Steganalysis of DCT-Embedding Based Adaptive Steganography and YASS

Size: px
Start display at page:

Download "Steganalysis of DCT-Embedding Based Adaptive Steganography and YASS"

Transcription

1 Steganalyss of DCT-Embeddng Based Adaptve Steganography and YASS Qngzhong Lu Department of Computer Sene Sam Houston State Unversty Huntsvlle, TX 77341, U.S.A. ABSTRACT Reently well-desgned adaptve steganograph systems, nludng ±1 embeddng n the DCT doman wth optmzed osts to aheve the mnmal-dstorton [8], have posed serous hallenges to steganalyzers. Addtonally, although the steganalyss of Yet Another Steganograph Sheme (YASS) was atvely onduted, the deteton of the YASS steganograms by a large B-blok parameter has not been well explored. In ths paper, we am to detet the state-of-the-art adaptve steganograph system n DCT-embeddng and to mprove the steganalyss of YASS. To detet DCT-embeddng based adaptve steganography, we desgn the features of dfferental neghborng jont densty on the absolute array of DCT oeffents between the orgnal JPEG mages and the albrated versons. To dsrmnate YASS steganograms from overs, the anddate bloks that are possbly used for embeddng and the nonanddate blok neghbors that are mpossbly used for nformaton hdng are dentfed frst. The dfferene of the neghborng jont densty between anddate bloks and the nonanddate neghbors s obtaned. Support Vetor Mahne (SVM) and logst regresson lassfers are employed for lassfaton. Expermental results show that our approah s very promsng when detetng DCT-embeddng based adaptve steganography. Compared to the steganalyss based on CC-PEV feature set, our method greatly mproves the deteton auray; the advantage s espeally noteable n the deteton of the steganograms wth low relatve payload. In steganalyss of YASS, our approah s superor to a prevous well-known steganalyss algorthm; our method remarkably mproves the deteton auray espeally n the deteton of the YASS steganograms that are produed wth a large B-blok sze, whh was not well addressed before. Categores and Subjet Desrptors I 4.9 [Image Proessng and Compute r Vson]: Applatons; K.6.m [Msellaneous]: Insurane and Seurty. General Terms Algorthms and Seurty Keywords Steganography, steganalyss, adaptve steganography, YASS, Permsson to make dgtal or hard opes of all or part of ths work for personal or lassroom use s granted wthout fee provded that opes are not made or dstrbuted for proft or ommeral advantage and that opes bear ths note and the full taton on the frst page. To opy otherwse, or republsh, to post on servers or to redstrbute to lsts, requres pror spef permsson and/or a fee. MM&Se 11, September 29 30, 2011, Buffalo, New York, USA. Copyrght 2011 ACM /11/09...$ DCT, JPEG, neghborng jont densty, logst regresson, SVM, CC-PEV. 1. INTRODUCTION Steganography ams to enable overt ommunaton by embeddng data nto dgtal fles and makng the hdden message nvsble. The potental of explotng steganography for overt dssemnaton s great: for example, a reent esponage ssue reveals that steganography has been used by governmental ntellgent ageny [1-3]. For several purposes, t s a heghtened need to realze effetve ountermeasures for steganography. In dgtal mages, to ths date, a few popular steganograph systems suh as LSB embeddng, LSB mathng [35, 41], spread spetrum steganography [34], Outguess [38], F5 [45], modelbased steganography [40], steghde [15], BCH syndrome ode based less detetable JPEG steganography [39], and reently welldesgned hghly undetetable steganography (HUGO) [37], have been suessfully steganalyzed [6, 10-14, 19, 20, 22, 24-26, 28, 33, 36, 42]. Although these remarkable advanes have been aheved, reently well-desgned steganograph systems, suh as Gbbs onstruton-based steganography [7], Syndrome-Trells Codes based steganography [9], have posed new hallenges for steganalyss. Fller and Frdrh reently presented a pratal framework of adaptve steganograph systems [8] by optmzng the parameters of addtve dstorton funtons and mnmzng the dstorton for ±1 embeddng n the DCT doman, whh greatly mproves the pror art of hdng data n wde-spread JPEG mages. The expermental results shown n [8] demonstrated DCTembeddng based adaptve system s undetetable, attaked by a prevous state-of-the-art steganalyss method, whle the relatve payload rato s less than 0.15 bpa. YASS was desgned to be a seure JPEG steganograph algorthm wth randomzed embeddng [43]. However, the loatons of the embeddng host bloks are not randomzed enough. The embeddng n YASS also ntrodues extra zero DCT oeffents nto the embeddng host bloks, and hene leaves a lue to be exposed L et al. presented a smple and effent deteton method by omparng the frequeny of zero oeffents of the embeddng host bloks and the neghborng bloks n DCT doman [23]. The deteton performane s very promsng whle the parameter of the bg blok (B-blok) sze s small (e.g., the sze s set to 9 and 10). However, the deteton performane apparently deterorates whle the parameter of B-blok sze nreases [23]. Reently, Kodovsky et al. desgned 1234 features to detet YASS and tested twelve dfferent onfguratons of YASS. In these twelve onfguratons, the parameter of B-blok sze s not larger than 11 [21]. However, the deteton. 77

2 performane on the YASS steganograms wth large parameter of B-blok sze (12, 13, 14, and 15) was mssng [21]. Amng to detet the state-of-the-art DCT-embeddng based adaptve steganography [8], we mprove a prevous JPEG steganalyss method [28]. We extrat the neghborng jont densty on the absolute array of DCT oeffents from the JPEG mage under examnaton, and we desgn a albrated algorthm to extrat the referene features; the dfferental features between the orgnal neghborng jont densty and the referene are alulated. Support Vetor Mahnes (SVM) [44] and logst regresson [16] lassfers are appled for lassfaton. Expermental results demonstrate promsng deteton performane of our approah. To mprove the deteton performane n steganalyss of YASS, we frst analyze the advantage and weakness of a prevous YASS deteton art presented n [23]. As ndated by L et al. [23], n YASS embeddng, the seleton of embeddng host blok s not random enough and the embeddng modfes the statsts of embeddng host bloks n DCT doman. However, ther deteton algorthm does not searh all possble anddate bloks that are possbly used for embeddng. Our study also fnds that the YASS embeddng not only nreases the zero oeffents of the host bloks, but t also modfes the neghborng jont densty of the DCT oeffents. Therefore, we desgn a new algorthm to mprove the deteton performane. The remander of the paper s organzed as follows: Seton 2 brefly ntrodues DCT-embeddng based adaptve steganography, YASS and a deteton art, and our prevous JPEG steganalyss method. Seton 3 presents our deteton method for steganalyss of DCT-embeddng based adaptve steganography, and seton 4 desrbes our approah to steganalyss of YASS. Seton 5 shows experments and analyss. Conlusons are made n seton BACKGROUND 2.1 DCT-embeddng based Adaptve Steganography Most steganograph systems am to mnmze the dstorton of orgnal over. In [8], a pratal framework for optmzng the parameters of addtve dstorton funtons to mnmze statstal detetablty was presented by defnng a rh parametr model. To realze DCT-embeddng based adaptve steganography, an nter/ntra-blok ost model was gven, as well as the performane of embeddng algorthms based on the nter/ntra-blok ost model when optmzed usng the L2-regularzed L2-loss (L2R_L2LOSS) rteron, attaked by a prevous state-of the art steganalyss method, whh was expermentally valdated as beng ompletely undetetable at low relatve payload [8]. In what follows, we brefly ntrodue ths pratal framework for DCTembeddng based adaptve steganography. Mnmal-dstorton steganography an be mplemented by mnmzng the followng ost funton where x, y n 1 D(x,y) x, y (1) s the ost hangng the th over pxel x to y. To desgn DCT-embeddng based adaptve steganography, an nter/ntra-blok ost model has been defned by Fller and Frdrh [8]. Let, r a be the model parameters desrbng the ost of dsturbng nter- and ntra-blok dependenes wth r = ( r,,, r,, r, ) and a =( a,,, a,, a, ). The ost of hangng any AC DCT oeffents x j to y j s gven by za j xj 1, xj, xj 1 0 f y xj (2) j (x, y) ( y) f yj otherwse 2 2 a, xj z z r, xj z r Where, N a and N r are ntra- and nter-blok neghborhoods. Based on the nter/ntra-blok ost model, whle the embeddng algorthms are optmzed by usng the mult-layered Syndrome- Trells Codes [9] to mnmze the L2R_L2LOSS rteron [8], wth SVM and CC-PEV feature set [20], and Cross-Doman Feature set [21, 8], respetvely, the experments show that proposed DCT-embeddng based adaptve steganography has greatly mproved the state of DCT-embeddng based steganography. More tehnal detals may be referred to [8]. 2.2 YASS and A Prevous Deteton Algorthm The orgnal YASS algorthm, presented n [43], nludes the followng steps: 1) Repeat-Aumulate error orreton ode s used to enode the payload; 2) The over mage s dvded nto bg bloks of T T (T =9, 10,, 15), denoted by B-blok; 3) In eah B-blok, an 8 8 blok s randomly seleted for payload embeddng; 4) The embeddng nludes the followng operatons: a) Seleted 8 8 blok s transformed usng a two-dmensonal DCT; b) The DCT oeffents are dvded by a quantzaton table, orrespondng to the hdng qualty fator QF h ; ) By usng QIM, bnary hdden bts are embedded nto the 19 low-frequeny AC DCT oeffents whose values are non-zeros; d) The modfed 8 8 blok s transformed bak to spatal doman; 5) The modfed mage s enoded n JPEG format wth the advertsng qualty fator QF a. Although YASS embeddng s not onfned to the 8 8 blok of the fnal JPEG ompresson n above step 5), the loaton of embeddng blok n B-blok s not random enough. Meanwhle, the QIM-based embeddng ntrodues addtonal zero DCT oeffents n the modfed 8 8 blok, and hene, L et al. desgned the followng algorthm to break YASS [23]. L et al. feature extraton algorthm for YASS deteton [23] Transform a JPEG mage under examnaton to spatal doman, denoted by I 1 ; For T = 9 to 15 For s = 1 to T (a) Dvde I s nto non-overlappng onseutve T T B- bloks; (b) Collet 8 8 bloks from the upper left of all B-bloks and perform 2D DCT; 78

