Natural Image Denoising: Optimality and Inherent Bounds

Size: px
Start display at page:

Download "Natural Image Denoising: Optimality and Inherent Bounds"

Transcription

1 atural Image Denosng: Optmalty and Inherent Bounds Anat Levn and Boaz adler Department of Computer Scence and Appled Math The Wezmann Insttute of Scence Abstract The goal of natural mage denosng s to estmate a clean verson of a gven nosy mage, utlzng pror knowledge on the statstcs of natural mages. The problem has been studed ntensvely wth consderable progress made n recent years. However, t seems that mage denosng algorthms are startng to converge and recent algorthms mprove over prevous ones by only fractonal db values. It s thus mportant to understand how much more can we stll mprove natural mage denosng algorthms and what are the nherent lmts mposed by the actual statstcs of the data. The challenge n evaluatng such lmts s that constructng proper models of natural mage statstcs s a long standng and yet unsolved problem. To overcome the absence of accurate mage prors, ths paper takes a non parametrc approach and represents the dstrbuton of natural mages usng a huge set of 0 0 patches. We then derve a smple statstcal measure whch provdes a lower bound on the optmal Bayesan mnmum mean square error (MMSE). Ths mposes a lmt on the best possble results of denosng algorthms whch utlze a fxed support around a denosed pxel and a generc natural mage pror. Our fndngs suggest that for small wndows, state of the art denosng algorthms are approachng optmalty and cannot be further mproved beyond 0.dB values.. Introducton atural mage denosng s defned as the problem of estmatng a clean verson of a nose corrupted mage, gven a- pror knowledge that the orgnal unknown sgnal s a natural mage. The man dea n ths settng s that mage denosng can be performed usng not only the nosy mage tself but rather usng a sutable pror on natural mage statstcs. Image denosng algorthms have drastcally advanced over the last few decades [7, 7, 20, 2, 3, 0, 8]. Varous mage prors have been learned [20, 2, 26], and partcularly mpressng results have been obtaned wth non parametrc technques such as non local means [3] or BM3D [8], and sparse representaton methods such as KSVD [0]. However, t seems that the performance of denosng algorthms s startng to converge. Recent technques typcally mprove over prevous ones by only fractonal db values. In some cases the dfference between the results of competng algorthms s so small and nconclusve, that one actually has to successvely toggle between mages on a montor to vsually compare ther denosng qualty. Ths rases the queston of whether the error rates of current denosng algorthms can be reduced much further, or whether there are nherent lmtatons mposed by the statstcal structure of natural mages? The goal of ths paper s to derve a lower bound on the best possble denosng error under a well defned statstcal framework. Such a bound can help us understand f there s hope to sgnfcantly mprove the current stateof-the-art mage denosng wth even better algorthms, or whether we have nearly approached the fundamental lmts. Understandng the lmts of natural mage denosng s also mportant as an nstance of a more fundamental computer and human vson challenge: modelng the statstcs of natural mages and understandng the nherent lmts of ther statstcal power. Several works attempted to estmate the entropy of natural mages [5, 4]. However, there s no drect relaton between entropy and our ablty to solve low level vson tasks. Furthermore, numerous research attempts have been devoted to the learnng of natural mage prors [27, 2, 2, 26, 8, 2]. However, t s not clear f these models actually capture the full statstcal structure of natural mages. In prncple, a perfectly accurate natural mage pror would allow us to compute optmal Bayesan estmators for low level vson tasks such as denosng, superresoluton, deconvoluton and others. Therefore, understandng how far from optmalty do current denosng algorthms stand, wll gve us an ndcaton of how far are we from understandng and modelng the statstcal structure of natural mages. Several works studed the lmts of mage denosng. Some methods focused mostly on SR arguments [3, 24, 23] wthout takng nto account the strength of natural mage prors. Smlarly, lmts on the related problem of superresoluton have been derved n the past [, 6], but they agan account for the numercal stablty of the lnear system beng nverted, and do not model the addtonal gan provded by natural mage prors. In fact, n the same paper, 2833

