng transformaton of PCA or MLDA s n general dfferent. Fgure 8 shows the data scatter of the whole dataset, and two classes by three eclpses, and the two projecton drectons are shown n dotted lne. Both of the above technques requre the orgnal feature set for the calculaton of the transformed features. If computaton s an ssue n the feature extracton process, or some of the orgnal feature component s of specal meanng and the user want to keep them unchanged, whle dmenson reducton s stll desred, then feature selecton technque s needed. [8] gves a detaled comparson on the feature selecton algorthms for labeled data. Here we propose a Prncple Feature Selecton algorthm for un-labeled data,.e., ths algorthm wll output the subset of the features that wll best represent orgnal data. 3. Prncple Component Analyss (PCA) Prncple components are the projecton of the orgnal features onto the egenvectors corresponds to the largest egenvalues of the covarance matrx of the orgnal feature set. Prncple components provde lnear representaton of the orgnal data usng the least number of components wth the mean-squared error mnmzed. (See [9] for detals) 3. Prncple Feature Analyss (PFA) Let X be a zero mean n-dmensonal random feature vector. Let Σ be the covarance matrx of X (whch could be n correlaton form as well). Let A be a matrx whose columns are the orthonormal egenvectors of the matrx Σ, computed usng the sngular value decomposton of the Σ: λ where Λ= and T A A = I n. 0 T Σ = AΛA () 0. λ n λ λ... λ, n Let A q be the frst q columns of A and let V, V,... V n R be the rows of the matrx A q. The vector V corresponds to the th feature (varable) n the vector X, and the coeffcents of V correspond to the weghts of that feature on each axes of the subspace. Features that are hghly correlated wll have smlar absolute value weght vectors (absolute value s necessary snce changng the sgn of one varable changes the sgns of the correspondng weghts but has no statstcal sgnfcance [9]). In order to fnd the best subset we wll use the structure of these rows to frst fnd the features whch are hghly correlated to each other and then choose from each group of correlated features the one whch wll represent that group optmally n terms of hgh spread n q the lower dmenson, reconstructon and nsenstvty to nose. The algorthm can be summarzed n the followng fve steps: Step Step Step 3 Compute the Sample Covarance matrx, or use the true covarance matrx f t s avalable. In some cases t would be preferred to use the Correlaton matrx nstead of the Covarance matrx. The Correlaton matrx s defned as the nxn matrx whose,j th entry s ρ j E = E[ x [ x x ] j ] E[ x Ths representaton s preferred n cases where the features have very dfferent varances from each other, and usng the regular covarance form wll cause the PCA to put very heavy weghts on the features wth the hghest varances. See [9] for more detals. Compute the Prncpal components and egenvalues of the Covarance/Correlaton matrx as defned n equaton (). Choose the subspace dmenson q and construct the matrx A q from A. Ths can be chosen by decdng how much of the varablty of the data s desred to be retaned. The retaned varablty can be computed usng: j Varablty Re taned = ] q = n = λ λ 00% Step 4 Cluster the vectors V, V,..., V n R to p q clusters usng K-Means algorthm. The dstance measure used for the K-Means algorthm s the Eucldean dstance. The vectors are clustered n p clusters and the means of each cluster s computed. Ths s an teratve stage whch repeats tself untl the p clusters are found and do not change. The reason to choose p greater then q n some cases s f the same retaned varablty as the PCA s desred, a slghtly hgher number of features s needed (Usually - more are enough). Step In each cluster, fnd the correspondng vector V whch s closest to the mean of the cluster. Choose the correspondng feature x as a prncpal feature. Ths wll yeld the choce of p features. The reason for choosng the vector nearest to the mean s twofold. Ths feature can be thought of as the central feature of that cluster- the one most domnant n t, and whch holds the least redundant nformaton of features n other clusters. Thus t satsfes both of the propertes we wanted to acheve- large spread n the lower dmensonal space, and good representaton of the orgnal data. Note that f a cluster has just two components n t, the chosen feature wll be the one q