3 () Quantze the DCT oeffents by usng QF a ; (d) Compute the probablty of zero rounded re-quantzed DCT oeffents n anddate embeddng bands and denote t by Z T (s); (e) Crop the frst s olumns and the frst s rows of I 1 to generate a new mage I s+1 for the next nner-loop; End 1 7 Compute the values of T Z () 1 T and T 7 1 T Z ( ) j T 6 T j as features. 7 End As shown by the above algorthm, the features are extrated from the anddate bloks along the dagonal dreton of B- bloks, not from all possble 8 8 anddate bloks n B-bloks. In a T T B-blok, there are (T-7) (T-7) blok anddates for embeddng. Unfortunately, the above algorthm only selets the (T-7) bloks along dagonal dreton, not all (T-7) (T-7) anddate bloks. As a result, the hane of the anddates along dagonal dreton only hts 1/(T-7). Whle the value of T s large, the ht rato s pretty low. For nstane, T=15, the ht rato s only 1/8 = The expermental results n the referene [23] also demonstrate that the deteton auray s not so good wth a large T value. 2.3 Neghborng Jont Densty based JPEG Steganalyss We have shown that nformaton-hdng n DCT doman generally modfes the neghborng jont densty [26, 28]. Aordngly, a JPEG-based steganalyss method based on neghborng jont densty was proposed, whh has been valdated to outperform Markov-proess based steganalyss that was orgnally presented n [6, 42]. Here we brefly ntrodue our pror deteton method. Our prevous study shows that ertan manpulatons suh as JPEG-based double ompresson, nformaton hdng, and resamplng, modfy the neghborng jont densty and leave a lue to reveal the operatons [26-29]. In general, neghborng jont densty of DCT oeffents s symmetr about the orgn. We desgned the neghborng jont densty features on the absolute array of DCT oeffents, desrbed as follows Neghborng Jont Densty on Intra-blok Let F denote the quantzed DCT oeffent array onsstng of M N bloks F j ( = 1, 2,, M; j = 1, 2,, N). The ntra-blok neghborng jont densty matrx on horzontal dreton absnj 1h and the matrx on vertal dreton absnj 1v are gven by: absnj ( x, y) 1v 1h absnj ( x, y) M N j1 m1 n1 M N j1 m1 n1 ( x, y) jmn 56MN jm( n1) ( x, y) jmn j( m1) n 56MN Where jmn s the DCT oeffent loated at the m th row and the n th olumn n the blok F j ; = 1 f ts arguments are satsfed, otherwse = 0; x and y are ntegers. For omputatonal effeny, we defne absnj 1 as the neghborng jont densty features on ntra-blok, alulated as follows: (3) (4) absnj ( x, y) absnj ( x, y) absnj ( x, y) /2 (5) 1 1h 1v In our pror deteton, the values of x and y are n the range [0, 5], and absnj 1 onssts of 36 features Neghborng Jont Densty on Inter-blok The nter-blok neghborng jont densty matrx on horzontal dreton absnj 2h and the matrx on vertal dreton absnj 2v are onstruted as follows: absnj 2h 2v ( x, y) absnj ( x, y) 8 8 M N1 m1 n1 1 j1 8 8 M1 N m1 n1 1 j1 ( x, y) jmn ( j1) mn 64 M( N 1) ( x, y) jmn ( 1) jmn 64( M 1) N We defne absnj 2 as the neghborng jont densty features on nter-blok, alulated as follows: absnj ( x, y) absnj ( x, y) absnj ( x, y) /2 (8) 2 2h 2v Smlarly, the values of x and y are n [0, 5] and absnj 2 has 36 features. 3. CALIBRATED NEIGHBORING JOINT DENSITY APPROACH TO STEGANALYSIS OF JPEG-BASED ADAPTIVE STEGANOGRAPHY Although DCT-embeddng based adaptve steganography ams to mnmze the dstorton ost through Syndrome-Trells Codes, we fnd that t does modfy the neghborng jont densty features proposed n referene [28], shown by Fgure 1. (a) JPEG over () Dfferene of NJ densty (b) JPEG steganogram (6) (7) (d) Dfferene of absnj densty Fgure 1. An example to demonstrate the modfaton of neghborng jont densty features by DCT-embeddng based adaptve steganography. 79

4 Fgure 1(a) and (b) show a JPEG over and the JPEG steganogram produed by usng DCT-embeddng based adaptve hdng algorthm [8] wth the hdng rato 0.4 bts per non-zero- AC (bpa). The over mage s downloaded from [4]. The adaptve hdng tool s avalable at [5]. Fgure 1() shows the dfferene of the ntra-blok based neghborng jont densty extrated from (a) and (b). Fgure 1(d) shows the dfferene of the neghborng jont densty on the absolute array of DCT oeffents, defned by equaton (5). Although the modfaton s small, the nformaton-hdng does modfy the neghborng jont densty. Consderng that the jont densty vares aross dfferent dgtal mages, to reflet the modfaton of the densty aused by the embeddng, based on a self-albraton approah that was presented n [11], we desgn a albrated neghborng jont densty, desrbed as follows: 1. The neghborng jont densty features absnj 1 (x,y) and absnj 2 (x,y), defned by equatons (5) and (8), are extrated from a JPEG mage under examnaton; 2. The testng JPEG mage s deoded n spatal doman, and ropped by rows and j olumns (0 <7, 0 j<7, and +j >0). The ropped mage s enoded n JPEG format wth the same quantzaton matrx, and the jont densty features, denoted by absnj 1,j (x,y) and absnj 2,j (x,y), are extrated from the ropped JPEG mages, here (, j) 0,1, 0,2,..., 1,0, 1,1,..., 7,7 ; 3. The mean values of absnj 1 and absnj 2 are alulated by 1 absnj 1 ( x, y) absnj1, j( x, y) 63 (9) (, j) 1 absnj 2 ( x, y) absnj2, j( x, y) 63 (10) (, j) 4. The dfferental jont densty features are gven by D 1 (, ) 1(, ) 1 (, ) (11) D 2 (, ) 2(, ) 2 (, ) (12) absnj x y absnj x y absnj x y absnj x y absnj x y absnj x y 5. The dfferental rato features are obtaned by D absnj (, ) (, ) (, ) (13) D absnj (, ) (, ) (, ) (14) R x y absnj x y absnj x y R x y absnj x y absnj x y The rato features are defned n equatons (13) and (14), denoted by dff-absnj-rato, and the features defned by equatons (9) to (12), denoted by ref-dff-absnj, are used to detet DCTembeddng based adaptve steganography. In our study to detet adaptve steganography, the nteger parameters x and y are set from 0 to 5, produng 36 features n (13) and 36 features n (14), so dff-absnj-rato ontans 72 features, and ref-dff-absnj ontans 144 features. 4. NEIGHBORING JOINT DESNITY BASED YASS-DETECTION ALGORITHM By searhng all possble 8 8 anddate bloks n B-bloks, we extrat the neghborng jont densty of the DCT oeffents from all anddate bloks and the 8 8 blok neghbors that mpossbly belong to the anddate set for nformaton hdng, and alulate the dfferene of the jont densty values of the anddates and the non-anddate neghbors. Our algorthm of feature desgn to detet YASS steganogram s desrbed as follows: 1. Deode an nput JPEG mage under srutny to spatal doman, and dvde t nto non-overlappng onseutve T T B-bloks (T = 9, 10,, 15); 2. In eah T T B-blok, searh all 8 8 bloks possbly used for nformaton hdng, total (T-7) 2 anddate bloks. The set of all anddate bloks of the mage under deteton s denoted by C. For eah anddate blok C() (=1,2,, n), subtrat 128 from eah pxel value, then apply two-dmensonal DCT transform, quantze the DCT oeffents by usng the quantzaton matrx orrespondng to QF a and obtan the absolute DCT oeffent array. The neghborng jont densty features, defned by equaton (5), are extrated from the absolute DCT oeffent array, denoted by absnj(; x,y). 3. From all adjaent 8 8 bloks to the anddate blok C() n the horzontal or vertal dreton but wthout any overlappng to C(), denoted by N(), we dentfy the adjaent 8 8 bloks that do not belong to C, the set of anddate bloks for YASS embeddng. The non-anddate blok neghbors are denoted by NC(). The neghborng jont densty defned by equaton (5) are extrated from these non-anddate neghborng bloks, and the average neghborng jont densty s denoted by avg_nc_absnj(; x, y), the dfferene of the jont densty from the anddate blok C() and the average neghborng jont densty s gven by dff_absnj(; x, y) = absnj(; x, y)-avg_nc_absnj(; x, y) (15) 4. The total number of anddate bloks s n. The mean values of the dfferental jont densty, whh are the features for YASS steganalyss n our algorthm, are gven by the followng dff_absnj(x,y) = dff_ absnj ; x, y n (16) It should be noted that n a T T B-blok, whh s not on the boundary of the mage under examnaton, f an 8 8 blok anddate s loated (a) nsde of the B-blok, t has four non-anddate neghbors, shown by Fgure 2(a); (b) on one of the four boundary borders of the B-blok but not on any orner, t has three non-anddate neghbors, shown by Fgure 2(b); () on one of the four orners of the B-blok, t has two nonanddate neghbors, shown by Fgure 2(); Fgures 2 (a), (b), and () llustrate the above senaros. The square n dash stands for a B-blok, a omplete blok n the B- blok represents a anddate blok for possble hdng and the non-anddate blok neghbors are aross the square. In our YASS deteton, the values of x and y are set n [0, 2] and dff_absnj ontans 9 features, orrespondng to eah value of T. We expet that the dff_absnj features extrated from overs are approxmately zero-valued, but the values of the features from YASS steganograms are not onstraned to zeros. Fgure 3 valdates our onjeture and mples the effetveness of our proposed features. (a) (b) () Fgure 2. A anddate blok s loated n a B-blok (dashed) and the non-anddate neghbors are aross two B-bloks. 80