2 Baker and Kanade [] show that ther super-resoluton lmts can be broken usng a face class pror. In the statstcal lterature, sharp bounds on mage denosng and edge detecton have been obtaned. However, these bounds assume over-smplfed mage models, such as pecewse constant regons separated by sharp edges wth smooth boundares [9, 4]. The non local means denosng algorthm [3] s proven to be asymptotcally optmal gven an nfntely large mage. However, ts expected error and devaton from optmalty on realstc fnte-sze mages s unclear. Recently, Chatteree and Mlanfar [5, 6] derved denosng bounds that do account for natural mage statstcs. Ther model assumes that the mage patches can be factorzed nto a small number of clusters, and each such cluster can be descrbed usng second order statstcs. Ths, however, s a strong assumpton whch can largely affect the conclusons. The lack of a more detaled analyss whch accounts for the statstcs of natural mages s due to the fact that modelng the statstcs of natural mages s an extremely hard problem, wth no agreeably good model suggested up to ths day. Thus, any evaluaton whch would rely on one of the exstng mage prors wll be strongly based by the naccuracy of the pror. In order to overcome ths challenge, our am n ths paper s to estmate a lower bound on mage denosng wthout usng any partcular parametrc model. Usng the sum of square dfferences error metrc, we derve a lower bound on the smallest possble reconstructon error of denosng algorthms whch consder a k k pxels support around a nosy pxel, and use a generc natural mage pror on k k patches. Thus, the possble gan by mage specfc prors [3, 0, 8], rather than generc ones, s beyond the scope of ths analyss. Our approach bulds on the success of recent large mage databases approaches n hgh level vson and graphcs applcatons [, 22], and represents the dstrbuton of natural mages n a non-parametrc way usng a huge set of 0 0 patches. We use ths patch set to approxmate the optmal Bayesan mnmum mean squared error (MMSE) estmator and ts expected error. Whle every fnte set of samples provdes only an approxmaton for the actual MMSE, we derve statstcal formulas whch, under certan condtons, allow us to compute both a lower bound and an upper bound on the MMSE. We show that for small support sze k or for large nose varance, the number of patches (0 0 ) s large enough. Hence, the lower and upper bounds concde and we can estmate the actual MMSE. At the more challengng cases of very large patch szes or very small nose levels, we only get a lower bound on the best possble denosng error. Our calculatons suggest that for the tested support szes, the state of the art denosng results of BM3D [8] are already close to optmalty, and cannot be further mproved beyond 0. db values. On the other hand, ncreasng the support sze does carry some potental for mproved denosng performance. However, ths ncrease should probably requre swtchng to parametrc approaches, snce non-parametrc approaches are nherently lmted by the densty of nearest neghbors, whch drastcally decreases wth wndow sze. Developng parametrc models whch can acheve state of the art denosng results s also useful snce they can be appled to other low-level vson tasks, and more mportantly, because ther generalzaton power can mprove our understandng of natural mages. 2. Boundng denosng performance 2.. Problem formulaton In mage denosng, one s gven a nosy verson y of a clean mage x y = x + n () and the am s to estmate a cleaner verson ŷ from y. In ths paper, we consder random nose vectors n whose entres are dstrbuted accordng to a zero mean..d. Gaussan wth varance σ 2. We restrct the dscusson to denosng algorthms whch denose each pxel separately usng a k k pxels support around t. Equvalently, we assume that x, y are k k patches and focus on the estmaton of the central pxel n each patch x, whch we denote by x c. The success of mage denosng algorthms s usually evaluated by PSR values, whch essentally measure the mean squared error (MSE) n reconstructon, averaged over a set of M test patches {(x, y )} M = «PSR = 0 log 0, MSE MSE M X (x,c ŷ,c) 2. (2) As dscussed n [25], the MSE may not provde an ultmate vsual qualty predcton. onetheless, we adopt the MSE for two reasons: a) t s the classcal measure beng optmzed by most exstng denosng algorthms, and b) t s correlated, though not perfectly, wth vsual qualty. Therefore, a lmt on MSE denosng s a reasonable proxy on how much we can mprove vsual qualty. The study of lower lmts correspondng to other qualty measures s an nterestng future research problem. Snce each nosy patch y could actually be generated by multple latent patches x, the problem s fundamentally ambguous and one cannot hope for a zero reconstructon error. Our goal s to understand what s the lowest possble MSE achevable by any denosng algorthm based on k k patches, gven a-pror knowledge that the nput mage has been randomly sampled from the set of natural mages. To ths end, we denote by p(x) the densty of k k patches of natural mages, and by p(y) the resultng densty of k k nosy patches. When context requres, we use the notaton p σ (y) to emphasze the explct dependence of the densty of nosy y patches on the nose level σ. The Gaussan..d. nose assumpton s not a hard one and the technque can be appled wth many other nose models. The non parametrc approach mostly requres the ablty to compute p(y x). 2834