6 whch has the hghest varance of the coeffcents, snce ths means that t has better separablty power n the subspace. By choosng the prncpal features usng ths algorthm, we choose the subset that represents well the entre feature set both n terms of retanng the varatons n the feature space (by usng the clusterng procedure), and keep the predcton error at a mnmum (by choosng the feature who s vector s closest to the mean of the cluster). (See [9] for more detals.) 3.3 Multvarate Lnear Dscrmnant Analyss (MLDA) Assume that S w s the n-class scatter matrx, S b s the nterclass scatter matrx, then the most dcrmnant projecton drectons are the egenvectors of the S w - S b assocated wth the largest egenvalues. (For detals see []) 4. Experments and Comparsons The frst set of experments s desgned to test the performance of the water-fllng features versus the wdely used waveletmoments as texture measures. Water-fllng features used n these experments are MFT&FC (see Secton for defnton), MFC&FT, GLC&MLC, and FTH&FC (7 bns) a total of 8 feature components per mage. Eucldean dstance s used as dstance measure. The datasets from Corel conssts of 769 mages, 400 of whch are arplanes, and 00 of them are Amercan eagles. Table shows the comparson n terms of averaged number of hts n top 0, 0, 40 and 80 returns for 00 arplanes and 00 eagles as the query mages, respectvely. Table. Water-fllng (WF) versus Wavelet Varances (WM) Arplanes top 0 Top 0 Top 40 Top 80 WF WV Eagles top 0 Top 0 Top 40 Top 80 WF WV Prncple feature analyss s appled on the jont feature set, whch s the unon of the Water-fllng features and wavelet moment features. For three datasets,,, and 3 pars of features from dfferent sets are clustered together, all wth low correlaton coeffcents (<0.), whch ndcates that waterfllng features contan mostly new nformaton than that s already expressed by wavelet moments. The second sets of experments are desgned to test the output of the prncple feature analyss algorthm (PF) for the purpose of mage retreval. Table shows the retreval results on the same sets of query mages as the frst experment (00 arplanes and 00 eagles). The fourth test n each case corresponds to the same number of features selected, but they span less number of clusters. One can see that the performance s consstently lower. Table 3 shows the testng results on the VISTEX texture database from MIT Meda Lab. In ths database the orgnal mages are tled nto 4X4=6 sub mages for testng purpose, so there are bg mages and 83 resultng sub mages. Snce we have the classes and labels, MLDA s tested to compare wth all the other cases. One can clearly see that MLDA outperform all the cases. Ths s reasonable snce t utlzes extra nformaton. Please note that the last test case n Table 3 have more feature components (6) than that of the prncple features selected (), but snce these feature components are chosen as clustered ones (feature, 4, 8 and 0 are all clustered nto one cluster for ths dataset!), so the performance s even worse, even though one mght ntutvely expect t to perform better snce they spanned all the three wavelet sub-bands! Table. Prncple Features (PF) Performance Comparson usng 0 wavelet varances (WV) for Corel dataset Arplanes top 0 Top 0 Top 40 Top 80 0 WV s PC s PF s (,,6,7,8,9,0) (,,3,4,8,9,0) Eagles top 0 Top 0 Top 40 Top 80 0 WV s PC s PF s (,,6,7,8,9,0) (,,3,4,8,9,0) Table 3. Prncple Features (PF) Performance Comparson usng 0 wavelet varances (WV) for VISTEX dataset VISTEX top 0 Top 0 Top 40 Top 80 0 WV s PC s MLDA s PF s (,,6,7,0) (,3,4,6,8,0) Table 4. Prncple Features (PF) Performance Comparson usng 9 color moments and 0 wavelet moments (WM) 40 random queres n Corel top 0 Top 0 All:9 color + 0 WM PF s (7 color, WM)

