Pixel-Based Texture Classification of Tissues in Computed Tomography
|
|
- Rudolf Gibson
- 6 years ago
- Views:
Transcription
1 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 Systems DePaul Unversty, Chago, Illnos, rsusombo@students.depaul.edu, {drau, furst}@s.depaul.edu Abstrat Prevous researh has been done to lassfy dfferent tssues/organs of nterest present n medal mages, n partular n Computed Tomography (CT) mages. ost of the researh used the anatomal struture present n the mages n order to lassfy the tssues. In ths paper, nstead of usng the anatomal struture, we propose a pxel-based texture approah for the representaton and lassfaton of the regons of nterest. The approah norporates varous texture features and deson trees to aomplsh tssue lassfaton n normal Computed Tomography (CT) mages of the hest and abdomen. Frst, we ntrodue a new dreton vs. dsplaement pars (DDP) approah to alulate a o-ourrene matrx for apturng all possble ombnaton between dretons and dsplaements neessary n alulatng the texture features at the pxel-level. Seond, we evaluate varous dfferent neghborhood szes for the pxel-based texture representaton n order to fnd the optmal wndow sze for dfferentatng among 8 organs/tssues of nterest: aorta, fat, kdney, lver, lung, musle, spleen, and trabeular bone. For all organs/tssues (exept for aorta), the optmal wndow was 13-by-13 allowng the lassfaton senstvty metr to be at least 96% for all organs/tssues. For aorta, the optmal wndow sze was 9-by-9 wth the lassfaton senstvty beng 81%. 1 Introduton Automat analyss of mages from varous medal magng modaltes s n hgh demand n order to nrease the produtvty of radologsts when nterpretng and dagnosng hundreds of mages every day. In medal magng researh, texture nformaton plays an mportant role for several reasons. Frst, the organ shapes are dfferent aross dfferent sldes due to the generated sequenes of two-dmenson (D) mages from three-dmenson (3D) struture. Sne sequenes of D mages represent a ross seton of human body, the shape of an organ throughout the entre slde of 3D s not onsstent. oreover, some organs an have dfferent szes from one patent to another. Therefore, the shape-based dsrmnaton tehnques [7, 8] may not be approprate to dentfy the organs. To avod these lmtatons, low-level features based on the gray-level nformaton present n the mage are used to apture the ontent of the medal mages. However, the gray -levels alone are not suffent as many soft tssues have overlappng gray level ranges [10]. Therefore, low-level features based on texture nformaton, that s
2 expeted to be homogenous and onsstent aross multple sldes for the same organ, are mostly used to perform automat mage analyss n the medal magng feld. The goal of the approah presented n ths paper s to aheve automat lassfaton of tssues of nterest n normal CT mages of the hest and of the abdomen. Our approah, dreton vs. dsplaement pars (DDP), s based on texture features alulated at the pxellevel nstead of a global-level as t has been done before n the texture analyss lterature: for eah pxel, we alulate a o-ourrene matrx for apturng all possble ombnatons between dretons and dsplaements wthn a small neghborhood entered on that pxel. The optmal neghborhood sze s determned by the wndow sze produng the best pxel-lassfaton auray of the regons of nterest. To lassfy the regons, 1) a Classfaton and Regresson (C &RT) deson tree s used to produe the pxel-level lassfaton and then ) a maorty vote sheme s appled to label eah of the regons of nterest based on the maorty of pxel labels wthn the orrespondng regon. The beauty of the proposed approah onssts n ts applablty to 1) perform automat segmentaton of regons of nterest by usng the lassfaton rules derved at the pxel-level; ) buld ontext-senstve reportng tools (for nstane, apply a omputer-aded dagnoss tool for lver only f the regon of nterest s a lver); and 3) develop eduatonal software for medne. Bakground Researh n the feld of mage lassfaton usng texture nformaton s mostly lmted to spef pathologes and a sngle organ tssue. In [11], Karkans et al. appled a multlayer feedforward neural network based on texture features to lassfy aner regons n olonosop mages. Baeg and Kehtarnavaz [13] proposed a three-layer bak-propagaton neural network based on two mage texture features to lassfy abnormaltes n dgtzed mammograms. Wroblewska et al. [15] proposed an automat feature seleton algorthm and multlayer feedforward bak-propagaton neural network to detet and lassfy of mro-alfatons n dgtal mammograms. Zaane et al. [14] proposed an automat lassfaton of bengn and malgn tumors n mammography mages based on an assoaton rule mnng approah. There