3 2.2. Bayesan Mnmum Mean Square Error There are two equvalent nterpretatons for the MSE the error nterpretaton and the varance nterpretaton. In the error vew, we randomly sample clean patches x from p(x), add nose to generate y, denose y and measure the reconstructon error (x,c ŷ,c ) 2. The average of ths reconstructon error s Z Z MSE = p(x) p(y x)(x c ŷ c) 2 dydx (3) An equvalent nterpretaton of the MSE, whch wll serve us below, s the varance nterpretaton start from a nosy patch y, and measure the varance of p(x y) around t. That s, compute the sum of weghted dstances between ŷ and all possble x explanatons: Z Z MSE = p(y) p(x y)(x c ŷ c) 2 dxdy (4) Eq. (4) s obtaned from Eq. (3) by swtchng the order of ntegraton and applyng Bayes rule. The advantage of the varance vew s that from t, one can easly derve the optmal estmator, namely the Bayesan mnmum mean squared error (MMSE) estmator [3], whch s smply gven by the condtonal mean µ(y) = E[x c y] = p(x y)x c dx (5) = p(y x) p(y) p(x)x cdx The MMSE at any fxed y s then the condtonal varance, Z V(y) = E[(x c µ(y)) 2 y] = p(x y)(x c µ(y)) 2 dx (6) whereas the overall MMSE s MMSE = E[V(y)] = p(y)v(y)dy. (7) In the framework of natural mage denosng, the MMSE n (7) s the lowest achevable denosng error by any denosng algorthm. To compute ths MMSE and the correspondng optmal estmator µ(y), we need access to the true natural mage densty p(x). However, as dscussed n the ntroducton, an accurate model for ths densty s a long standng research problem [20, 2, 26, 9]. One opton would be to ft some parametrc model q(x), for example, a feld of experts model [2] or a Gaussan mxture model. We can then compute an approxmated mean ˆµ(y) = /q(y) p(y x)q(x)x c dx. However, snce q(x) s only an approxmaton to the true natural mage densty p(x), the error of the estmator ˆµ(y), lke the error of any denosng algorthm, s only an upper bound on the MMSE. The dfference n MSE depends on the qualty of the approxmaton, whch s not easy to evaluate, n partcular snce p(x) s unknown. Our goal, n contrast, s to estmate the MMSE wthout commttng to any specfc approxmate parametrc model, or essentally, wthout even explctly knowng p(x). The key dea s that even though p(x) s unknown, we are stll able to sample from t. Therefore we consder a non-parametrc representaton of the dstrbuton, usng a large set of about = 0 0 natural mage patches. Ths allows us to approxmate the ntegral of Eq. (5) by averagng over samples: ˆµ(y) = where for Gaussan nose p(y x) = P p(y x)x,c P p(y x), (8) x y 2 (2πσ 2 e 2σ ) d/2 2 (9) and d = k 2. As, ˆµ(y) ndeed converges to the true MMSE estmator. However, for any fnte set, Eq. (8) s only an approxmaton, and thus, lke any other denosng algorthm, ts average error provdes an upper bound on the MMSE. However, our goal s to bound the MMSE from below, stll, wthout knowng p(x) explctly, but havng at our dsposal samples from t. The man dea n dervng both an upper and a lower bound on the MMSE, s to use the two MSE formulatons n Eqs. (3) and (4). Gven a set of M clean and nosy pars {( x, y )} M = and another ndependent set of clean patches {x } =, both randomly sampled from natural mages, we compute MMSE U = (ˆµ(y ) x,c ) 2 (0) M MMSE L = ˆV(y ) () M where ˆV(y ) s the approxmated varance: ˆV(y ) = P p(y x)(ˆµ(y) x,c)2 P p(y x) (2) MMSE U and MMSE L are both random varables whch depend on the partcular set of x, y samples. When the sample sze approaches nfnty, they converge to the exact MMSE. However, we show that for a fnte sample, n expectaton, MMSE U and MMSE L provde upper and lower bounds on the best possble MMSE. ote the key dfference between these two quanttes: MMSE U uses explct knowledge of the orgnal nose-free patch x, whle MMSE L does not nvolve t. Snce MMSE U bascally measures the error of the estmator ˆµ(y ), t provdes, lke every denosng algorthm, an upper bound on the MMSE. The term MMSE L s analyzed n Sec When s suffcently large, the two computed values MMSE U and MMSE L are smlar, and we get an accurate estmate for the actual optmal MMSE. As we show n secton 3 ths happens when the patch sze k s small, or when 2835