7 random selected (I) (7 color, WM) random selected (II) (7 color, WM) random selected (III) (7 color, WM) Table 4 shows results on the 9 color moments plus 0 wavelet moments on the Corel database. 7 out 9 color features and out of 0 wavelet feature are selected by PFA. 40 randomly selected query mages are pcked. The performance of the orgnal feature set and the prncple set are tested aganst 3 randomly selected (but the same number of) features. It clearly shows that Prncple features yeld comparable results as that of orgnal set and sgnfcantly hgher results than any random pcks.. CONCLUSIONS In ths paper we proposed a revsed edge-based structural feature extracton approach by followng gudelnes obtaned from summarzng the exstng feature extracton approaches. A prncple feature selecton algorthm s also proposed for new feature analyss and feature selecton. The results of the PFA s tested and compared to the orgnal feature set, random selectons, as well as those from Prncple Component Analyss and multvarate lnear dscrmnant analyss. The experments showed that the proposed algorthm yeld comparable results to orgnal set, and better results than random sets. 6. REFERENCES [] Haralck, R. M., K. Shanmugam, and I. Dnsten, Texture feature for mage classfcaton, IEEE Trans. System Man and Cybernetcs (Nov., 973) [] Tamura, Hdeyuk, S. Mor, and T. Yamawak, Texture Features Correspondng to Vsual Percepton, IEEE Trans. System Man and Cybernetcs, 8 (6), (978). [3] Corman, T. H., C. E. Leserson, R. L. Rvest. Introducton to algorthms, McGraw-Hll, New York, 997 [4] Flckner, M. et al., Query by mage and vdeo content: The qbc system, IEEE Computers, 99 [] Gonzalez, R. C. and Woods, Dgtal Image Processng, Addson-Wesley, 99 [6] Hu, M. K., Vsual pattern recognton by moment nvarants, IRE Trans. Informaton Theory, 8, 96 [7] Iqbal, Q. and J. K. Aggarwal, Applyng perceptual groupng to content-based mage retreval: Buldng mages, Proc. IEEE CVPR 99, 4-48, 999 [8] Jan, A. K., Fundamentals of Dgtal Image Processng, Prentce Hall, 989 [9] Jollffe, I.T., Prncpal Component Analyss, Sprnger- Verlag, New-York, 986. [0] Lane, A., J. Fan. Texture classfcaton by wavelet packet sgnatures. IEEE Trans. Pattern Anal. Machne Intell., 993 [] Ru, Y., T. S. Huang, M. Ortega, and S. Mehrotra, Relevance Feedback: A Power Tool n Interactve Content-Based Image Retreval, IEEE Tran on Crcuts and Systems for Vdeo Technology, Vol 8, No., Sept., 644-6, 998 [] Smth, J. R. and Chang, Transform features for texture classfcaton and dscrmnaton n large mage databases, Proc. IEEE ICIP, 99 [3] Tabb, M., N. Ahuja, Multscale Image Segmentaton by Integrated Edge and Regon Detecton, IEEE trans. Image Processng, Vol. 6, No., May 997 [4] Vasconcelos, N. and A. Lppman, Embedded mxture modelng for effcent probablstc content-based ndexng and retreval, n SPIE Multmeda Storage and Archvng Systems III, Boston, 998 [] Wlks, S.S., Mathematcal Statstcs. New York: Wley, (963) [6] Zahn, C. T. and Roskes, Fourer descrptors for plane closed curves, IEEE Trans. Computers, 97 [7] Zhou, X. S., Y. Ru, and T. S. Huang, Water-fllng: a novel way for mage structural feature extracton, Proc. ICIP, 999 [8] Jan, A., A.,Zongker, Feature Selecton: Evaluaton, Applcaton, and Small Sample Performance, IEEE Trans. PAMI, VOL. 9, (Feb, 997) [9] I. Cohen, Q. Tan, X. Zhou, T. S. Huang, Feature Selecton and Dmensonalty Reducton Usng Prncple Feature Analyss", submtted to the Sxteenth Conference on Uncertanty n Artfcal Intellgence (UAI-000) Workshop on Fuson of Doman Knowledge wth Data For Decson Support, Stanford Unversty, Stanford, CA, June 30,