s also some work done wth respet to the dentfaton of several organs at the abdomen level. In [9], a Hopfeld neural network was appled to perform organ dentfaton based on Haralk texture features appled at the pxel level. The advantage of ths tehnque s that t does not requre a pror knowledge. However, the tehnque gves poor results n dentfyng lver, spleen and musle beause of a blok-wse ontour effet from alulatng the texture features at the blok-level. Lee at al. [1] proposed a mult-module ontextual neural network and fuzzy rules to overome the dffultes n Koss s tehnque. The mult-module ontextual neural network segments mages by lassfyng the pxels nto dsont regon based on the gray-level and ontextual nformaton of neghborng pxels. Then, the fuzzy rule s appled to the regon features, nludng relatve loaton, relatve dstane, tssue-ntensty, area, ompatness, and elongatedness, n order to gve the fnal segmented organs. Furthermore, the texture feature analyss has been performed ether at the global-level (organ/tssue) or loal-level (small regon wthn the tssue/organ of nterest) rather than at the pxel-level. The DDP approah proposed n the paper wll extrat all possble texture haratersts wth respet to dreton and dsplaement nstead of onsderng all pxel par wthn a small neghborhood entered on the orrespondng pxel; furthermore, a deson tree s employed to do a pxel-based lassfaton. Sne ths proess s done at the pxel-level, the
3 lmtatons of the prevous approahes n dealng wth nonsstent organ shapes and postons, and overlappng gray-levels wll be overome by the proposed approah. The rest of ths paper s organzed as follows. The methodology of the approah for organ lassfaton s dsussed n Seton 3 and the evaluaton of the approah s desrbed n Seton 4. The prelmnary expermental results are presented n Seton 5; fnally, onlusons and future work are dsussed n Seton 6. 3 ethodology Our approah to pure path tssue lassfaton onssts of three stages: 1) pre-proessng; ) pxel-level texture extraton; and 3) lassfaton. Fgure 1 llustrates the overvew of our proposed system. Frst, the pre-proessng stage strengthens the ontrast n the mage and redues the nose n the data. Seond, the pxel-level texture stage extrats the haratersts of the texture at the loal level. Fnally, a supervsed learnng tehnque s utlzed n order to lassfy eah pxel, and further eah tssue of nterest. CT Images Pre-Proessng: Clpped Bnnng Enhane Image eghborhood of a pxel Pxel-Level Texture Extraton: Haralk Classfaton: Deson Tree (Pxel Level) Classfed Annotated mages 3.1 Pre-proessng: Gray-level reduton usng lpped Sne most of the soft tssues have overlappng gray-levels and the ontrast among the tssues s very low wthn the CT mages, espeally gven the large range of gray-levels n DICO (Dgtal Imagng and Communatons n edne) format of 4096 gray-levels for a 1 bt resoluton, an mage enhanement pre-proessng tehnque s needed to nrease the gray-level ontrast among the tssues. We apply a lpped bnnng tehnque [1] to enhane the ontrast wthn the soft tssues neessary for good texture feature extraton. The lpped bnnng tehnque norporates 1) a K- means algorthm that automatally determnes the range of the gray levels for the soft tssues n the gven CT mages; ) the gray-levels that are lower than the soft tssue range and the graylevels that are hgher than the soft tssue range wll be assgned to the mnmum bn and maxmum bn, respetvely; the gray values wthn the soft tssue range wll be lnearly dvded nto equal bns. 3. Pxel-level texture extraton Fgure 1: Dagram of the proposed system. Pxel-level texture extraton wll be used to dsover the texture of eah pxel wthn a small neghborhood. These overlappng regons allow a ertan ambguty to be norporated n the texture property for mprovng organ/tssue lassfaton. In order to apture the spatal dstrbuton of the gray-levels wthn the neghborhood, a two-dmensonal o-ourrene matrx
4 [3] s appled to alulate the loal texture nformaton around that pxel. Whle Kalnn et al. [4] and Koss et al. [9] alulated the loal texture based on an all-pars (AP) approah, whh onsders all the pxel pars wthn the neghborhood, we propose a new dreton vs. dsplaement pars (DDP) approah, whh onsders every dreton and dsplaement separately around the pxel of nterest. In ths way, the heker board problem ntrodued by the AP approah (two textures havng dfferent patterns wll have the same o-ourrene matrx f they have same values for the gray-levels) wll be overome by the proposed DDP approah. The DDP s based on the estmaton of ont ondtonal probablty of pxel par ourrenes P ( d, θ ). The P denotes the normalzed o-ourrene matrx by total number of the ourrene of two neghborng pxels between