4 the nose varance σ 2 s hgh, snce n such cases there s a large number of vald nearest neghbors around each patch. In the harder cases, we cannot compute the MMSE exactly, but MMSE L stll provdes a lower bound on the best denosng results we can expect The sample varance as a lower bound on the MMSE Our goal s to show that the approxmated varance estmated from a fnte set of samples (Eq. (2)), provdes a lower bound on the correct varance (Eq. (6)). To see ths at an ntutve level, consder the numerator of Eq. (2). Snce ˆµ(y) s the weghted mean of {x } =, t mnmzes the mean squared dstance from these patches, and n partcular, t acheves a smaller squared dstance compared to the exact unknown mean µ(y): p(y x )(x,c ˆµ(y )) 2 p(y x )(x,c µ(y )) 2 Thus, n expectaton " # X E p(y x )(x,c ˆµ(y )) 2 " # X E p(y x )(x,c µ(y )) 2 = p(y )V(y ), (3) where E [ ] denotes expectaton wth respect to all possble sets of..d. patches from p(x). The denomnator of Eq. (2) s also a random varable whose expectaton s p(y ), thus, roughly, E [ˆV(y)] V(y). However, snce both numerator and denomnator are random varables, a more careful analyss of ˆV(y) s needed. The followng clam derves a second order approxmaton to E [ˆV(y)] for a fxed patch y. Clam Asymptotcally, as the tranng set sze, the expected value of the approxmated varance defned n Eq. (2) s «E [ˆV(y)] = V(y) + C(y)B(y) + o, (4) wth B(y) = E σ [x 2 c y] 3E σ [x c y] 2 2E σ [x 2 c y] +4E σ [x c y]e σ [x c y] (5) C(y) = p σ (y) (4πσ 2 ) d/2 p σ (y) 2 (6) where p σ, E σ [ ], p σ, E σ [ ] denote probablty and expectaton of random varables wth nose standard dervaton σ and σ respectvely. σ s the actual standard devaton and σ = σ/ 2. The proof of Clam s provded n the appendx. It uses an asymptotc expanson of ˆV(y) n, and neglects hgh V(y) 4 x Correct Varance Approx. Varance, =5 Approx. Varance, =20 Approx. Varance, =00 σ y 0 σ y y B(y ) x (a) (b) Fgure. Evaluatng the bas of ˆV(y). (a) For a D Gaussan densty p(x), we compare the expected sample varance E [ˆV(y)] to ts exact value V(y), whch s constant n y for a Gaussan dstrbuton. As predcted theoretcally, E [ˆV(y)] V(y), wth the bas decreasng wth ncreasng number of samples or at hgh densty regons. (b) Plug-n bas estmates ˆB(y) of real natural mage patches, sorted n ascendng order. ote that up to small numercal errors, the bas estmates of all patches are negatve. order terms. It s hence vald for suffcently large and suffcently common y. ext, we consder a local Gaussan approxmaton for p(x) around x = y, e.g., a Laplace approxmaton consstng of a second order Taylor expanson of lnp(x). For a Gaussan dstrbuton we can analytcally compute the leadng order bas term C(y)B(y) from Eqs. (5) and (6). The followng clam, proven n the appendx, shows that B(y) s negatve for all y values. Snce C(y) > 0, t follows that E [ˆV(y)] V(y). Thus, n expectaton, MMSE L provdes a lower bound on the MMSE, as we am to show. Clam 2 For a Gaussan dstrbuton, B(y) 0 for all y. Fgure (a) consders a smple case of a D Gaussan dstrbuton and compares the exact varance V(y) wth E [ˆV(y)], averaged over 0, 000 realzatons of sets of = 5, 0 or 5 samples. The bas s always negatve and E [ˆV(y)] V(y). As evdent from Eq. (6), C(y) s large when p(y) s small, and so for a Gaussan dstrbuton, the bas ncreases exponentally wth y. The densty of natural mage patches s not Gaussan, so a Laplace approxmaton mght be naccurate and Clam 2 may not drectly apply. To evaluate ths on real data, we used a plug-n estmator ˆB(y) for the term B(y). That s, we consdered M = 2, 000 nosy 3 3 patches {y } wth nose varance σ = 8. For each y the expectatons n Eq. (5) were estmated by averagng over an ndependent large set of clean patches {x }. Fgure (b) shows the resultng ˆB(y ) values, sorted n ascendng order for vsualzaton purposes. As seen n the plot, the estmated values ˆB(y ) are all negatve up to small numercal errors. 3. Experments To evaluate the MMSE we use a set of 20, 000 mages from the LabelMe dataset [22]. We thus mplctly consder ths dataset an unbased representatve of nose-free 2836