Feature Reduction and Selection

Feature Reduction and Selection Feature Reducton and Selecton Dr. Shuang LIANG School of Software Engneerng TongJ Unversty Fall, 2012 Today s Topcs Introducton Problems of Dmensonalty Feature Reducton Statstc methods Prncpal Components

More information

A Fast Content-Based Multimedia Retrieval Technique Using Compressed Data

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

More information

A 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

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

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

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

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

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

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

Detection of an Object by using Principal Component Analysis

Detection of an Object by using Principal Component Analysis Detecton of an Object by usng Prncpal Component Analyss 1. G. Nagaven, 2. Dr. T. Sreenvasulu Reddy 1. M.Tech, Department of EEE, SVUCE, Trupath, Inda. 2. Assoc. Professor, Department of ECE, SVUCE, Trupath,

More information

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

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

12/2/2009. Announcements. Parametric / Non-parametric. Case-Based Reasoning. Nearest-Neighbor on Images. Nearest-Neighbor Classification

12/2/2009. Announcements. Parametric / Non-parametric. Case-Based Reasoning. Nearest-Neighbor on Images. Nearest-Neighbor Classification Introducton to Artfcal Intellgence V22.0472-001 Fall 2009 Lecture 24: Nearest-Neghbors & Support Vector Machnes Rob Fergus Dept of Computer Scence, Courant Insttute, NYU Sldes from Danel Yeung, John DeNero

More information

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

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

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

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

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

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

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

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

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

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

Fuzzy C-Means Initialized by Fixed Threshold Clustering for Improving Image Retrieval

Fuzzy C-Means Initialized by Fixed Threshold Clustering for Improving Image Retrieval Fuzzy -Means Intalzed by Fxed Threshold lusterng for Improvng Image Retreval NAWARA HANSIRI, SIRIPORN SUPRATID,HOM KIMPAN 3 Faculty of Informaton Technology Rangst Unversty Muang-Ake, Paholyotn Road, Patumtan,

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

Laplacian Eigenmap for Image Retrieval

Laplacian Eigenmap for Image Retrieval Laplacan Egenmap for Image Retreval Xaofe He Partha Nyog Department of Computer Scence The Unversty of Chcago, 1100 E 58 th Street, Chcago, IL 60637 ABSTRACT Dmensonalty reducton has been receved much

More information

Comparison Study of Textural Descriptors for Training Neural Network Classifiers

Comparison Study of Textural Descriptors for Training Neural Network Classifiers Comparson Study of Textural Descrptors for Tranng Neural Network Classfers G.D. MAGOULAS (1) S.A. KARKANIS (1) D.A. KARRAS () and M.N. VRAHATIS (3) (1) Department of Informatcs Unversty of Athens GR-157.84

More information

Image Alignment CSC 767

Image Alignment CSC 767 Image Algnment CSC 767 Image algnment Image from http://graphcs.cs.cmu.edu/courses/15-463/2010_fall/ Image algnment: Applcatons Panorama sttchng Image algnment: Applcatons Recognton of object nstances

More information

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

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

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

More information

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

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

More information

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

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

An Entropy-Based Approach to Integrated Information Needs Assessment

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

More information

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

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

More information

Local Quaternary Patterns and Feature Local Quaternary Patterns

Local Quaternary Patterns and Feature Local Quaternary Patterns Local Quaternary Patterns and Feature Local Quaternary Patterns Jayu Gu and Chengjun Lu The Department of Computer Scence, New Jersey Insttute of Technology, Newark, NJ 0102, USA Abstract - Ths paper presents

More information

Modular PCA Face Recognition Based on Weighted Average

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

More information

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

Video Content Representation using Optimal Extraction of Frames and Scenes

Video Content Representation using Optimal Extraction of Frames and Scenes Vdeo Content Representaton usng Optmal Etracton of rames and Scenes Nkolaos D. Doulam Anastasos D. Doulam Yanns S. Avrths and Stefanos D. ollas Natonal Techncal Unversty of Athens Department of Electrcal