5 (a) YASS steganogram (T=9) () YASS steganogram (T=10) (b) dff_absnj features of the over and the left YASS stego-mage (T = 9) (d) dff_absnj features of the over and the left YASS stego-mage (T = 10) steganograms. A logst regresson lassfer [16, 32] and Support Vetor Mahnes (SVM) [42], are used for the deteton. In eah experment, 50% samples are randomly seleted for tranng, and the other 50% samples are used for testng. In eah experment, the testng results an be dvded nto True Negatve (TN), False Negatve (FN), False Postve (FP), and True Postve (TP). Wthout losng a generalty, we measure the deteton auray by 0.5*TN/(TN+FP)+0.5*TP/(TP+FN). To ompare the deteton performane, two hundred experments are operated for eah feature set at eah hdng rato by usng eah lassfer, and the mean deteton auray over 200 experments s obtaned. In the applaton of SVM, we partularly adopt two popular SVM algorthms, LbSVM [46] and SVM_lght [17], and we ompare the deteton performane of these two SVM mplementaton algorthms wth lnear, polynomal, and radal bass funton (RBF) kernels. On average, n our experments, a lnear LbSVM hts the hghest deteton auray Expermental Results Table 1 lsts the mean values of deteton auray on testng feature sets over two hundred experments by usng the 72- dmensonal dff-absnj-rato feature set, 144-dmensonal ref-dffabsnj feature set, and 548-dmensonal CC-PEV feature set wth lnear LbSVM and logst regresson (denoted by LogtReg) lassfer. The expermental results show that the dff-absnj-rato and ref-dff-absnj feature sets outperform CC-PEV feature set regardng deteton auray. Espeally at the relatve payload parameter of 0.1 bpa and 0.15 bpa, dff-absnj-rato and ref-dffabsnj feature sets mprove the deteton auray by about 15~20%, ether usng SVM or usng logst regresson lassfer. Table 1. Average deteton auray (%) over 200 experments at dfferent hdng ratos (measured by relatve payload, bpa), by applyng SVM and logst regresson lassfer to 548-dm CC- PEV, 72-dm dff-absnj-rato, and 144-dm ref-dff-absnj. (e) YASS steganogram (T=15) (f) dff_absnj features of the over and the left YASS stego-mage (T = 15) Fgure 3. Modfaton of the dff-absnj features by YASS embeddng (QF h = QF a = 75) wth B-blok sze T=9, 10, and 15, wheren the feature ndes from 1, 2,, to 9 orrespond to the (x, y) pars n equaton (16) from (0, 0), (0, 1),, to (2,2). 5. EXPERIMENTS 5.1 Steganalyss of DCT-Embeddng based Adaptve Steganography Setup 1000 BOSSRank over mages downloaded from [4] are frst onverted nto JPEG mages wth the qualty fator 75. The JPEG-based adaptve steganograms are produed by usng the DCT-embeddng based hdng tool [5], and the parameter of hdng bts per non-zero-ac (bpa) s set from 0.1 to 0.4 wth the step of 0.05 bpa. We extrat 72-dmensonal rato features, defned by (13) and (14), or dff-absnj-rato, and 144- dmensonal features, or ref-dff-absnj, from the JPEG overs and the adaptve steganograms. To ompare our feature sets and a reently well-desgned feature set, CC-PEV [20, 36], we also extrat the 548-dmensonal CC-PEV features from the overs and bpa CC-PEV dff-absnj-rato ref-dff-absnj SVM LogtReg SVM Logt Reg SVM LogtReg Addtonally, the mean and the standard devaton (STD) values of true negatve rate (TNR) and true postve rate (TPR) over 200 experments are gven by Fgure 4. Whle we onsder the mean deteton auray wth the standard devaton together, n deteton of the steganograms at low relatve payload, ether usng SVM or logst regresson lassfer, the deteton auray wth CC-PEV s not mpressve, but the orrespondng standard devaton are pretty hgh, whh means that the deteton performane by usng CC-PEV s very unstable aross dfferent experments. It also lams the undetetabllty of DCTembeddng based adaptve steganography aganst CC-PEV feature set. In omparson to CC-PEV, our feature sets demonstrate the superorty, ether n terms of deteton auray or the deteton stablty aross dfferent experments. 81

6 (a) Mean TNR wth LbSVM (b) Mean TPR wth LbSVM () Mean TNR wth LogtReg (d) Mean TPR wth LogtReg (e) Std(TNR) wth LbSVM (f) Std(TPR) wth LbSVM (g) Std(TNR) wth LogtReg (h) Std(TPR) wth LogtReg Fgure 4. The mean and standard devaton of true negatve rate (TNR) and true postve rate (TPR) by applyng LbSVM and logst regresson to dff-absnj-rato, ref-dff-absnj, and CC-PEV feature sets. 5.2 Steganalyss of YASS Setup Smlarly, the orgnal 1000 BOSSRank over mages downloaded from [4] are used for YASS embeddng. We set QF h = QF a = 75 and QF h = QF a = 50 respetvely. Aordngly, we enode the 1000 BOSSRank over mages n JPEG format at the qualty fator of 75 and 50 respetvely, whh are used as JPEG overs. In reaton of YASS steganograms, QF h and QF a are set to the same quantzaton fator n order to avod double JPEG ompresson, beause the YASS steganograms ould be deteted by usng the deteton method to expose double JPEG ompresson. In our experments, the embeddng parameter T of B-blok sze s set from 9 to 15. To ondut a omparatve study, we extrat the dff_absnj features defned n (16), and the zero-valued probablty features presented by L et al. [23]. A lnear LbSVM and logst regresson lassfer are used for lassfaton (smlar to the steganalyss of DCT-embeddng based adaptve steganography, we ompared LbSVM and SVM_lght wth lnear, polynomal and RBF kernels and fnally seleted LbSVM wth lnear kernel n these experments). In eah experment, 50% samples are randomly seleted for tranng, and the other 50% samples are used for testng; 200 experments are operated for eah feature set at eah B-blok sze by usng eah learnng lassfer Expermental Results The mean value and standard devaton of the deteton auray on testng feature sets over 200 experments are lsted n Tables 2 and 3. The deteton auray on testng set s alulated by the half of the sum of true postve rate and true negatve rate, or 0.5*TP/(TP+FN)+0.5*TN/(TN+FN). Expermental results show that the deteton method presented n [23] delvers good performane n deteton of the YASS steganograms that are produed wth small B-blok sze (e.g., T = 9, the deteton auray s over 99%). However, the deteton performane apparently deterorates whle the parameter of B- blok sze nreases (e.g., T = 15, ether usng SVM or logst regresson lassfer, the deteton auray s less than 70%). As a omparson, our method performs well ether n small or large parameter of B-blok sze, usng SVM or logst regresson lassfer. Whle T=9, 10, and 11 n the ase QF h = QF a = 75, and T=9, 10, 11, 12, and 13, n the ase QF h = QF a = 50, all deteton auray values obtaned by our approah are over 99%; whle T= 15, our approah hts the deteton auray over 90% for QF h = QF a = 75, and 95% for QF h = QF a = 50, ether usng SVM or logst regresson lassfer. The standard devaton values by usng the dff_absnj feature set are smaller than the values obtaned by ompared method, whh mples that our method s more stable than the ompared method. In summary, the expermental results shown by tables 2 and 3 ndate that our method s more effetve and relable ompared to the pror deteton art based on zero-valued DCT densty feature set. Table 2. The average deteton auray ± standard devaton (%) over 200 experments wth dff_absnj and zero-valued DCT densty feature set [23], usng LbSVM and LogtReg for the deteton of YASS steganograms (QF h = QF a = 75) T Dff_absNJ Zero-valued DCT densty SVM LogtReg SVM LogtReg ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±1.2 82