gray-ntensty at vertal dreton (row) and another gray-level at horzontal dreton (olumn) of spefy by dsplaement vetor d and angleθ. There are four dfferent dretons nludng 0, 45, 90, and 135 are generally used n mage proessng as n Fgure and (n-1) s a number of dsplaement vetor n n-by-n wndow sze, where n denotes resoluton vetor n row and olumn Fgure : From the entered pxel; pxel 1 represents 0 at d=1; pxel represents 45 ; pxel 3 represents 90 and pxel 4 represents 135 at d= Orgnal Image Co-Ourrene Fgure 3: Co-ourrene matrx for dstane 1, dreton 0 In terms of omputatonal tme, the DDP approah s Ο( n ) sne the number of pxel omparsons of DDP s gven by formula (1), where n denotes the sze of the wndow: n n n + (1) = 1 = 1 Furthermore, for omputatonal effeny purposes, the o-ourrene matrx mplementaton represents only the gray-levels that appear wthn the pxel neghborhood under onsderaton. One the o-ourrene matrx s alulated around eah pxel, ten Haralk texture desrptors [] measurng dfferent propertes of the texture are obtaned; 1) Entropy to measure the randomness of gray-level dstrbuton; ) Energy to measure the ourrene of repeated pars wthn an mage; 3) Contrast to apture the loal ontrast n an mage; 4) Homogenety to measure the homogenety of the mage; 5) Sum Average to provde the mean of the gray ntensty wthn an mage; 6) Varane to estmate the varaton of gray level dstrbuton; 7) Correlaton to measure a orrelaton of pxel pars on gray-levels; 8) axmum probablty to represent the most predomnant pxel par n an mage, 9) Inverse Dfferene oment (ID) to measure the smoothness of an mage and 10) Cluster Tendeny to measure the groupng of pxels that have smlar gray-level values. The alulaton of Haralk features are shown n Appendx. At the end of ths stage, eah pxel wll be haraterzed along eah dstane and dsplaement by a 10-dmensonal vetor: [ d,, d ]
5 3.3 Classfaton The lassfaton stage onssts of stages: pxel-based texture lassfaton and pure path lassfaton Pxel-based texture lassfaton. There are many types of lassfers that ould be used to dfferentate among the organs/tssues n pxel-based spae. Deson trees have a relatvely fast learnng speed ompared wth other lassfaton tehnques and also have the ablty to elmnate the rrelevant attrbutes from the set of features. Furthermore, the deson trees do not make any assumptons about the dstrbuton of the data and ths property makes them approprate to be used when the dstrbuton of the data s unknown. In our approah, a Classfaton and Regresson Tree (C&RT) lassfer s appled to dsrmnate the pxel-based pattern among dfferent tssues [16]. C&RT mplements the lassfaton proess based on splttng the urrent node nto two hld nodes based on the predtors values whh are the texture desrptors n our ase. The best predtor s hosen usng the Gn mpurty ndex suh that eah hld node s more pure than ts parent node [16]. The goal s to produe subsets, leaf nodes, whh are as homogeneous as possble wth respet to the lass label. In order to not overft the data, there are several stoppng rtera used for the tree growth [5]: axmum tree depth: d nmum number of ases for the parent node (nternal node): n p nmum number of ases for hld nodes (left/ termnal node): n nmum hange n mpurty: mp Dependng on these parameter values (d, np, n, mp), a dfferent tree wll be produed. The optmal tree wll be the tree whose parameters produe the hghest lassfaton auray at the pxel level on the testng data. One the optmal deson tree s developed, ts deson rules wll be used to lassfy eah pxel wthn the tssue regon of nterest Tssue lassfaton. The lassfaton of a tssue regon nvolves a maorty sheme n whh eah pxel from the regon s lassfed usng the rules derved from the C&RT. The most frequent label wthn the orrespondng regon wll beome the lassfaton label of that path. Therefore, the tssue lassfaton s the representaton of the maorty of predted tssues at the pxel level. Sne a set of rules s appled to eah pxel wthn the regon, the tssue lassfaton s ndependent of the tssue regons szes, so regons of dfferent szes an be automatally lassfed by our proposed approah. Ths advantage wll allow us to use the tehnque for mage segmentaton n regons havng not neessary equal szes as t s the ase for multple organs/tssues segmentaton n CT mages. 4 Evaluaton model The evaluaton of our approah s performed wth respet to eght dfferent pure tssue pathes nludng aorta, fat, kdneys, lver, lung, musle, spleen, and trabeular bone. The segmented pure tssues have been manually generated by a lnal expert. The proposed results wll be ompared aganst the lnal expert s manual path labels; n order to assess the auray of the approah, eah tssue wll be evaluated separately.