More information

Helsinki University Of Technology, Systems Analysis Laboratory Mat Independent research projects in applied mathematics (3 cr)

Helsinki University Of Technology, Systems Analysis Laboratory Mat Independent research projects in applied mathematics (3 cr) Helsnk Unversty Of Technology, Systems Analyss Laboratory Mat-2.08 Independent research projects n appled mathematcs (3 cr) "! #$&% Antt Laukkanen 506 R ajlaukka@cc.hut.f 2 Introducton...3 2 Multattrbute

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

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

3D vector computer graphics

3D vector computer graphics 3D vector computer graphcs Paolo Varagnolo: freelance engneer Padova Aprl 2016 Prvate Practce ----------------------------------- 1. Introducton Vector 3D model representaton n computer graphcs requres

More information

A Gradient Difference based Technique for Video Text Detection

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

More information

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

Human Face Recognition Using Generalized. Kernel Fisher Discriminant

Human Face Recognition Using Generalized. Kernel Fisher Discriminant Human Face Recognton Usng Generalzed Kernel Fsher Dscrmnant ng-yu Sun,2 De-Shuang Huang Ln Guo. Insttute of Intellgent Machnes, Chnese Academy of Scences, P.O.ox 30, Hefe, Anhu, Chna. 2. Department of

More information

User Authentication Based On Behavioral Mouse Dynamics Biometrics

User Authentication Based On Behavioral Mouse Dynamics Biometrics User Authentcaton Based On Behavoral Mouse Dynamcs Bometrcs Chee-Hyung Yoon Danel Donghyun Km Department of Computer Scence Department of Computer Scence Stanford Unversty Stanford Unversty Stanford, CA

More information

A Gradient Difference based Technique for Video Text Detection

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

More information

USING GRAPHING SKILLS

USING GRAPHING SKILLS Name: BOLOGY: Date: _ Class: USNG GRAPHNG SKLLS NTRODUCTON: Recorded data can be plotted on a graph. A graph s a pctoral representaton of nformaton recorded n a data table. t s used to show a relatonshp

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

A Clustering Algorithm for Key Frame Extraction Based on Density Peak

A Clustering Algorithm for Key Frame Extraction Based on Density Peak Journal of Computer and Communcatons, 2018, 6, 118-128 http://www.scrp.org/ournal/cc ISSN Onlne: 2327-5227 ISSN Prnt: 2327-5219 A Clusterng Algorthm for Key Frame Extracton Based on Densty Peak Hong Zhao

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

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

A Deflected Grid-based Algorithm for Clustering Analysis

A Deflected Grid-based Algorithm for Clustering Analysis A Deflected Grd-based Algorthm for Clusterng Analyss NANCY P. LIN, CHUNG-I CHANG, HAO-EN CHUEH, HUNG-JEN CHEN, WEI-HUA HAO Department of Computer Scence and Informaton Engneerng Tamkang Unversty 5 Yng-chuan

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

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

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

More information

The Research of Support Vector Machine in Agricultural Data Classification

The Research of Support Vector Machine in Agricultural Data Classification The Research of Support Vector Machne n Agrcultural Data Classfcaton Le Sh, Qguo Duan, Xnmng Ma, Me Weng College of Informaton and Management Scence, HeNan Agrcultural Unversty, Zhengzhou 45000 Chna Zhengzhou

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

Mathematics 256 a course in differential equations for engineering students

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

More information

COMPLEX WAVELET TRANSFORM-BASED COLOR INDEXING FOR CONTENT-BASED IMAGE RETRIEVAL

COMPLEX WAVELET TRANSFORM-BASED COLOR INDEXING FOR CONTENT-BASED IMAGE RETRIEVAL COMPLEX WAVELET TRANSFORM-BASED COLOR INDEXING FOR CONTENT-BASED IMAGE RETRIEVAL Nader Safavan and Shohreh Kasae Department of Computer Engneerng Sharf Unversty of Technology Tehran, Iran skasae@sharf.edu