7 Table 3. The average deteton auray ± standard devaton (%) over 200 experments wth dff_absnj and zero-valued DCT densty feature set [23], usng LbSVM and LogtReg for the deteton of YASS steganograms (QF h = QF a = 50) T Dff_absNJ Zero-valued DCT densty SVM LogtReg SVM LogtReg ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Dsusson In steganalyss of DCT-embeddng based adaptve steganography, to obtan the albrated neghborng jont densty features, the JPEG mage under srutny s ropped 63 tmes ndvdually wth the shftng from (0, 1) to (7,7), the neghborng jont densty features are extrated from these 63 ropped versons, and the mean values of the features are used as albrated features. Compared to the albraton that only takes one-roppng (e.g., only shftng by 4 rows and 4 olumns), the omputaton ost s relatvely hgh. However, the albrated neghborng jont densty obtaned by 63-roppng s generally loser to the neghborng jont densty of orgnal over, shown by Fgure 5, wheren the mean absolute values of the dfferene of the neghborng jont densty between 1000 overs and the albrated versons are gven. Fgure 6 plots the mean values of the relatve dfferene on the 1000 overs and the albrated versons. Relatve dfferene s absnj x, y absnj x, y absnj x, y alulated by wheren absnj x,y and absnj x, y stand for the neghborng jont densty from un-albrated mage and from the albrated verson respetvely. Beause DCT-embeddng based adaptve steganography has been well desgned to reman orgnal statstal property through Syndrome-Trells Codes and mnmze the dstorton ost, the dfferenes of the features from a over and from the steganogram are very small; n suh ase, f the albrated features are loser to those from orgnal over, t s better to mprove the deteton auray. Our experments show that f we only take one-roppng (ropped by 4 rows and 4 olumns) to obtan albrated features, ompared to the deteton by 63-tmes-roppng, the deteton auray dereases by about 6% n detetng the steganograms at relatve payload 0.1 bpa. It s worth notng that 63-tmes-roppng s not only useful to produe albrated features, but also very effetve to expose msalgned-roppng and reompresson-based forgery n JPEG mages. We have desgned shft-reompresson-based forgery deteton, whh s very promsng to reveal the relevant forgery manpulatons n JPEG mages [31]. Fgure 5. A omparson of the dfferene of neghborng jont densty between one-roppng and 63-tmes-roppng. Fgure 6. A omparson of the relatve dfferene of neghborng jont densty between one-roppng and 63-tmes-roppng. 83

8 In steganalyss of YASS, although L et al ndated that the weakness of the YASS steganograph system, ther deteton algorthm does not searh all anddate host bloks that are possbly used for nformaton hdng, and the deteton performane s not so well whle the B-blok sze s large. By searhng all possble anddate bloks and omparng the neghborng jont densty of these anddate bloks and the nonanddate neghborng bloks, we have greatly mproved the deteton performane. We should menton that n YASS embeddng, f the embeddng postons of bnary hdden bts are not lmted nto the 19 low-frequeny AC DCT oeffents (e.g., the AC DCT oeffents are randomly seleted for embeddng), our approah s stll effetve for the deteton, beause our feature extraton s not lmted to the poston of 19 lowfrequeny AC oeffents. In the orgnal YASS embeddng algorthm, the upper-left of the frst B-blok s overlapped wth the upper-left of the frst 8 8 blok. If the frst B-blok randomly s msplaed from the upperleft pont of the frst 8 8 blok, we an searh all possblty of msmathng. There are 64 ombnatons nludng the orgnal exat mathng, aordngly we an retreve the dff_absnj features n eah msmathng, n order to detet suh polymorphsm of YASS steganograph system. In addton to SVM and logst regresson lassfer [32], other learnng lassfers, suh as evolvng neuro-fuzzy nferene system [18] and ensemble lassfer have been appled to steganalyss [12-14, 22, 24]. In terms of both deteton auray and omputaton ost, logst regresson s one good opton. The expermental results also demonstrate that the deteton auray under low ompresson qualty QF h = QF a = 50 s generally hgher than the deteton auray under hgh ompresson qualty QF h = QF a = 75. From our standpont, the low ompresson qualty fator takes large quantzaton steps durng JPEG ompresson to obtan quantzed DCT oeffents, and hene produes a smaller magntude of quantzed DCT oeffents. The hane of the modfaton to these small magntude quantzed DCT oeffents by YASS QIM embeddng aordngly nreases, and the amount of relatve modfaton nreases. As a result, the deteton auray on the YASS steganograms that are produed at low qualty s generally hgher than the results on the hgh qualty fator. To desgn undetetable steganography n JPEG mages, based on the relatonshp between mage omplexty and deteton performane [24, 25], and our pror study of JPEG steganalyss [28], we have desgned a JPEG-based statstally nvsble steganography by ±1 embeddng n the large magntude of quantzed DCT oeffents n the 8 8 bloks wth omplated texture, whh s smple and straghtforward [30]. By ombnng ths method wth the methodology of adaptve steganography [8], we surmse that the steganograph system of beng hghly undetetable an be desgned n DCT doman. 6. CONCLUSIONS In ths paper, we propose an mproved approah based on neghborng jont densty to detet a well-desgned adaptve steganography n DCT doman, whh has greatly mproved earler DCT-embeddng arts. We also propose a new approah to steganalyss of YASS, by omparng the neghborng jont densty of all anddate host bloks that are possbly used for data embeddng and the non-anddate neghborng bloks. Support vetor mahne and logst regresson lassfers are employed for lassfaton Experments show that, n steganalyss of DCT-embeddng based adaptve steganography, our approah has ganed onsderable good deteton performane ompared to a prevous state-of-the art JPEG steganalyss; the advantage of our approah s espeally noteable when detetng the steganograms at low payload embeddng. In steganalyss of YASS, our method has sgnfantly mproved a prevous deteton method, espeally n the deteton of YASS steganograms that are produed by adoptng a large B-blok sze, whh was not well addressed before. 7. ACKNOWLEDGMENTS Ths projet was supported by Award No DN-BX-K223 awarded by the Natonal Insttute of Juste, Offe of Juste Programs, U. S. Department of Juste. The opnons, fndngs, and onlusons or reommendatons expressed n ths publaton/program/exhbton are those of the authors and do not neessarly reflet those of the Department of Juste. Part of the support for ths study from a 2011 Sam Houston State Unversty (SHSU) Researh Enhanement grant s also greatly appreated. We are grateful to Dr. Bn L for provdng us ther feature extraton ode and to anonymous revewers for ther nsghtful omments and suggestons to mprove our work. In addton to BOSSRank mages [4], the hdng tool of DCT-embeddng based adaptve steganography [5], the ode to extrat CC-PEV features [47], MATLABArsenal [48] and PRtools [49] are used n ths study. We are truly grateful to these authors and provders. Speal thanks go to Mrs. Sharla Mles and Mrs. Dela Gallnaro at SHSU for ther proofreadng. 8. REFERENCES [1] _sene-sene/ [2] [3] [4] W&tmpl=materals [5] [6] Chen C and Sh Y (2008). JPEG mage steganalyss utlzng both ntrablok and nterblok orrelatons. Pro IEEE Internatonal Symposum on Cruts and Systems, pp [7] Fller T and Frdrh J (2010). Gbbs onstruton n steganography. IEEE Trans. on Info. Forenss and Seurty, 5(4): [8] Fller T and Frdrh J (2011). Desgn of adaptve steganograph shemes for dgtal mages. Pro. SPIE, Eletron Imagng, Meda Watermarkng, Seurty, and Forenss XIII, San Franso, CA, January 23-26, [9] Fller T, Judas J and Frdrh J (2011). Mnmzng addtve dstorton n steganography usng syndrome-trells odes. IEEE Trans. on Info. Forenss and Seurty, to appear. [10] Frdrh J, Goljan M and Hogeam D (2002). Steganalyss of JPEG mages: breakng the F5 algorthm. Pro. of 5 th Informaton Hdng Workshop, pp [11] Frdrh J (2004). Feature-based steganalyss for JPEG mages and ts mplatons for future desgn of steganograph shemes. LNCS, 3200, pp