6 Two lassfaton metrs are used to measure the lassfaton performane: TP senstvty = () TP + F T spefty = (3) T + FP n whh TP denote true postves, FP denote false postves, T denotes true negatves, and F denote false negatves. True postve s a number of the pathes that belong to that organ that are orretly lassfed as that organ. False postve s a number of pathes that belong to that organ that are norretly lassfed as that organ. True negatve s a number of pathes that belong to other organs that are orretly lassfed as other organs. False negatve s a number of the pathes that belong to that organ that are norretly lassfed as other organs. 5 Expermental results 5.1 Data In ths study, 440 pure pathes (55 pathes per organ/tssue) from multple, seral, axal normal CT mages derved from helal, mult-detetor CT aqustons of 5 patents were manually segmented and annotated by an expert radologst. The mages were n DICO format, sze of 51 by 51 wth 1 bt resoluton. The segmented tssues have been manually annotated n one of the followng ategores: aorta, fat, kdneys, lver, lung, musle, spleen, and trabeular bone; some examples of pure pathes are shown n Fgure 4. Kdney Lver Orgnal Clpped Orgna Clpped Spleen Trabeular bone Orgnal Clpped Orgna Clpped Fgure 4: Result of pure pathes ontrast enhanement usng lpped bnnng wth 56 bns. 5. Pre-proessng We found that the soft tssues n DICO format (wth 4096 gray levels) are nearly always plaed nto an approxmately gray-level range between 856 and Therefore, the lpped bnnng approah alloates all gray ntensty wthn a lower bound (0-855) to a sngle lowest bn and alloates a sngle maxmum bn to an upper bound ( ). The rest of the gray levels are equally dvded nto equal szed bns; by varyng the number of gray levels, we found that 56 equal-szed bns s an optmum number for ontrast enhanement of soft tssues n CT mages [1].
7 5.3 Texture features We assess the o-ourrene matrx alulaton for both the AP and DDP approahes: AP alulates the o-ourrene by onsderng all the pxel pars wthn the neghborhood; DDP s an ntegraton of every possble o-ourrene whh onsders dreton and dsplaement separately. For the DDP approah, we onsder four dfferent dretons of 0, 45, 90, and 135 ; n terms of dsplaement, sne DDP approah onsders every dreton and dsplaement separately, the values of the dsplaement wll vary aordngly to the pxel wndow sze. For example, n a 5-by-5 wndow, we wll have 4 dfferent dsplaements and thus, 16 (4 dretons by 4 dsplaements) o-ourrene matres n total. The ten Haralk texture features are alulated for eah o-ourrene matrx. 5.4 Classfaton To obtan the C&RT deson tree, we dvded the 440 pure pathes nto a tranng set ontanng a random samplng of 66% of the pxels and a testng set ontanng the other 34%. Sne the texture feature have dfferent value as a pre-proessng step, we saled all the features to the range 0 to 1; the salng was done usng the max-mn normalzaton approah [5]. The optmal tree was found for number of parents equal to 100 and number of hldren equal to 5. For ths tree, 155 rules and 188 rules were found for the AP and DDP approahes, respetvely, for a wndow sze equal to 13 by 13. In both approahes, the lowest number of rules was obtaned for fat, lung, and musle, and the hghest number of rules was obtaned for lver and spleen; ths ndates that there s not muh dfferene among the texture pxels for fat/lung/musle pxels, whle there s sgnfant dfferene among the tssue pxels of a sngle organ/tssue suh as lver or spleen. One all the pxels n the pure-path are lassfed usng the derved deson rules, the hghest frequeny of the organ/tssue label n the path wll be used as the label representaton of ths path n order to get the fnal lassfaton. In addton to the lassfaton task, the deson tree approah helps us dentfy the most mportane features for dsrmnatng among the organs/tssues. We found that the ID desrptor s the most mportant feature for the lassfaton task. 5.5 Path Classfaton Results Wth respet to the way of alulatng the texture features, n all ases, the DDP approah produed at least the same or better results than the AP approah (Tables 1 & ) regardless of the organ onsderaton. Varyng the wndow sze from 5 by 5 to 13 by 13, had a slght effet on the result of AP approah, but had a sgnfant effet on the DDP approah. An explanaton of ths result may rely on the fat that, as the wndow beomes larger, more loal texture propertes are aptured for the alulaton of the o-ourrene matrx, and thus of the texture features used for lassfaton. The wndow sze was vared from 5-by-5 to 13-by-13 beause 1) senstvty of aorta starts to derease from 9-by-9 wndow; ) at 13-by-13 most of tssue lassfatons get almost 100% lassfaton senstvty; and 3) the omputaton tme nvolved n DDP oourrene alulaton nreases ubally wth the sze of the wndow. Table 1 presents the sgnfant mprovement of the DDP approah s senstvty wth the wndow sze. For example, the senstvty of kdneys rses from 64% to 97% and there s a