5 PSR MMSE U MMSE L LMMSE BM3D KSVD GSM Support sze k PSR MMSE U MMSE L LMMSE BM3D KSVD GSM Support sze k PSR MMSE U MMSE L LMMSE BM3D KSVD GSM Support sze k (a) σ = 8 (b) σ = 55 (c) σ = 70 Fgure 2. PSR values of several recent denosng algorthms along wth our MMSE lower and upper bounds. As predcted by the theory, the performance of all algorthms are bounded by our MMSE L estmate, although BM3D approaches the bound by fractonal db values. (ote that snce PSR = 0log 0(MSE), the MMSE lower bound turns nto an upper bound on the best achevable PSR). natural mage statstcs. 2 To avod quantzaton and JPEG artfacts we frst low-passed all mages and down-sampled them by a factor of two. These mages provde a set of about = 0 0 natural mage patches {x }. We use another M = 2, 000 clean and nosy pars of patches {( x, y )}, denose them by computng the weghted mean of Eq. (8), and measure the mean squared error MMSE U (Eq. (0)), and the lower bound MMSE L (Eq. ()). Ths s an ntensve computaton, whch took about a week of computaton on a 00 CPUs cluster. In Fgure 2 we compare the PSR values of varous denosng algorthms to MMSE U and MMSE L wth three nose levels σ = 8, 55, 70, for mage ntenstes n the [0, 255] range. We plot our bounds as a functon of the wndow sze k. We tred to match the support sze of the tested algorthms as well. We used the BM3D [8] algorthm wth patch szes of 4, 8 and 2 pxels (for 8, 2 pxels we used the authors parameters, and for 4 pxels, our adaptaton). We assocate the KSVD [0] and Portlla et al. GSM[20] algorthms wth the largest support sze k consdered, snce these are global denosng algorsms that do not consder fnte support patches. As a baselne comparson we also present a lnear mnmum mean square error (LMMSE) estmator [3]. Ths estmator, also known as the Wener flter, uses only the second order statstcs of the data, by fttng a sngle k 2 dmensonal Gaussan to the set of mage patches. When computng our MMSE U and MMSE L scores we used the bootstrappng method to evaluate the varance, by drawng multple patch subsets. The standard devaton of the estmaton s rather small, rangng from 0.05dB for small k values to 0.2dB at the large ones. The results n Fgure 2 suggest several observatons whch we dscuss below. 2 Even f these nput mages do contan a small amount of nose, t s neglgble w.r.t. the nose level added n our experments. If nonetheless, these patches do contan a small amount of nose, then the goal s formulated as the reconstructon of sgnals wth ths slghtly perturbed dstrbuton, whch s stll a statstcally well defned problem. Tghtness of bounds: For small wndow szes or hgh nose, the upper and lower bounds concde and hence we have obtaned an accurate estmate of the exact optmal MMSE value. For harder cases, there s a gap between the upper and lower bounds. Whle we are unable to estmate the exact MMSE, we stll get a vald lower bound on t. Ths happens when there are too few samples at a normalzed dstance of σ 2 from the nosy patch. To demonstrate that, we measured the number of nearest neghbors wthn normalzed dstance of one standard devaton from each nosy patch. Under the assumpton of Gaussan nose the dfference y x 2 follows a χ 2 dstrbuton wth k degrees of freedom. Therefore, for each nosy patch y we consder all patches at a squared dstance of σ 2 ( + 2/k 2 ): n = # x k 2 x y 2 < σ 2 + «ff 2 k (7) In Fgure 3 we plot the cumulatve dstrbuton of n values, for two wndow szes k = 3 and k = 9, both wth σ = 8. We see that for small k, most patches have a large number of nearest neghbors, whereas for k = 9, 3% of the patches have zero neghbors. Accordngly, at k = 9, σ = 8, there s a gap between MMSE U and MMSE L n Fg 2(a). ear optmalty of state of the art denosng results: As predcted by the theory, the emprcal errors of all denosng algorthms are larger than our lower bound. onetheless, for most cases the PSR values of BM3D are wthn 0.dB of the optmal ones, suggestng that the BM3D results are qute close to optmalty, for small patch szes. Thus, future denosng algorthms that use small patches have very lttle room for mprovement over BM3D 3. However, whle 0.dB seems lke a very small mprovement, t may stll be vsually notceable. 3 Ths does not drectly mply that BM3D cannot be further mproved, snce whle BM3D s a patch based algorthm, t utlzes a wder support by lookng for neghbors n the entre mage. 2837

6 00 00 Examples percent, h(n) Examples percent, h(n) umber of neghbors, n umber of neghbors, n (a) k = 3 (b) k = 9 Fgure 3. Cumulatve dstrbuton functon of the number of neghbors wthn dstance σ 2 ( + 2/k), computed at σ = 8. For 3 3 patches, 99% of the examples had more than 2, 000 neghbors. In contrast, for a 9 9 patch sze, 3% of the examples had no neghbors wthn ths dstance. (a) Orgnal mage (b) osy nput Support sze: It seems that ncreasng the support sze carres some potental for ncreasng the PSR. Due to the curse of dmensonalty, non-parametrc technques suffer from an unavodable tradeoff between patch sze and the number of vald nearest neghbors, and hence are unlkely to be applcable for large wndow szes. Developng algorthms whch utlze a larger support sze thus probably requres swtchng from non-parametrc approaches to parametrc ones. Unfortunately, at the moment parametrc denosng algorthms are far behnd the non parametrc ones. (c) Opt. MMSE, PSR=23.93dB (d) BM3D, PSR=23.86dB Denosng at extreme nose levels: In Fg 2(c), the nose level s σ = 70. At ths hgh nose level, the mage pror results are very smlar to the results of a smple lnear mnmum mean square error (LMMSE) estmator whch utlzes only the second order statstcs of the data. Ths happens snce at hgh nose levels the sgnal content s lost and all the pror can do s to estmate a flat mage. The square error of such an estmator s proportonal to the varance of the data, therefore natural mage prors or Gaussan prors utlzng second order statstcs gve smlar results. The other extreme that we dd not evaluate here s very low nose. For low nose the effect of the pror s small and the error s proportonal to the nose varance. Therefore, natural mage prors have an nterestng effect only at medum nose levels. Vsual results: In Fgure 4 we used the Peppers mage at half resoluton to vsualze the denosng results of several recent denosng algorthms compared wth our approach, whch s the best possble denosng wth a generc natural mages pror. We used σ = 75 and 2 2 wndows for our approach and for BM3D. umercally, our result outperforms BM3D by 0.07dB, and ths dfference leads to a slghtly better vsual qualty. The vsualzaton of optmal denosng n Fg 4(c) s only a proof-of-concept, but not a practcal denosng algorthm- denosng ths mage requred two weeks of computaton on 00 CPUs. However, one can thnk of several ways to accelerate the (d) KSVD, PSR=22.4dB (e) GSM, PSR=23.28dB Fgure 4. Vsual comparson of our optmal MMSE and other algorthms, for σ = 75. The optmal MMSE and BM3D use 2 2 wndows. Our approach acheves slghtly hgher PSR comparng to BM3D, and a somewhat better vsual qualty. search wth approxmated nearest neghbor technques and a smarter encodng of the large set of 0 0 mage patches. Image specfc bounds: Our analyss s based on the knowledge of a generc mage pror. An nterestng queston for future research, as recently consdered by [5, 6], s whether there s a sgnfcant advantage n adaptng the pror to the specfc statstcs of the observed mage, as done by some recent denosng algorthms [3, 0, 8]. In fact, such algorthms can be thought of as nstances of a generc pror whose support sze s the entre mage, snce the support sze essentally means that an algorthm can see and use n the estmaton all pxels wthn a k k wndow around a nosy pxel. Unfortunately, we cannot evaluate tght bounds at very large wndow szes. 2838