9 [12] Frdrh J, Kodovsky J, Holub V and Goljan M (2011). Breakng HUGO the proess dsovery. In Pro. 13th Informaton Hdng Workshop, Prague, Czeh Republ, May 18 20, [13] Frdrh J, Kodovsky J, Holub V and Goljan M (2011). Steganalyss of ontent-adaptve steganography n spatal doman. In Pro. 13th Informaton Hdng Workshop, Prague, Czeh Republ, May 18 20, [14] Gul G and Kurugollu F (2011). A new methodology n steganalyss: breakng hghly undetetable steganography (HUGO). In Pro. 13th Informaton Hdng Workshop, Prague, Czeh Republ, May 18 20, [15] Hetzl S and Mutzel P (2005). A graph-theoret approah to steganography. LNCS, 3677, pp [16] Hlbe, JM (2009). Logst Regresson Models. Chapman & Hall/CRC Press. ISBN [17] Joahms T (2000). Estmatng the generalzaton performane of a SVM effently. Pro. of the Internatonal Conferene on Mahne Learnng, Morgan Kaufman, SVM_lght s avalable at [18] Kasabov NK and Song Q (2002). DENFIS: dynam evolvng neural-fuzzy nferene system and ts applaton for tme-seres predton. IEEE Transatons on Fuzzy Systems, 10(2): [19] Ker A (2005). Improved deteton of LSB steganography n graysale mages. LNCS, 3200, pp [20] Kodovsky J and Frdrh J (2009). Calbraton revsted. Proeedngs of the 11th ACM Multmeda and Seurty Workshop, Prneton, NJ, September 7-8, [21] Kodovsky J, Pevny T and Frdrh J (2010). Modern steganalyss an detet YASS, Pro. SPIE, Eletron Imagng, Meda Forenss and Seurty XII, San Jose, CA, January 17 21, pp , [22] Kodovsky J and Frdrh J (2011). Steganalyss n hgh dmensons: fusng lassfers bult on random subspaes. Pro. SPIE 7880, 78800L (2011); do: / [23] L B, Sh Y and Huang J (2009). Steganalyss of YASS. IEEE Trans. Informaton Forenss and Seurty, 4(3): [24] Lu Q, Sung AH, Chen H and Xu J (2008). Feature mnng and pattern lassfaton for steganalyss of LSB mathng steganography n graysale mages. Pattern Reognton, 41(1): [25] Lu Q, Sung AH, Rbero BM, We M, Chen Z and Xu J (2008). Image omplexty and feature mnng for steganalyss of least sgnfant bt mathng steganography. Informaton Senes, 178(1): [26] Lu Q, Sung AH and Qao M (2009). Improved deteton and evaluaton for JPEG steganalyss. Pro. 17 th ACM Multmeda, pp [27] Lu Q and Sung AH (2009). A new approah for JPEG resze and mage splng deteton. Pro. ACM Multmeda Workshop on Multmeda n Forenss 2009, pp [28] Lu Q, Sung AH and Qao M (2011). Neghborng jont densty-based JPEG steganalyss. ACM Trans. on Intellgent Systems and Tehnology, 2(2): artle 16. [29] Lu Q, Sung AH and Qao M (2011). A method to detet JPEG-based double ompresson. In Pro. of 8 th Internatonal Symposum on Neural Networks (2), pages [30] Lu Q, Sung AH, Chen Z and Huang X (2011). A JPEGbased statstally nvsble steganography. In Pro of 3 rd ACM Internatonal Conferene on Internet Multmeda Computng and Serve, August 5-7, 2011, Chengdu, Chna. [31] Lu Q. Deteton of msalgned roppng and reompresson wth the same quantzaton matrx and relevant forgery, ACM Workshop on Multmeda n Forenss and Intellgene Under revew. [32] Lubenko I and Ker AD (2011). Steganalyss usng logst regresson. In Pro. SPIE 7880, 78800K (2011); do: / [33] Lyu S and Fard H (2005). How realst s photorealst. IEEE Trans. on Sgnal Proessng, 53(2): [34] Marvel L, Bonelet C and Retter C (1999). Spread spetrum mage steganography. IEEE Trans. Image Proessng, 8(8): [35] Melkanen J (2006). LSB mathng revsted. IEEE Sgnal Proessng Letters 13(5): [36] Pevny T and Frdrh J (2007). Mergng Markov and DCT features for mult-lass JPEG steganalyss. Pro. SPIE, Vol. 6505, , 2007; DOI: / [37] Pevny T, Fller T and Bas P (2010). Usng hgh-dmensonal mage models to perform hghly undetetable steganography, Pro. 12 th Informaton Hdng, Calgary, Alberta, Canada, June 28-30, [38] Provos N (2001). Defendng aganst statstal steganalyss. Pro. 10 th USENIX Seurty Symposum, vol.10, pp [39] Sahnev V, Km HJ and Zhang R (2009). Less detetable JPEG steganography method based on heurst optmzaton and BCH syndrome ode. In: Proeedngs of the 11 th ACM Multmeda & Seurty Workshop. Pp [40] Sallee P (2004). Model based steganography. LNCS, 2939, pp [41] Sharp T (2001). An mplementaton of key-based dgtal sgnal steganography. LNCS, 2137, pp.13 26, [42] Sh Y, Chen C and Chen W (2007). A Markov proess based approah to effetve attakng JPEG steganography. LNCS, 4437, pp [43] Solank K, Sarkar A and Manjunath B (2007). YASS: Yet another steganograph sheme that ressts blnd steganalyss. LNCS, 4567, pp [44] Vapnk V (1998), Statstal Learnng Theory, John Wley. [45] Westfeld A (2001). Hgh apaty despte better steganalyss (F5 a steganograph algorthm). LNCS 2137, pp [46] [47] [48] al/matlabarsenal.htm [49] 85

Color Texture Classification using Modified Local Binary Patterns based on Intensity and Color Information

Color Texture Classification using Modified Local Binary Patterns based on Intensity and Color Information Color Texture Classfaton usng Modfed Loal Bnary Patterns based on Intensty and Color Informaton Shvashankar S. Department of Computer Sene Karnatak Unversty, Dharwad-580003 Karnataka,Inda shvashankars@kud.a.n

More information

Progressive scan conversion based on edge-dependent interpolation using fuzzy logic

Progressive scan conversion based on edge-dependent interpolation using fuzzy logic Progressve san onverson based on edge-dependent nterpolaton usng fuzzy log P. Brox brox@mse.nm.es I. Baturone lum@mse.nm.es Insttuto de Mroeletróna de Sevlla, Centro Naonal de Mroeletróna Avda. Rena Meredes

More information

Distortion Function Designing for JPEG Steganography with Uncompressed Side-image

Distortion Function Designing for JPEG Steganography with Uncompressed Side-image Dstorton Functon Desgnng for JPEG Steganography wth Uncompressed Sde-mage Fangjun Huang School of Informaton Scence and Technology, Sun Yat-Sen Unversty, GD 56, Chna huangfj@mal.sysu.edu.cn Jwu Huang School

More information

Research on Neural Network Model Based on Subtraction Clustering and Its Applications

Research on Neural Network Model Based on Subtraction Clustering and Its Applications Avalable onlne at www.senedret.om Physs Proeda 5 (01 ) 164 1647 01 Internatonal Conferene on Sold State Deves and Materals Sene Researh on Neural Networ Model Based on Subtraton Clusterng and Its Applatons

More information

Performance Evaluation of TreeQ and LVQ Classifiers for Music Information Retrieval

Performance Evaluation of TreeQ and LVQ Classifiers for Music Information Retrieval Performane Evaluaton of TreeQ and LVQ Classfers for Mus Informaton Retreval Matna Charam, Ram Halloush, Sofa Tsekerdou Athens Informaton Tehnology (AIT) 0.8 km Markopoulo Ave. GR - 19002 Peana, Athens,

More information

LOCAL BINARY PATTERNS AND ITS VARIANTS FOR FACE RECOGNITION

LOCAL BINARY PATTERNS AND ITS VARIANTS FOR FACE RECOGNITION IEEE-Internatonal Conferene on Reent Trends n Informaton Tehnology, ICRTIT 211 MIT, Anna Unversty, Chenna. June 3-5, 211 LOCAL BINARY PATTERNS AND ITS VARIANTS FOR FACE RECOGNITION K.Meena #1, Dr.A.Suruland

More information

Link Graph Analysis for Adult Images Classification

Link Graph Analysis for Adult Images Classification Lnk Graph Analyss for Adult Images Classfaton Evgeny Khartonov Insttute of Physs and Tehnology, Yandex LLC 90, 6 Lev Tolstoy st., khartonov@yandex-team.ru Anton Slesarev Insttute of Physs and Tehnology,

More information

Keywords: watermarking, Security, Encode and Decode,Fragile watermarking,psnr ratio.

Keywords: watermarking, Security, Encode and Decode,Fragile watermarking,psnr ratio. Volume 5, Issue 9, September 2015 ISSN: 2277 128X Internatonal Journal of Advaned Researh n Computer Sene and Software Engneerng Researh Paper Avalable onlne at: www.jarsse.om Improved Reversble Watermarng

More information

TAR based shape features in unconstrained handwritten digit recognition

TAR based shape features in unconstrained handwritten digit recognition TAR based shape features n unonstraned handwrtten dgt reognton P. AHAMED AND YOUSEF AL-OHALI Department of Computer Sene Kng Saud Unversty P.O.B. 578, Ryadh 543 SAUDI ARABIA shamapervez@gmal.om, yousef@s.edu.sa

More information

Multilabel Classification with Meta-level Features

Multilabel Classification with Meta-level Features Multlabel Classfaton wth Meta-level Features Sddharth Gopal Carnege Mellon Unversty Pttsburgh PA 523 sgopal@andrew.mu.edu Ymng Yang Carnege Mellon Unversty Pttsburgh PA 523 ymng@s.mu.edu ABSTRACT Effetve

More information

Adaptive Class Preserving Representation for Image Classification

Adaptive Class Preserving Representation for Image Classification Adaptve Class Preservng Representaton for Image Classfaton Jan-Xun M,, Qankun Fu,, Wesheng L, Chongqng Key Laboratory of Computatonal Intellgene, Chongqng Unversty of Posts and eleommunatons, Chongqng,

More information

Boosting Weighted Linear Discriminant Analysis

Boosting Weighted Linear Discriminant Analysis . Okada et al. / Internatonal Journal of Advaned Statsts and I&C for Eonoms and Lfe Senes Boostng Weghted Lnear Dsrmnant Analyss azunor Okada, Arturo Flores 2, Marus George Lnguraru 3 Computer Sene Department,

More information

Gabor-Filtering-Based Completed Local Binary Patterns for Land-Use Scene Classification