8 sgnfant mprovement from 61% to 100% for spleen, when the wndow sze grows up from 5- by-5 to 13-by-13. The senstvty of aorta went up from 45% to 81% when wndow sze nreased from 5-by-5 to 9-by-9; however, the senstvty started droppng at that pont. Thus, from these prelmnary results, t seems lke there are dfferent patterns wthn the aorta tssue pathes and a 9-by-9 wndow sze wll be more approprate for aorta. On the other hand, the senstvty for fat, lver, lung and bakbone dd not nrease sgnfantly beause the senstvty was already almost 100% at a wndow sze of 5-by-5. Table presents the spefty results for the eght organs of nterest. Regardless of the wndow sze, the spefty value for eah organ was above 90% for both AP and DDP approahes. Table 1: Senstvty for the pure path lassfaton for DDP and AP Aorta Fat Kdney Lver Lung usle Spleen Trabeular Bone Wndow DDP AP DDP AP DDP AP DDP AP DDP AP DDP AP DDP AP DDP AP 5x5 45.8% 0.00% 98.18% 98.18% 63.64% 36.36% 100% 100% 100% 100% 100% 100% 60.47% 44.44% 100% 9.86% 7x7 67.9% 1.8% 100% 98.15% 81.48% 44.44% 100% 100% 100% 100% 100% 100% 83.33% 67.44% 100% 93.75% 9x % 7.7% 100% 100% 78.85% 51.9% 100% 98.18% 100% 100% 100% 100% 94.44% 83.33% 100% 95.4% 11x % 30.61% 100% 100% 86.67% 55.56% 100% 98.18% 100% 100% 100% 100% 94.34% 86.79% 100% 100% 13x % 30.43% 100% 100% 96.55% 7.41% 100% 100% 100% 100% 100% 100% 100% 96.67% 100% 100% Table : Spefty for the pure path lassfaton for DDP and AP Aorta Fat Kdney Lver Lung usle Spleen Trabeular Bone Wndow DDP AP DDP AP DDP AP DDP AP DDP AP DDP AP DDP AP DDP AP 5x % 99.44% 100% 100% 96.4% 96.08% 94.04% 91.60% 100% 100% 99.70% 99.7% 94.56% 90.78% 95.9% 91.08% 7x % 98.80% 100% 100% 96.39% 96.11% 96.37% 95.50% 100% 100% 100% 99.70% 96.79% 91.30% 96.55% 9.4% 9x % 99.3% 100% 100% 98.04% 95.96% 99.00% 96.60% 100% 100% 100% 100% 95.54% 90.17% 96.6% 9.18% 11x % 99.30% 100% 100% 95.43% 97.% 98.93% 97.48% 100% 100% 100% 100% 96.30% 91.43% 96.50% 9.78% 13x13 100% 100% 100% 100% 9.97% 95.43% 98.43% 97.13% 100% 100% 100% 100% 97.64% 96.55% 99.4% 96.59% 6 Conluson and future work Our prelmnary results show that that the o-ourrene alulaton sheme at the pxel level and the wndow sze have a sgnfant nfluene upon the lassfaton results, espeally for the soft tssues. Wth respet to the DDP approah, the optmal wndow sze was 13-by-13 allowng the senstvty metr to be at least 96% for all organs exept for aorta. For aorta, the optmal wndow sze was 9-by-9 wth the lassfaton senstvty beng 81%. Therefore, dfferent wndow szes are approprate for dfferent organs and the larger the wndow sze, the more loal nformaton wll be aptured to alulate the texture. In terms of the overall lassfaton result and operatonal effeny, we suggest that the best wndow sze for dentfy regons of nterest s 9-by-9. As future work, the proposed approah an be extended to 1) nlude other texture features; and ) assgn probablst labels to the regons of nterest nstead of ust the label whh most predomnant wthn the orrespondng regon. Ths wll allow applyng the proposed approah for 1) probablst annotaton of unknown pure pathes; ) automat probablst segmentaton of pure pathes n CT mages usng splt and merge segmentaton algorthms; and 3) reaton of ontext-senstve tools for CAD systems. We also plan to nvestgate the effet of
9 usng three dmensonal texture models and norporatng more sophstated deson algorthms to move from a pxel lassfaton to a path lassfaton. Fnally, we wll begn testng the applablty of texture-based lassfaton to ertan pathologes. Referenes [1] R. Lerman, D. S. Rau, and J. D. Furst, Contrast enhanement of soft tssues n Computed Tomography mages, SPIE edal Imagng, 006. [] R.. Haralk, K. Shanmugam, and I. Dnsten, Textural Features for Image Classfaton, IEEE Trans. on Systems, an, and Cybernets, vol. Sm-3, no.6, pp , [3] R. Jan, R. Kastur, and B.G. Shunk, ahne Vson, ew York: Graw-Hll, [4]. Kalnn, D.S. Rau, J. Furst, and D.S. Channn, A lassfaton Approah for anatomal regons segmentaton, IEEE Int. Conf. on Image Proessng, 005. [5] D. Channn, D. S. Rau, J. D. Furst, D. H. Xu, L. Llly, and C. Lmpsangsr, "Classfaton of Tssues n Computed Tomography usng Deson Trees", The 90th Sentf Assembly and Annual eetng of Radology Soety of orth Amera (RSA04), 004. [6]. Kass, A. Wtkn, and D. Terzopoulos, Snakes: Atve ontour models, Int l. J. of Comp. Vs. 1(4), [7] E. Persoon, K. S. Fu, Shape dsrmnaton usng Fourer desrptors, IEEE Trans. Pattern Analyss & ahne Intellgene., vol. Pam-8, no. 3, pp , [8] R.. Dave, T. Fu, Robust shape deteton usng Fuzzy lusterng: Pratal applaton, Fuzzy Sets Syst., vol. 65, pp , [9] J. E. Koss, F. D. ewman, T. K. Johnson, D. L. Krh, Abdomnal organ segmentaton usng texture transform and a Hopfeld neural network, IEEE Trans. edal Imagng, vol. 18, no. 7, pp , [10]. Kobash, L. G. Shapro, Knowledge-based organ dentfaton from CT mages, Pattern Reognton, vol. 8, no. 4, pp , [11]S. A. Karkans, et al., Detetng abnormaltes n olonosop mages by texture desrptors and neural networks, Pro. of the Workshop ahne Learnng n ed. App., pp 59-6, [1]C. C. Lee, P. C. Chung, and H.. Tsa, Identfyng multple abdomnal organs from CT mage seres usng a mult-module ontextual neural network and spatal fuzzy rules, IEEE Trans. on Info. Teh. Bomedal, vol. 7, no. 3, 003. [13]S. Baeg,. Kehtarnavaz, Texture based Classfaton of ass Abnormaltes n ammograms, Pro. of IEEE CBS Symposum, 000. [14]O. R. Zaane,. L. Antone, A. Coman, ammography Classfaton by an Assoaton Rule-based Classfer, Int. Workshop on ultmeda Data nng., 00 [15]A. Wroblewska, P. Bonnsk, A. Przelaskowsk, Kazubek, Segmentaton and feature extraton for relable lassfaton of mro alfatons n dgtal mammograms, Opto- Eletron. Rev., vol. 11, no. 3, 003. [16]R. O. Duda, P. E. Hart, D. G. Stork, Pattern Classfaton, A Wley-Intersene, 001