More information

R s s f. m y s. SPH3UW Unit 7.3 Spherical Concave Mirrors Page 1 of 12. Notes

R s s f. m y s. SPH3UW Unit 7.3 Spherical Concave Mirrors Page 1 of 12. Notes SPH3UW Unt 7.3 Sphercal Concave Mrrors Page 1 of 1 Notes Physcs Tool box Concave Mrror If the reflectng surface takes place on the nner surface of the sphercal shape so that the centre of the mrror bulges

More information

A Multi-step Strategy for Shape Similarity Search In Kamon Image Database

A Multi-step Strategy for Shape Similarity Search In Kamon Image Database A Mult-step Strategy for Shape Smlarty Search In Kamon Image Database Paul W.H. Kwan, Kazuo Torach 2, Kesuke Kameyama 2, Junbn Gao 3, Nobuyuk Otsu 4 School of Mathematcs, Statstcs and Computer Scence,

More information

A Novel Adaptive Descriptor Algorithm for Ternary Pattern Textures

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

More information

An Improved Image Segmentation Algorithm Based on the Otsu Method

An Improved Image Segmentation Algorithm Based on the Otsu Method 3th ACIS Internatonal Conference on Software Engneerng, Artfcal Intellgence, Networkng arallel/dstrbuted Computng An Improved Image Segmentaton Algorthm Based on the Otsu Method Mengxng Huang, enjao Yu,

More information

Performance Evaluation of Information Retrieval Systems

Performance Evaluation of Information Retrieval Systems Why System Evaluaton? Performance Evaluaton of Informaton Retreval Systems Many sldes n ths secton are adapted from Prof. Joydeep Ghosh (UT ECE) who n turn adapted them from Prof. Dk Lee (Unv. of Scence

More information

Lecture 5: Multilayer Perceptrons

Lecture 5: Multilayer Perceptrons Lecture 5: Multlayer Perceptrons Roger Grosse 1 Introducton So far, we ve only talked about lnear models: lnear regresson and lnear bnary classfers. We noted that there are functons that can t be represented

More information

A New Feature of Uniformity of Image Texture Directions Coinciding with the Human Eyes Perception 1

A New Feature of Uniformity of Image Texture Directions Coinciding with the Human Eyes Perception 1 A New Feature of Unformty of Image Texture Drectons Concdng wth the Human Eyes Percepton Xng-Jan He, De-Shuang Huang, Yue Zhang, Tat-Mng Lo 2, and Mchael R. Lyu 3 Intellgent Computng Lab, Insttute of Intellgent

More information

Background Removal in Image indexing and Retrieval

Background Removal in Image indexing and Retrieval Background Removal n Image ndexng and Retreval Y Lu and Hong Guo Department of Electrcal and Computer Engneerng The Unversty of Mchgan-Dearborn Dearborn Mchgan 4818-1491, U.S.A. Voce: 313-593-508, Fax:

More information

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

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

More information

Analysis of Continuous Beams in General

Analysis of Continuous Beams in General Analyss of Contnuous Beams n General Contnuous beams consdered here are prsmatc, rgdly connected to each beam segment and supported at varous ponts along the beam. onts are selected at ponts of support,

More information

Histogram-Enhanced Principal Component Analysis for Face Recognition

Histogram-Enhanced Principal Component Analysis for Face Recognition Hstogram-Enhanced Prncpal Component Analyss for Face ecognton Ana-ara Sevcenco and Wu-Sheng Lu Dept. of Electrcal and Computer Engneerng Unversty of Vctora sevcenco@engr.uvc.ca, wslu@ece.uvc.ca Abstract

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

Wishing you all a Total Quality New Year!