Gabor-Filtering-Based Completed Local Binary Patterns for Land-Use Scene Classification Gabor-Flterng-Based Completed Loal Bnary Patterns for Land-Use Sene Classfaton Chen Chen 1, Lbng Zhou 2,*, Janzhong Guo 1,2, We L 3, Hongjun Su 4, Fangda Guo 5 1 Department of Eletral Engneerng, Unversty

More information

Matrix-Matrix Multiplication Using Systolic Array Architecture in Bluespec

Matrix-Matrix Multiplication Using Systolic Array Architecture in Bluespec Matrx-Matrx Multplaton Usng Systol Array Arhteture n Bluespe Team SegFault Chatanya Peddawad (EEB096), Aman Goel (EEB087), heera B (EEB090) Ot. 25, 205 Theoretal Bakground. Matrx-Matrx Multplaton on Hardware

More information

Avatar Face Recognition using Wavelet Transform and Hierarchical Multi-scale LBP

Avatar Face Recognition using Wavelet Transform and Hierarchical Multi-scale LBP 2011 10th Internatonal Conferene on Mahne Learnng and Applatons Avatar Fae Reognton usng Wavelet Transform and Herarhal Mult-sale LBP Abdallah A. Mohamed, Darryl D Souza, Naouel Bal and Roman V. Yampolsky

More information

Fuzzy Modeling for Multi-Label Text Classification Supported by Classification Algorithms

Fuzzy Modeling for Multi-Label Text Classification Supported by Classification Algorithms Journal of Computer Senes Orgnal Researh Paper Fuzzy Modelng for Mult-Label Text Classfaton Supported by Classfaton Algorthms 1 Beatrz Wlges, 2 Gustavo Mateus, 2 Slva Nassar, 2 Renato Cslagh and 3 Rogéro

More information

Session 4.2. Switching planning. Switching/Routing planning

Session 4.2. Switching planning. Switching/Routing planning ITU Semnar Warsaw Poland 6-0 Otober 2003 Sesson 4.2 Swthng/Routng plannng Network Plannng Strategy for evolvng Network Arhtetures Sesson 4.2- Swthng plannng Loaton problem : Optmal plaement of exhanges

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

TECHNICAL POINTS ABOUT ADAPTIVE STEGANOGRAPHY BY ORACLE (ASO) 161, rue Ada, 34095, Montpellier Cedex 05, France

TECHNICAL POINTS ABOUT ADAPTIVE STEGANOGRAPHY BY ORACLE (ASO) 161, rue Ada, 34095, Montpellier Cedex 05, France 20th European Sgnal Processng Conference (EUSIPCO 2012) Bucharest, Romana, August 27-31, 2012 TECHNICAL POINTS ABOUT ADAPTIVE STEGANOGRAPHY BY ORACLE (ASO) Sarra Kouder 2, Marc Chaumont 1,2, Wllam Puech

More information

Pattern Classification: An Improvement Using Combination of VQ and PCA Based Techniques

Pattern Classification: An Improvement Using Combination of VQ and PCA Based Techniques Ameran Journal of Appled Senes (0): 445-455, 005 ISSN 546-939 005 Sene Publatons Pattern Classfaton: An Improvement Usng Combnaton of VQ and PCA Based Tehnques Alok Sharma, Kuldp K. Palwal and Godfrey

More information

AVideoStabilizationMethodbasedonInterFrameImageMatchingScore

AVideoStabilizationMethodbasedonInterFrameImageMatchingScore Global Journal of Computer Sene and Tehnology: F Graphs & vson Volume 7 Issue Verson.0 Year 207 Type: Double Blnd Peer Revewed Internatonal Researh Journal Publsher: Global Journals In. (USA) Onlne ISSN:

More information

Improved Accurate Extrinsic Calibration Algorithm of Camera and Two-dimensional Laser Scanner

Improved Accurate Extrinsic Calibration Algorithm of Camera and Two-dimensional Laser Scanner JOURNAL OF MULTIMEDIA, VOL. 8, NO. 6, DECEMBER 013 777 Improved Aurate Extrns Calbraton Algorthm of Camera and Two-dmensonal Laser Sanner Janle Kong, Le Yan*, Jnhao Lu, Qngqng Huang, and Xaokang Dng College

More information

Bit-level Arithmetic Optimization for Carry-Save Additions

Bit-level Arithmetic Optimization for Carry-Save Additions Bt-leel Arthmet Optmzaton for Carry-Sae s Ke-Yong Khoo, Zhan Yu and Alan N. Wllson, Jr. Integrated Cruts and Systems Laboratory Unersty of Calforna, Los Angeles, CA 995 khoo, zhanyu, wllson @sl.ula.edu

More information

Connectivity in Fuzzy Soft graph and its Complement

Connectivity in Fuzzy Soft graph and its Complement IOSR Journal of Mathemats (IOSR-JM) e-issn: 2278-5728, p-issn: 2319-765X. Volume 1 Issue 5 Ver. IV (Sep. - Ot.2016), PP 95-99 www.osrjournals.org Connetvty n Fuzzy Soft graph and ts Complement Shashkala

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

Computing Cloud Cover Fraction in Satellite Images using Deep Extreme Learning Machine

Computing Cloud Cover Fraction in Satellite Images using Deep Extreme Learning Machine Computng Cloud Cover Fraton n Satellte Images usng Deep Extreme Learnng Mahne L-guo WENG, We-bn KONG, Mn XIA College of Informaton and Control, Nanjng Unversty of Informaton Sene & Tehnology, Nanjng Jangsu

More information

MULTIPLE OBJECT DETECTION AND TRACKING IN SONAR MOVIES USING AN IMPROVED TEMPORAL DIFFERENCING APPROACH AND TEXTURE ANALYSIS

MULTIPLE OBJECT DETECTION AND TRACKING IN SONAR MOVIES USING AN IMPROVED TEMPORAL DIFFERENCING APPROACH AND TEXTURE ANALYSIS U.P.B. S. Bull., Seres A, Vol. 74, Iss. 2, 2012 ISSN 1223-7027 MULTIPLE OBJECT DETECTION AND TRACKING IN SONAR MOVIES USING AN IMPROVED TEMPORAL DIFFERENCING APPROACH AND TEXTURE ANALYSIS Tudor BARBU 1

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

A Robust Algorithm for Text Detection in Color Images

A Robust Algorithm for Text Detection in Color Images A Robust Algorthm for Tet Deteton n Color Images Yangng LIU Satosh GOTO Takesh IKENAGA Abstrat Tet deteton n olor mages has beome an atve researh area sne reent deades. In ths paper we present a novel

More information

Edge Detection in Noisy Images Using the Support Vector Machines

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

More information

A Fast Way to Produce Optimal Fixed-Depth Decision Trees

A Fast Way to Produce Optimal Fixed-Depth Decision Trees A Fast Way to Produe Optmal Fxed-Depth Deson Trees Alreza Farhangfar, Russell Grener and Martn Znkevh Dept of Computng Sene Unversty of Alberta Edmonton, Alberta T6G 2E8 Canada {farhang, grener, maz}@s.ualberta.a

More information

A Modified Median Filter for the Removal of Impulse Noise Based on the Support Vector Machines

A Modified Median Filter for the Removal of Impulse Noise Based on the Support Vector Machines A Modfed Medan Flter for the Removal of Impulse Nose Based on the Support Vector Machnes H. GOMEZ-MORENO, S. MALDONADO-BASCON, F. LOPEZ-FERRERAS, M. UTRILLA- MANSO AND P. GIL-JIMENEZ Departamento de Teoría

More information

Bottom-Up Fuzzy Partitioning in Fuzzy Decision Trees

Bottom-Up Fuzzy Partitioning in Fuzzy Decision Trees Bottom-Up Fuzzy arttonng n Fuzzy eson Trees Maej Fajfer ept. of Mathemats and Computer Sene Unversty of Mssour St. Lous St. Lous, Mssour 63121 maejf@me.pl Cezary Z. Janow ept. of Mathemats and Computer

More information

Interval uncertain optimization of structures using Chebyshev meta-models

Interval uncertain optimization of structures using Chebyshev meta-models 0 th World Congress on Strutural and Multdsplnary Optmzaton May 9-24, 203, Orlando, Florda, USA Interval unertan optmzaton of strutures usng Chebyshev meta-models Jngla Wu, Zhen Luo, Nong Zhang (Tmes New

More information

Measurement and Calibration of High Accuracy Spherical Joints

Measurement and Calibration of High Accuracy Spherical Joints 1. Introduton easurement and Calbraton of Hgh Auray Spheral Jonts Ale Robertson, Adam Rzepnewsk, Alexander Sloum assahusetts Insttute of Tehnolog Cambrdge, A Hgh auray robot manpulators are requred for

More information

Optimal shape and location of piezoelectric materials for topology optimization of flextensional actuators

Optimal shape and location of piezoelectric materials for topology optimization of flextensional actuators Optmal shape and loaton of pezoeletr materals for topology optmzaton of flextensonal atuators ng L 1 Xueme Xn 2 Noboru Kkuh 1 Kazuhro Satou 1 1 Department of Mehanal Engneerng, Unversty of Mhgan, Ann Arbor,

More information

Deep learning is a good steganalysis tool when embedding key is reused for different images, even if there is a cover source-mismatch