10 Appendx Suppose and represent resoluton vetor at row and olumn respetvely. The ten Haralk features are alulated as the followng: Entropy: P log P Energy: P Contrast: ( ) P P Homogenety: ; Sum Average: 1 ( P + P ) Varane: µ + µ Correlaton: axmum Probablty: 1, ax P, (( ) P ( ) P ) r ( µ r)( µ ) P σ σ r P Inverse Dfferene oment: 1 + ( ) Cluster Tendeny: ( µ + µ ) P where µ r, µ, σ r, σ are the mean and varane of row and olumn defned as follow: µ r = P, µ = P r ( r) P, σ ( ) = µ P σ = µ r
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 informationPerformance 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 informationLOCAL 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 informationBoosting 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 informationProgressive 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 informationLink 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 informationTAR 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 informationPattern 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 informationSteganalysis of DCT-Embedding Based Adaptive Steganography and YASS
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. lu@shsu.edu ABSTRACT Reently well-desgned
More informationMatrix-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 informationFEATURE 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 informationCluster ( 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 informationAvatar 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 informationInterval 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 informationComputing 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 informationAdaptive 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 informationMultilabel 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 informationBit-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 informationA 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 informationImprovement 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 informationFuzzy 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 informationA 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 informationCS 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 informationConnectivity 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 informationFUZZY 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 informationBottom-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 informationGabor-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 informationDETECTING 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 informationResearch 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 informationMULTIPLE 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 informationMicroprocessors 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 informationEvaluation 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 informationA 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 informationAn 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 informationPerformance 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 informationarxiv: 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 informationMeasurement 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 informationInternational Journal of Pharma and Bio Sciences HYBRID CLUSTERING ALGORITHM USING POSSIBILISTIC ROUGH C-MEANS ABSTRACT
Int J Pharm Bo S 205 Ot; 6(4): (B) 799-80 Researh Artle Botehnology Internatonal Journal of Pharma and Bo Senes ISSN 0975-6299 HYBRID CLUSTERING ALGORITHM USING POSSIBILISTIC ROUGH C-MEANS *ANURADHA J,
More informationSession 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 informationElsevier Editorial System(tm) for NeuroImage Manuscript Draft
Elsever Edtoral System(tm) for NeuroImage Manusrpt Draft Manusrpt Number: Ttle: Comparson of ampltude normalzaton strateges on the auray and relablty of group ICA deompostons Artle Type: Tehnal Note Seton/Category:
More information25. Mining Knowledge in Computer Tomography Image Databases
25. Mnng Knowledge n Computer Tomography Image Databases Danela Stan Racu Summary: Ths chapter presents our research results obtaned for texture extracton, classfcaton, segmentaton, and retreval of normal
More informationContent 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 informationData Mining: Model Evaluation
Data Mnng: Model Evaluaton Aprl 16, 2013 1 Issues: Evaluatng Classfcaton Methods Accurac classfer accurac: predctng class label predctor accurac: guessng value of predcted attrbutes Speed tme to construct
More informationClassifying 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 informationOn the End-to-end Call Acceptance and the Possibility of Deterministic QoS Guarantees in Ad hoc Wireless Networks
On the End-to-end Call Aeptane and the Possblty of Determnst QoS Guarantees n Ad ho Wreless Networks S. Srram T. heemarjuna Reddy Dept. of Computer Sene Dept. of Computer Sene and Engneerng Unversty of
More informationMixture Models and the Segmentation of Multimodal Textures. Roberto Manduchi. California Institute of Technology.