7 4. Dscusson Ths paper derved a statstcal measure for the best possble denosng results utlzng a generc natural mage pror. Our fndngs suggest that for small wndows, state of the art denosng algorthms cannot be further mproved beyond fractonal db values. Consderng a wder support carres some potental for mproved results. However, ncreasng the support sze of non-parametrc algorthms mght lead to a dead-end, and parametrc approaches are requred. Unfortunately, at the moment parametrc algorthms perform well behnd non-parametrc ones. Developng parametrc algorthms whch can acheve state of the art results s also mportant because they can generalze for other lowlevel vson tasks. The evaluaton methodology used here can be easly appled to other low-level vson problems such as superresoluton, deconvoluton, or npantng. However, some of these problems, such as the deconvoluton of a wde support kernel, rely on a large support of pxels. Hence we may not be able to fnd a suffcent number of nearest neghbors for a tght lower bound estmaton. Yet, understandng the lmts of mage denosng s mportant as a step toward other lowlevel vson problems, and moreover, as a step toward an understandng of the nherent lmts of natural mage statstcs. Acknowledgments: The authors thank the ISF, BSF and ERC for provdng fnancal support. 5. Appendx Proof of Clam : For future use we denote by A(σ, k) = (4πσ 2 ) k2 /2 and recall that σ = σ/ 2. The frst step n the proof s to nsert Eq. (8) nto Eq. (2), whch yelds the followng more convenent expresson for ˆV(y), p(y x )x ˆV(y) 2,c = ( p(y x )x,c ) 2 p(y x ) ( p(y x )) 2 = A A 2. (8) The man dea s to analyze the bas of each of these terms separately, whereby for each term we further consder the mean and fluctuatons of ther numerator and denomnator. For the term A we shall use the followng equaltes E[p(y x)] = R p(x)p(y x)dx = p σ(y) E[p(y x)x 2 c] = p σ(y)e σ[x 2 c y] (9) where p σ (y) s the densty of y-patches at nose level σ, and E σ [ y] denotes expectaton wth respect to p σ (y). The two expressons n Eq.(9) are nothng but the mean of the denomnator and numerator of A, respectvely. We thus rewrte the term A as ( p σ (y)e[x 2 c y] + A = ( p σ (y) + p(y x)x ) 2,c pσ(y)e[x2 c y] p σ(y)e[x 2 c y] p(y x) p σ(y) p σ(y) ) (20) ext, we assume and that the patch y s not too rare, such that p(y x ) p σ (y) p σ (y) Then, usng a Taylor expanson for small ǫ, + ǫ = ǫ + ǫ2 + O(ǫ 3 ) we obtan the followng asymptotc expanson for the frst term, (where for ease of notaton we wrte p(y) for p σ (y))! A E[x 2 c y] + X p(y x )x 2,c p(y)e[x 2 c y] p(y)e[x 2 c y]! X p(y x ) p(y) + 2! X p(y x ) p(y) p(y) p(y) (2) We now take the expectaton of Eq. (2) over x samples. We use the fact that E[p(y x) p(y)] = 0, E[p(y x)x 2 c p(y)e[x 2 c y]] = 0. We also neglect all O(/2 ) terms. E [A ] = E[x 2 c y] + E[p(y x) 2 ] p(y) 2 E[p(y x) 2 x 2 c] p(y) 2 E[x 2 c y] + o(/) Smlarly, for the second term A 2 we use the fact that «(22) E[p(y x) 2 ] = R p(x) e x y 2 /σ 2 dx = A(σ, k)p σ (y) (2πσ 2 ) k2 E[p(y x) 2 x c] = A(σ, k)p σ (y)e σ [x c y] E[p(y x) 2 x 2 c] = A(σ, k)p σ (y)e σ [x 2 c y] and hence rewrte the second term A 2 as + A 2 = ˆµ(y) 2 = E[x c y] 2 + = E[x c y] 2 P + 2 p(y x )x,c p(y)e[x c y] p(y)e[x c y] + E[x c y] P + 2 p(y x ) p σ(y) p σ(y) + P p(y x )x,c p(y)e[x c y] p(y)e[x c y] + P 2 p(y x ) p σ(y) p σ(y) + 3 P p(y x )x,c p(y)e[x 2 c y] p(y)e[x c y] P 2 p(y x ) p σ(y) p σ(y) (23) P «p(y x )x,c p(y)e[x 2 c y] p(y)e[x c y] «2 P p(y x ) p σ(y) p σ(y) P «p(y x )x,c p(y)e[x 2 c y] p(y)e[x c y] P «2 p(y x ) p σ(y) p σ(y) (24) Takng expectatons and omttng O(/ 2 ) terms, we get: E [A 2] = E[x c y] 2 4 E[p(y x) 2 x c] p(y) 2 E[x + E[p(y x) 2 x 2 c] c y] p(y) 2 E[x c y] «E[p(y x) 2 ] + o(/) p(y) 2 (25) 2839