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

More information

Module Management Tool in Software Development Organizations

Module Management Tool in Software Development Organizations Journal of Computer Scence (5): 8-, 7 ISSN 59-66 7 Scence Publcatons Management Tool n Software Development Organzatons Ahmad A. Al-Rababah and Mohammad A. Al-Rababah Faculty of IT, Al-Ahlyyah Amman Unversty,

More information

X- Chart Using ANOM Approach

X- Chart Using ANOM Approach ISSN 1684-8403 Journal of Statstcs Volume 17, 010, pp. 3-3 Abstract X- Chart Usng ANOM Approach Gullapall Chakravarth 1 and Chaluvad Venkateswara Rao Control lmts for ndvdual measurements (X) chart are

More information

An Optimal Algorithm for Prufer Codes *

An Optimal Algorithm for Prufer Codes * J. Software Engneerng & Applcatons, 2009, 2: 111-115 do:10.4236/jsea.2009.22016 Publshed Onlne July 2009 (www.scrp.org/journal/jsea) An Optmal Algorthm for Prufer Codes * Xaodong Wang 1, 2, Le Wang 3,

More information

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

Learning-Based Top-N Selection Query Evaluation over Relational Databases

Learning-Based Top-N Selection Query Evaluation over Relational Databases Learnng-Based Top-N Selecton Query Evaluaton over Relatonal Databases Lang Zhu *, Wey Meng ** * School of Mathematcs and Computer Scence, Hebe Unversty, Baodng, Hebe 071002, Chna, zhu@mal.hbu.edu.cn **

More information

Load Balancing for Hex-Cell Interconnection Network

Load Balancing for Hex-Cell Interconnection Network Int. J. Communcatons, Network and System Scences,,, - Publshed Onlne Aprl n ScRes. http://www.scrp.org/journal/jcns http://dx.do.org/./jcns.. Load Balancng for Hex-Cell Interconnecton Network Saher Manaseer,

More information

Related-Mode Attacks on CTR Encryption Mode

Related-Mode Attacks on CTR Encryption Mode Internatonal Journal of Network Securty, Vol.4, No.3, PP.282 287, May 2007 282 Related-Mode Attacks on CTR Encrypton Mode Dayn Wang, Dongda Ln, and Wenlng Wu (Correspondng author: Dayn Wang) Key Laboratory

More information

Tsinghua University at TAC 2009: Summarizing Multi-documents by Information Distance

Tsinghua University at TAC 2009: Summarizing Multi-documents by Information Distance Tsnghua Unversty at TAC 2009: Summarzng Mult-documents by Informaton Dstance Chong Long, Mnle Huang, Xaoyan Zhu State Key Laboratory of Intellgent Technology and Systems, Tsnghua Natonal Laboratory for

More information

Face Recognition using 3D Directional Corner Points

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

More information

Using Fuzzy Logic to Enhance the Large Size Remote Sensing Images

Using Fuzzy Logic to Enhance the Large Size Remote Sensing Images Internatonal Journal of Informaton and Electroncs Engneerng Vol. 5 No. 6 November 015 Usng Fuzzy Logc to Enhance the Large Sze Remote Sensng Images Trung Nguyen Tu Huy Ngo Hoang and Thoa Vu Van Abstract

More information

Type-2 Fuzzy Non-uniform Rational B-spline Model with Type-2 Fuzzy Data

Type-2 Fuzzy Non-uniform Rational B-spline Model with Type-2 Fuzzy Data Malaysan Journal of Mathematcal Scences 11(S) Aprl : 35 46 (2017) Specal Issue: The 2nd Internatonal Conference and Workshop on Mathematcal Analyss (ICWOMA 2016) MALAYSIAN JOURNAL OF MATHEMATICAL SCIENCES

More information

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

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

Combination of Color and Local Patterns as a Feature Vector for CBIR

Combination of Color and Local Patterns as a Feature Vector for CBIR Internatonal Journal of Computer Applcatons (975 8887) Volume 99 No.1, August 214 Combnaton of Color and Local Patterns as a Feature Vector for CBIR L.Koteswara Rao Asst.Professor, Dept of ECE Faculty