Mxture Models and the Segmentaton of Multmodal Textures oberto Manduh Jet ropulson Laboratory Calforna Insttute of Tehnology asadena, CA 91109 manduh@pl.nasa.gov 1 Introduton Abstrat Aproblem wth usng
More informationFULLY 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 informationA 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 informationA Toolbox for Easily Calibrating Omnidirectional Cameras
A oolbox for Easly Calbratng Omndretonal Cameras Davde Saramuzza, Agostno Martnell, Roland Segwart Autonomous Systems ab Swss Federal Insttute of ehnology Zurh EH) CH-89, Zurh, Swtzerland {davdesaramuzza,
More informationSemi-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 informationAn 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 informationEdge 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 informationMulti-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 informationOutline. 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 informationSubspace 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 informationON THE USE OF THE SIFT TRANSFORM TO SELF-LOCATE AND POSITION EYE-IN-HAND MANIPULATORS USING VISUAL CONTROL
XVIII Congresso Braslero de Automáta / a 6-setembro-00, Bonto-MS ON THE USE OF THE SIFT TRANSFORM TO SELF-LOCATE AND POSITION EYE-IN-HAND MANIPULATORS USING VISUAL CONTROL ILANA NIGRI, RAUL Q. FEITOSA
More informationActive Contours/Snakes
Actve Contours/Snakes Erkut Erdem Acknowledgement: The sldes are adapted from the sldes prepared by K. Grauman of Unversty of Texas at Austn Fttng: Edges vs. boundares Edges useful sgnal to ndcate occludng
More informationEfficient automatic correction and segmentation based 3D visualization of magnetic resonance images
Lousana State Unverst LSU Dgtal Commons LSU Dotoral Dssertatons Graduate Shool 5 Effent automat orreton and segmentaton based 3D vsualzaton of magnet resonane mages Mkhal V. Mlhenko Lousana State Unverst
More informationAVideoStabilizationMethodbasedonInterFrameImageMatchingScore
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 informationOptimal 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 informationOutline. 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 informationLanguage Understanding in the Wild: Combining Crowdsourcing and Machine Learning
Language Understandng n the Wld: Combnng Crowdsourng and Mahne Learnng Edwn Smpson Unversty of Oxford, UK edwn@robots.ox.a.uk Pushmeet Kohl Mrosoft Researh, Cambrdge, UK pkohl@mrosoft.om Matteo Venanz
More informationImproved 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 informationA 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 informationTN348: 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 informationClustering incomplete data using kernel-based fuzzy c-means algorithm
Clusterng noplete data usng ernel-based fuzzy -eans algorth Dao-Qang Zhang *, Song-Can Chen Departent of Coputer Sene and Engneerng, Nanjng Unversty of Aeronauts and Astronauts, Nanjng, 210016, People
More informationFitting: Deformable contours April 26 th, 2018
4/6/08 Fttng: Deformable contours Aprl 6 th, 08 Yong Jae Lee UC Davs Recap so far: Groupng and Fttng Goal: move from array of pxel values (or flter outputs) to a collecton of regons, objects, and shapes.