8 Substtutng the terms from Eq. (23) n Eqs. (22) and (25) yelds the desred Eq. (4). Proof of Clam 2: We note that B(y) (Eq. (5)) can be wrtten as: B(y) = V σ[x c y] 2V σ [x c y] 2(E σ [x c y] E σ[x c y]) 2 (26) where V[x c y] = E[x 2 c y] E[x c y] 2 denotes varance. Thus, t s suffcent to show that V σ [x y] 2V σ [x y] 0. We denote by Φ the covarance matrx of p(x) and by Ψ σ, Ψ σ the covarance matrces of p σ (x y), p σ (x y): ««Ψ σ = σ 2I d + Φ Ψ σ = σ 2I d + Φ, (27) where I d s the d dmensonal dentty matrx. The varance at x c s the (c, c) entry of these matrces: V σ [x y] = Ψ σ (c, c), V σ [x y] = Ψ σ (c, c). We denote by {λ l }, {γ l }, {γl } the egenvalues of the matrces Φ, Ψ σ, Ψ σ respectvely. We note that all these matrces are dagonal n the same bass and we can wrte Ψ σ(c, c) = X l u 2 lγ l, Ψ σ (c, c) = X l u 2 lγ l, (28) (u l s the c entry of egenvector l). We can also relate the egenvalues of Ψ σ, Ψ σ to the egenvalues of the uncondtonal covarance Φ: γ l = (λ l + σ 2 ), γ l = (λ l + (σ ) 2 ). (29) A smple calculaton shows that for every l: «γ l 2γl = λ l σ 2 λ l + σ 2 0 (30) 2 2λ l + σ 2 Usng Eqs. (30) and (28) we conclude: References Ψ σ(c, c) 2Ψ σ(c, c) = X l u 2 l(γ l 2γ l ) 0. (3) [] S. Baker and T. Kanade. Lmts on super-resoluton and how to break them. PAMI, [2] A. Bell and T. Senowsk. The ndependent components of natural scenes are edge flters. Vson Res., 997. [3] A. Buades, B. Coll, and J. Morel. A revew of mage denosng methods, wth a new one. Multscale Model. Smul., [4] D. Chandler and D. Feld. Estmates of the nformaton content and dmensonalty of natural scenes from proxmty dstrbutons. J. Opt. Soc. Am., [5] P. Chatteree and P. Mlanfar. Is denosng dead? IEEE Trans Image Processng, 200. [6] P. Chatteree and P. Mlanfar. Practcal bounds on mage denosng: From estmaton to nformaton. IEEE Trans Image Processng, 20. [7] R. Cofman and D. Donoho. Translaton-nvarant denosng. In Wavelets and Statstcs, 995. [8] K. Dabov, A. Fo, V. Katkovnk, and K. Egazaran. Image denosng by sparse 3-d transform-doman collaboratve flterng. IEEE Trans Image Processng, [9] J. Echhorn, F. Snz, and M. Bethge. atural mage codng n v: How much use s orentaton selectvty? PLoS Comput Bol, [0] M. Elad and M. Aharon. Image denosng va sparse and redundant representatons over learned dctonares. IEEE Trans Image Processng, [] J. Hays and A. Efros. Scene completon usng mllons of photographs. SIGGRAPH, [2] A. Hyvarnen, J. Hurr, and P. Hoyer. atural Image Statstcs A probablstc approach to early computatonal vson. Sprnger-Verlag, [3] Steven M. Kay. Fundamentals of statstcal sgnal processng: estmaton theory. Prentce-Hall, Inc., 993. [4] A. Korostelev and A. Tsybakov. Mnmax Theory of Image Reconstructon. Sprnger-Verlag, ew York, 993. [5] A. Lee, K. Pedersen, and D. Mumford. The nonlnear statstcs of hgh-contrast patches n natural mages. IJCV, [6] Z. Ln and H. Shum. Fundamental lmts of reconstructonbased superresoluton algorthms under local translaton. PAMI, [7] S. Osher, A. Sole, and L. Vese. Image decomposton and restoraton usng total varaton mnmzaton and the h norm. Multscale Model. Smul., [8] S. Osndero and G. Hnton. Modelng mage patches wth a drected herarchy of markov random felds. IPS, [9] J. Polzehl and V. Spokony. Image denosng: Pontwse adaptve approach. Annals of Statstcs, 3:30 57, [20] J. Portlla, V. Strela, M. Wanwrght, and E. Smoncell. Image denosng usng scale mxtures of gaussans n the wavelet doman. IEEE Trans Image Processng, [2] S. Roth and M.J. Black. Felds of experts: A framework for learnng mage prors. In CVPR, [22] B. Russell, A. Torralba, K. Murphy, and W. Freeman. Labelme: a database and web-based tool for mage annotaton. IJCV, [23] T. Trebtz and Y. Schechner. Recovery lmts n pontwse degradaton. In ICCP, [24] M. Unser, B. Trus, and A. Steven. A new resoluton crteron based on spectral sgnal-to-nose ratos. Ultramcroscopy, 987. [25] Z. Wang, A. C. Bovk, H. R. Shekh, and E. P. Smoncell. Image qualty assessment: From error vsblty to structural smlarty. IEEE Trans. on Image Processng, [26] Y. Wess and W. T. Freeman. What makes a good model of natural mages? In CVPR, [27] S. Zhu and D. Mumford. Pror learnng and gbbs reactondffuson. PAMI,