More information

Feature Selection for Target Detection in SAR Images

Feature Selection for Target Detection in SAR Images Feature Selecton for Detecton n SAR Images Br Bhanu, Yngqang Ln and Shqn Wang Center for Research n Intellgent Systems Unversty of Calforna, Rversde, CA 95, USA Abstract A genetc algorthm (GA) approach

More information

Feature Extractions for Iris Recognition

Feature Extractions for Iris Recognition Feature Extractons for Irs Recognton Jnwook Go, Jan Jang, Yllbyung Lee, and Chulhee Lee Department of Electrcal and Electronc Engneerng, Yonse Unversty 134 Shnchon-Dong, Seodaemoon-Gu, Seoul, KOREA Emal:

More information

Detection of Human Actions from a Single Example

Detection of Human Actions from a Single Example Detecton of Human Actons from a Sngle Example Hae Jong Seo and Peyman Mlanfar Electrcal Engneerng Department Unversty of Calforna at Santa Cruz 1156 Hgh Street, Santa Cruz, CA, 95064 {rokaf,mlanfar}@soe.ucsc.edu

More information

Edge/Structural Features for Content Based Image Retrieval

Edge/Structural Features for Content Based Image Retrieval Edge/Structural Features for Content Based Image Retrieval Xiang Sean Zhou, Thomas S. Huang Beckman Institute for Advanced Science and Technology University of Illinois at Urbana Champaign, Urbana, IL

More information

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

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

More information

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

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

More information

Machine Learning. Topic 6: Clustering

Machine Learning. Topic 6: Clustering Machne Learnng Topc 6: lusterng lusterng Groupng data nto (hopefully useful) sets. Thngs on the left Thngs on the rght Applcatons of lusterng Hypothess Generaton lusters mght suggest natural groups. Hypothess

More information

Classification of Face Images Based on Gender using Dimensionality Reduction Techniques and SVM

Classification of Face Images Based on Gender using Dimensionality Reduction Techniques and SVM Classfcaton of Face Images Based on Gender usng Dmensonalty Reducton Technques and SVM Fahm Mannan 260 266 294 School of Computer Scence McGll Unversty Abstract Ths report presents gender classfcaton based

More information

CLASSIFICATION OF ULTRASONIC SIGNALS

CLASSIFICATION OF ULTRASONIC SIGNALS The 8 th Internatonal Conference of the Slovenan Socety for Non-Destructve Testng»Applcaton of Contemporary Non-Destructve Testng n Engneerng«September -3, 5, Portorož, Slovena, pp. 7-33 CLASSIFICATION

More information

Mining Image Features in an Automatic Two- Dimensional Shape Recognition System

Mining Image Features in an Automatic Two- Dimensional Shape Recognition System Internatonal Journal of Appled Mathematcs and Computer Scences Volume 2 Number 1 Mnng Image Features n an Automatc Two- Dmensonal Shape Recognton System R. A. Salam, M.A. Rodrgues Abstract The number of

More information

Edge-Based Structural Features for Content-Based Image Retrieval

Edge-Based Structural Features for Content-Based Image Retrieval Appeared in Pattern Recognition Letters, April 2001 Edge-Based Structural Features for Content-Based Image Retrieval Xiang Sean Zhou, Thomas S. Huang Beckman Institute for Advanced Science and Technology

More information

UB at GeoCLEF Department of Geography Abstract

UB at GeoCLEF Department of Geography   Abstract UB at GeoCLEF 2006 Mguel E. Ruz (1), Stuart Shapro (2), June Abbas (1), Slva B. Southwck (1) and Davd Mark (3) State Unversty of New York at Buffalo (1) Department of Lbrary and Informaton Studes (2) Department

More information

A New Approach For the Ranking of Fuzzy Sets With Different Heights

A New Approach For the Ranking of Fuzzy Sets With Different Heights New pproach For the ankng of Fuzzy Sets Wth Dfferent Heghts Pushpnder Sngh School of Mathematcs Computer pplcatons Thapar Unversty, Patala-7 00 Inda pushpndersnl@gmalcom STCT ankng of fuzzy sets plays

More information