More informationParallelism 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 informationRecognizing 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 informationMultiscale Heterogeneous Modeling with Surfacelets
759 Multsale Heterogeneous Modelng wth Surfaelets Yan Wang 1 and Davd W. Rosen 2 1 Georga Insttute of Tehnology, yan.wang@me.gateh.edu 2 Georga Insttute of Tehnology, davd.rosen@me.gateh.edu ABSTRACT Computatonal
More informationMULTISPECTRAL 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 informationThe 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 informationMachine Learning: Algorithms and Applications
14/05/1 Machne Learnng: Algorthms and Applcatons Florano Zn Free Unversty of Bozen-Bolzano Faculty of Computer Scence Academc Year 011-01 Lecture 10: 14 May 01 Unsupervsed Learnng cont Sldes courtesy of
More informationSteps for Computing the Dissimilarity, Entropy, Herfindahl-Hirschman and. Accessibility (Gravity with Competition) Indices
Steps for Computng the Dssmlarty, Entropy, Herfndahl-Hrschman and Accessblty (Gravty wth Competton) Indces I. Dssmlarty Index Measurement: The followng formula can be used to measure the evenness between
More informationA 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 information3D Scene Reconstruction System from Multiple Synchronized Video Images
3D Sene Reonstruton Sstem from Multple Snhronzed Vdeo Images aewoo Han 1, Juho Lee 2, Hung S. Yang 3 AIM Lab., EE/CS Dept., KAIS 1,2,3 373-1, Guseong-dong, Yuseong-gu, Daejon, Republ of Korea { bluebrd
More informationAnalysis 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 informationTime Synchronization in WSN: A survey Vikram Singh, Satyendra Sharma, Dr. T. P. Sharma NIT Hamirpur, India
Internatonal Journal of Enhaned Researh n Sene Tehnology & Engneerng, ISSN: 2319-7463 Vol. 2 Issue 5, May-2013, pp: (61-67), Avalable onlne at: www.erpublatons.om Tme Synhronzaton n WSN: A survey Vkram
More informationShape 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 informationSimulation: Solving Dynamic Models ABE 5646 Week 11 Chapter 2, Spring 2010
Smulaton: Solvng Dynamc Models ABE 5646 Week Chapter 2, Sprng 200 Week Descrpton Readng Materal Mar 5- Mar 9 Evaluatng [Crop] Models Comparng a model wth data - Graphcal, errors - Measures of agreement
More informationA 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 informationFace 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 informationFuzzy 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 informationClassifier 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 informationImage Representation & Visualization Basic Imaging Algorithms Shape Representation and Analysis. outline
mage Vsualzaton mage Vsualzaton mage Representaton & Vsualzaton Basc magng Algorthms Shape Representaton and Analyss outlne mage Representaton & Vsualzaton Basc magng Algorthms Shape Representaton and
More informationA Symbolic Representation of Time Series, with Implications for Streaming Algorithms
A Symbol Representaton of Tme Seres, th Implatons for Streamng Algorthms Jessa Ln Eamonn Keogh Stefano Lonard Bll Chu Unversty of Calforna - Rversde Computer Sene & Engneerng Department Rversde, CA 9252,
More informationClustering Data. Clustering Methods. The clustering problem: Given a set of objects, find groups of similar objects
Clusterng Data The lusterng problem: Gven a set of obets, fnd groups of smlar obets Cluster: a olleton of data obets Smlar to one another wthn the same luster Dssmlar to the obets n other lusters What
More informationA 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 informationA 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 informationAUTOMATICALLY MULTIPLE FEATURES DETECTION OF FACE SKETCH BASED ON MAXIMUM LINE GRADIENT
AUTOMATICALLY MULTIPLE FEATURES DETECTION OF FACE SKETCH BASED ON MAXIMUM LINE GRADIENT Arf Muntasa, Mohamad Harad, Maurdh Her Purnomo 3,,3 Eletral Engneerng Department, Insttut Teknolog Sepuluh Nopember,
More informationS1 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 informationMinimize 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 informationEYE 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 informationHierarchical clustering for gene expression data analysis
Herarchcal clusterng for gene expresson data analyss Gorgo Valentn e-mal: valentn@ds.unm.t Clusterng of Mcroarray Data. Clusterng of gene expresson profles (rows) => dscovery of co-regulated and functonally
More information1. Introduction. Abstract
Image Retreval Usng a Herarchy of Clusters Danela Stan & Ishwar K. Seth Intellgent Informaton Engneerng Laboratory, Department of Computer Scence & Engneerng, Oaland Unversty, Rochester, Mchgan 48309-4478
More informationA PATTERN RECOGNITION APPROACH TO IMAGE SEGMENTATION
1 THE PUBLISHING HOUSE PROCEEDINGS OF THE ROMANIAN ACADEMY, Seres A, OF THE ROMANIAN ACADEMY Volume 4, Number 2/2003, pp.000-000 A PATTERN RECOGNITION APPROACH TO IMAGE SEGMENTATION Tudor BARBU Insttute
More informationPictures at an Exhibition
1 Pctures at an Exhbton Stephane Kwan and Karen Zhu Department of Electrcal Engneerng Stanford Unversty, Stanford, CA 9405 Emal: {skwan1, kyzhu}@stanford.edu Abstract An mage processng algorthm s desgned
More informationREGISTRATION 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 informationModeling Radiometric Uncertainty for Vision with Tone-mapped Color Images
1 Modelng Radometr Unertanty for Vson wth Tone-mapped Color Images Ayan Chakrabart, Yng Xong, Baohen Sun, Trevor Darrell, Danel Sharsten, Todd Zkler, and Kate Saenko arxv:1311.6887v [s.cv] 9 Apr 14 Abstrat
More informationThe 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