Deep learning is a good steganalysis tool when embedding key is reused for different images, even if there is a cover source-mismatch Deep learnng s a good steganalyss tool when embeddng key s reused for dfferent mages, even f there s a cover source-msmatch Lonel PIBRE 2,3, Jérôme PASQUET 2,3, Dno IENCO 2,3, Marc CHAUMONT 1,2,3 (1) Unversty

More information

An Adaptive Filter Based on Wavelet Packet Decomposition in Motor Imagery Classification

An Adaptive Filter Based on Wavelet Packet Decomposition in Motor Imagery Classification An Adaptve Flter Based on Wavelet Paket Deomposton n Motor Imagery Classfaton J. Payat, R. Mt, T. Chusak, and N. Sugno Abstrat Bran-Computer Interfae (BCI) s a system that translates bran waves nto eletral

More information

A Novel Dynamic and Scalable Caching Algorithm of Proxy Server for Multimedia Objects

A Novel Dynamic and Scalable Caching Algorithm of Proxy Server for Multimedia Objects Journal of VLSI Sgnal Proessng 2007 * 2007 Sprnger Sene + Busness Meda, LLC. Manufatured n The Unted States. DOI: 10.1007/s11265-006-0024-7 A Novel Dynam and Salable Cahng Algorthm of Proxy Server for

More information

A MPAA-Based Iterative Clustering Algorithm Augmented by Nearest Neighbors Search for Time-Series Data Streams

A MPAA-Based Iterative Clustering Algorithm Augmented by Nearest Neighbors Search for Time-Series Data Streams A MPAA-Based Iteratve Clusterng Algorthm Augmented by Nearest Neghbors Searh for Tme-Seres Data Streams Jessa Ln 1, Mha Vlahos 1, Eamonn Keogh 1, Dmtros Gunopulos 1, Janwe Lu 2, Shouan Yu 2, and Jan Le

More information

Minimize Congestion for Random-Walks in Networks via Local Adaptive Congestion Control

Minimize Congestion for Random-Walks in Networks via Local Adaptive Congestion Control Journal of Communatons Vol. 11, No. 6, June 2016 Mnmze Congeston for Random-Walks n Networks va Loal Adaptve Congeston Control Yang Lu, Y Shen, and Le Dng College of Informaton Sene and Tehnology, Nanjng

More information

FULLY AUTOMATIC IMAGE-BASED REGISTRATION OF UNORGANIZED TLS DATA

FULLY AUTOMATIC IMAGE-BASED REGISTRATION OF UNORGANIZED TLS DATA FULLY AUTOMATIC IMAGE-BASED REGISTRATION OF UNORGANIZED TLS DATA Martn Wenmann, Bors Jutz Insttute of Photogrammetry and Remote Sensng, Karlsruhe Insttute of Tehnology (KIT) Kaserstr. 12, 76128 Karlsruhe,

More information

A Real-Time Detecting Algorithm for Tracking Community Structure of Dynamic Networks

A Real-Time Detecting Algorithm for Tracking Community Structure of Dynamic Networks A Real-Tme Detetng Algorthm for Trakng Communty Struture of Dynam Networks Jaxng Shang*, Lanhen Lu*, Feng Xe, Zhen Chen, Jaa Mao, Xueln Fang, Cheng Wu* Department of Automaton, Tsnghua Unversty, Beng,,

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

DETECTING AND ANALYZING CORROSION SPOTS ON THE HULL OF LARGE MARINE VESSELS USING COLORED 3D LIDAR POINT CLOUDS

DETECTING AND ANALYZING CORROSION SPOTS ON THE HULL OF LARGE MARINE VESSELS USING COLORED 3D LIDAR POINT CLOUDS ISPRS Annals of the Photogrammetry, Remote Sensng and Spatal Informaton Senes, Volume III-3, 2016 XXIII ISPRS Congress, 12 19 July 2016, Prague, Czeh Republ DETECTING AND ANALYZING CORROSION SPOTS ON THE

More information

Cluster ( Vehicle Example. Cluster analysis ( Terminology. Vehicle Clusters. Why cluster?

Cluster (  Vehicle Example. Cluster analysis (  Terminology. Vehicle Clusters. Why cluster? Why luster? referene funton R R Although R and R both somewhat orrelated wth the referene funton, they are unorrelated wth eah other Cluster (www.m-w.om) A number of smlar ndvduals that our together as

More information

Pixel-Based Texture Classification of Tissues in Computed Tomography

Pixel-Based Texture Classification of Tissues in Computed Tomography Pxel-Based Texture Classfaton of Tssues n Computed Tomography Ruhaneewan Susomboon, Danela Stan Rau, Jaob Furst Intellgent ultmeda Proessng Laboratory Shool of Computer Sene, Teleommunatons, and Informaton

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

A Flexible Solution for Modeling and Tracking Generic Dynamic 3D Environments*

A Flexible Solution for Modeling and Tracking Generic Dynamic 3D Environments* A Flexble Soluton for Modelng and Trang Gener Dynam 3D Envronments* Radu Danesu, Member, IEEE, and Sergu Nedevsh, Member, IEEE Abstrat The traff envronment s a dynam and omplex 3D sene, whh needs aurate

More information

Identify the Attack in Embedded Image with Steganalysis Detection Method by PSNR and RGB Intensity

Identify the Attack in Embedded Image with Steganalysis Detection Method by PSNR and RGB Intensity Internatonal Journal of Computer Systems (ISSN: 394-1065), Volume 03 Issue 07, July, 016 Avalable at http://www.jcsonlne.com/ Identfy the Attack n Embedded Image wth Steganalyss Detecton Method by PSNR

More information

Feature Reduction and Selection

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

More information

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

A Model-Based Approach for Automated Feature Extraction in Fundus Images

A Model-Based Approach for Automated Feature Extraction in Fundus Images A Model-Based Approah for Automated Feature Extraton n Fundus Images Huq L Shool of Computng Natonal Unversty of Sngapore dslhq@nus.edu.sg Opas Chutatape Shool of Eletral and Eletron Engneerng Nanyang

More information

Key-Selective Patchwork Method for Audio Watermarking

Key-Selective Patchwork Method for Audio Watermarking Internatonal Journal of Dgtal Content Technology and ts Applcatons Volume 4, Number 4, July 2010 Key-Selectve Patchwork Method for Audo Watermarkng 1 Ch-Man Pun, 2 Jng-Jng Jang 1, Frst and Correspondng

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

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

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

Microprocessors and Microsystems

Microprocessors and Microsystems Mroproessors and Mrosystems 36 (2012) 96 109 Contents lsts avalable at SeneDret Mroproessors and Mrosystems journal homepage: www.elsever.om/loate/mpro Hardware aelerator arhteture for smultaneous short-read

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

Multi-scale and Discriminative Part Detectors Based Features for Multi-label Image Classification

Multi-scale and Discriminative Part Detectors Based Features for Multi-label Image Classification Proeedngs of the wenty-seventh Internatonal Jont Conferene on Artfal Intellgene (IJCAI-8) Mult-sale and Dsrmnatve Part Detetors Based Features for Mult-lael Image Classfaton Gong Cheng, Deheng Gao, Yang

More information

arxiv: v3 [cs.cv] 31 Oct 2016

arxiv: v3 [cs.cv] 31 Oct 2016 Unversal Correspondene Network Chrstopher B. Choy Stanford Unversty hrshoy@a.stanford.edu JunYoung Gwak Stanford Unversty jgwak@a.stanford.edu Slvo Savarese Stanford Unversty sslvo@stanford.edu arxv:1606.03558v3

More information

Semi-analytic Evaluation of Quality of Service Parameters in Multihop Networks

Semi-analytic Evaluation of Quality of Service Parameters in Multihop Networks U J.T. (4): -4 (pr. 8) Sem-analyt Evaluaton of Qualty of Serve arameters n Multhop etworks Dobr tanassov Batovsk Faulty of Sene and Tehnology, ssumpton Unversty, Bangkok, Thaland bstrat

More information

Information Hiding Watermarking Detection Technique by PSNR and RGB Intensity

Information Hiding Watermarking Detection Technique by PSNR and RGB Intensity www..org 3 Informaton Hdng Watermarkng Detecton Technque by PSNR and RGB Intensty 1 Neha Chauhan, Akhlesh A. Waoo, 3 P. S. Patheja 1 Research Scholar, BIST, Bhopal, Inda.,3 Assstant Professor, BIST, Bhopal,

More information

FUZZY SEGMENTATION IN IMAGE PROCESSING

FUZZY SEGMENTATION IN IMAGE PROCESSING FUZZY SEGMENTATION IN IMAGE PROESSING uevas J. Er,, Zaldívar N. Danel,, Roas Raúl Free Unverstät Berln, Insttut für Inforat Tausstr. 9, D-495 Berln, Gerany. Tel. 0049-030-8385485, Fax. 0049-030-8387509

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

REGISTRATION OF TERRESTRIAL LASER SCANNER DATA USING IMAGERY INTRODUCTION

REGISTRATION OF TERRESTRIAL LASER SCANNER DATA USING IMAGERY INTRODUCTION EGISTATION OF TEESTIAL LASE SCANNE DATA USING IMAGEY Khall Al-Manasr, Ph.D student Clve S. Fraser, Professor Department of Geomats Unversty of Melbourne Vtora 3010 Australa k.al-manasr@pgrad.unmelb.edu.au.fraser@unmelb.edu.au

More information

High-Boost Mesh Filtering for 3-D Shape Enhancement

High-Boost Mesh Filtering for 3-D Shape Enhancement Hgh-Boost Mesh Flterng for 3-D Shape Enhancement Hrokazu Yagou Λ Alexander Belyaev y Damng We z Λ y z ; ; Shape Modelng Laboratory, Unversty of Azu, Azu-Wakamatsu 965-8580 Japan y Computer Graphcs Group,

More information

Performance Analysis of Hybrid (supervised and unsupervised) method for multiclass data set

Performance Analysis of Hybrid (supervised and unsupervised) method for multiclass data set IOSR Journal of Computer Engneerng (IOSR-JCE) e-issn: 2278-0661,p-ISSN: 2278-8727, Volume 16, Issue 4, Ver. III (Jul Aug. 2014), PP 93-99 www.osrjournals.org Performane Analyss of Hybrd (supervsed and

More information

Neighboring Joint Density-Based JPEG Steganalysis

Neighboring Joint Density-Based JPEG Steganalysis Neighboring Joint Density-Based JPEG Steganalysis QINGZHONG LIU, Sam Houston State University ANDREW H. SUNG and MENGYU QIAO, NewMexicoTech The threat posed by hackers, spies, terrorists, and criminals,

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

A Secured Method for Image Steganography Based On Pixel Values

A Secured Method for Image Steganography Based On Pixel Values A Secured Method for Image Steganography Based On Pxel Values Tarun Gulat #, Sanskrt Gupta * # Assocate Professor, Electroncs and Communcaton Engneerng Department, MMEC, M.M.U., Mullana, Ambala, Haryana,

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

Statistical Steganalyis of Images Using Open Source Software

Statistical Steganalyis of Images Using Open Source Software Statstcal Steganalys of Images Usng Open Source Software Bhargav Kapa, Stefan A. Robla Department of Computer Scence Montclar State Unversty Montclar, NJ 07043 roblas@mal.montclar.edu Abstract In ths paper

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

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

The Simulation of Electromagnetic Suspension System Based on the Finite Element Analysis

The Simulation of Electromagnetic Suspension System Based on the Finite Element Analysis 308 JOURNAL OF COMPUTERS, VOL. 8, NO., FEBRUARY 03 The Smulaton of Suspenson System Based on the Fnte Element Analyss Zhengfeng Mng Shool of Eletron & Mahanal Engneerng, Xdan Unversty, X an, Chna Emal:

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

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

Machine Learning 9. week

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

More information

Evaluation of Segmentation in Magnetic Resonance Images Using k-means and Fuzzy c-means Clustering Algorithms

Evaluation of Segmentation in Magnetic Resonance Images Using k-means and Fuzzy c-means Clustering Algorithms ELEKTROTEHIŠKI VESTIK 79(3): 129-134, 2011 EGLISH EDITIO Evaluaton of Segmentaton n Magnet Resonane Images Usng k-means and Fuzzy -Means Clusterng Algorthms Tomaž Fnkšt Unverza v Lublan, Fakulteta za stronštvo,

More information

Programming in Fortran 90 : 2017/2018

Programming in Fortran 90 : 2017/2018 Programmng n Fortran 90 : 2017/2018 Programmng n Fortran 90 : 2017/2018 Exercse 1 : Evaluaton of functon dependng on nput Wrte a program who evaluate the functon f (x,y) for any two user specfed values

More information

A Semi-parametric Approach for Analyzing Longitudinal Measurements with Non-ignorable Missingness Using Regression Spline

A Semi-parametric Approach for Analyzing Longitudinal Measurements with Non-ignorable Missingness Using Regression Spline Avalable at http://pvamu.edu/aam Appl. Appl. Math. ISSN: 93-9466 Vol., Issue (June 5), pp. 95 - Applatons and Appled Mathemats: An Internatonal Journal (AAM) A Sem-parametr Approah for Analyzng Longtudnal

More information

Detecting MP3Stego using Calibrated Side Information Features

Detecting MP3Stego using Calibrated Side Information Features 2628 JOURNAL OF SOFTWARE, VOL. 8, NO. 10, OCTOBER 2013 Detectng P3Stego usng Calbrated Sde Informaton Features Xanmn Yu School of Informaton Scence and Engneerng, Nngbo Unversty Emal: mlhappy1016@163.com

More information

A Hybrid Semi-Blind Gray Scale Image Watermarking Algorithm Based on DWT-SVD using Human Visual System Model

A Hybrid Semi-Blind Gray Scale Image Watermarking Algorithm Based on DWT-SVD using Human Visual System Model A Hybrd Sem-Blnd Gray Scale Image Watermarkng Algorthm Based on DWT-SVD usng Human Vsual System Model Rajesh Mehta r Scence & Engneerng, USICT Guru Gobnd Sngh Indrarprastha Unversty New Delh, Inda rajesh00ust@gmal.com

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

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

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 Lossless Watermarking Scheme for Halftone Image Authentication

A Lossless Watermarking Scheme for Halftone Image Authentication IJCSNS Internatonal Journal of Computer Scence and Network Securty, VOL.6 No.2B, February 2006 147 A Lossless Watermarkng Scheme for Halftone Image Authentcaton Jeng-Shyang Pan, Hao Luo, and Zhe-Mng Lu,

More information

High Payload Reversible Data Hiding Scheme Using Difference Segmentation and Histogram Shifting

High Payload Reversible Data Hiding Scheme Using Difference Segmentation and Histogram Shifting JOURNAL OF ELECTRONIC SCIENCE AND TECHNOLOGY, VOL. 11, NO. 1, MARCH 2013 9 Hgh Payload Reversble Data Hdng Scheme Usng Dfference Segmentaton and Hstogram Shftng Yung-Chen Chou and Huang-Chng L Abstract

More information

A GENETIC APPROACH FOR THE AUTOMATIC ADAPTATION OF SEGMENTATION PARAMETERS

A GENETIC APPROACH FOR THE AUTOMATIC ADAPTATION OF SEGMENTATION PARAMETERS A GENETIC APPROACH FOR THE AUTOMATIC ADAPTATION OF SEGMENTATION PARAMETERS R. Q. Fetosa a, *, G. A. O. P. Costa a, T. B. Cazes a, B. Fejo b a Dept. of Eletral Engneerng, b Dept of Informats, Cathol Unversty

More information

Solving two-person zero-sum game by Matlab

Solving two-person zero-sum game by Matlab Appled Mechancs and Materals Onlne: 2011-02-02 ISSN: 1662-7482, Vols. 50-51, pp 262-265 do:10.4028/www.scentfc.net/amm.50-51.262 2011 Trans Tech Publcatons, Swtzerland Solvng two-person zero-sum game by

More information

Wavefront Reconstructor

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

More information

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

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

Assignment # 2. Farrukh Jabeen Algorithms 510 Assignment #2 Due Date: June 15, 2009.

Assignment # 2. Farrukh Jabeen Algorithms 510 Assignment #2 Due Date: June 15, 2009. Farrukh Jabeen Algorthms 51 Assgnment #2 Due Date: June 15, 29. Assgnment # 2 Chapter 3 Dscrete Fourer Transforms Implement the FFT for the DFT. Descrbed n sectons 3.1 and 3.2. Delverables: 1. Concse descrpton

More information

Machine Learning. Support Vector Machines. (contains material adapted from talks by Constantin F. Aliferis & Ioannis Tsamardinos, and Martin Law)

Machine Learning. Support Vector Machines. (contains material adapted from talks by Constantin F. Aliferis & Ioannis Tsamardinos, and Martin Law) Machne Learnng Support Vector Machnes (contans materal adapted from talks by Constantn F. Alfers & Ioanns Tsamardnos, and Martn Law) Bryan Pardo, Machne Learnng: EECS 349 Fall 2014 Support Vector Machnes

More information

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

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

More information

The stream cipher MICKEY-128 (version 1) Algorithm specification issue 1.0

The stream cipher MICKEY-128 (version 1) Algorithm specification issue 1.0 The stream cpher MICKEY-128 (verson 1 Algorthm specfcaton ssue 1. Steve Babbage Vodafone Group R&D, Newbury, UK steve.babbage@vodafone.com Matthew Dodd Independent consultant matthew@mdodd.net www.mdodd.net

More information

Shape-adaptive DCT and Its Application in Region-based Image Coding

Shape-adaptive DCT and Its Application in Region-based Image Coding Internatonal Journal of Sgnal Processng, Image Processng and Pattern Recognton, pp.99-108 http://dx.do.org/10.14257/sp.2014.7.1.10 Shape-adaptve DCT and Its Applcaton n Regon-based Image Codng Yamn Zheng,

More information

CHAPTER 3 SEQUENTIAL MINIMAL OPTIMIZATION TRAINED SUPPORT VECTOR CLASSIFIER FOR CANCER PREDICTION

CHAPTER 3 SEQUENTIAL MINIMAL OPTIMIZATION TRAINED SUPPORT VECTOR CLASSIFIER FOR CANCER PREDICTION 48 CHAPTER 3 SEQUENTIAL MINIMAL OPTIMIZATION TRAINED SUPPORT VECTOR CLASSIFIER FOR CANCER PREDICTION 3.1 INTRODUCTION The raw mcroarray data s bascally an mage wth dfferent colors ndcatng hybrdzaton (Xue

More information