Study and Comparison of Various Techniques of Image Edge Detection

Study and Comparison of Various Techniques of Image Edge Detection Gureet Sngh et al Int. Journal of Engneerng Research Applcatons RESEARCH ARTICLE OPEN ACCESS Study Comparson of Varous Technques of Image Edge Detecton Gureet Sngh*, Er. Harnder sngh** *(Department of

More information

Parameter Estimates of a Random Regression Test Day Model for First Three Lactation Somatic Cell Scores

Parameter Estimates of a Random Regression Test Day Model for First Three Lactation Somatic Cell Scores Parameter Estmates of a Random Regresson Test Day Model for Frst Three actaton Somatc Cell Scores Z. u, F. Renhardt and R. Reents Unted Datasystems for Anmal Producton (VIT), Hedeweg 1, D-27280 Verden,

More information

Using Past Queries for Resource Selection in Distributed Information Retrieval

Using Past Queries for Resource Selection in Distributed Information Retrieval Purdue Unversty Purdue e-pubs Department of Computer Scence Techncal Reports Department of Computer Scence 2011 Usng Past Queres for Resource Selecton n Dstrbuted Informaton Retreval Sulleyman Cetntas

More information

4.2 Scheduling to Minimize Maximum Lateness

4.2 Scheduling to Minimize Maximum Lateness 4. Schedulng to Mnmze Maxmum Lateness Schedulng to Mnmzng Maxmum Lateness Mnmzng lateness problem. Sngle resource processes one ob at a tme. Job requres t unts of processng tme and s due at tme d. If starts

More information

International Journal of Emerging Technologies in Computational and Applied Sciences (IJETCAS)

International Journal of Emerging Technologies in Computational and Applied Sciences (IJETCAS) Internatonal Assocaton of Scentfc Innovaton and Research (IASIR (An Assocaton Unfyng the Scences, Engneerng, and Appled Research Internatonal Journal of Emergng Technologes n Computatonal and Appled Scences

More information

Copy Number Variation Methods and Data

Copy Number Variation Methods and Data Copy Number Varaton Methods and Data Copy number varaton (CNV) Reference Sequence ACCTGCAATGAT TAAGCCCGGG TTGCAACGTTAGGCA Populaton ACCTGCAATGAT TAAGCCCGGG TTGCAACGTTAGGCA ACCTGCAATGAT TTGCAACGTTAGGCA

More information

Subject-Adaptive Real-Time Sleep Stage Classification Based on Conditional Random Field

Subject-Adaptive Real-Time Sleep Stage Classification Based on Conditional Random Field Subject-Adaptve Real-Tme Sleep Stage Classfcaton Based on Condtonal Random Feld Gang Luo, PhD, Wanl Mn, PhD IBM TJ Watson Research Center, Hawthorne, NY {luog, wanlmn}@usbmcom Abstract Sleep stagng s the

More information

310 Int'l Conf. Par. and Dist. Proc. Tech. and Appl. PDPTA'16

310 Int'l Conf. Par. and Dist. Proc. Tech. and Appl. PDPTA'16 310 Int'l Conf. Par. and Dst. Proc. Tech. and Appl. PDPTA'16 Akra Sasatan and Hrosh Ish Graduate School of Informaton and Telecommuncaton Engneerng, Toka Unversty, Mnato, Tokyo, Japan Abstract The end-to-end

More information

Combined Temporal and Spatial Filter Structures for CDMA Systems

Combined Temporal and Spatial Filter Structures for CDMA Systems Combned Temporal and Spatal Flter Structures for CDMA Systems Ayln Yener WINLAB, Rutgers Unversty yener@wnlab.rutgers.edu Roy D. Yates WINLAB, Rutgers Unversty ryates@wnlab.rutgers.edu Sennur Ulukus AT&T

More information

Modeling the Survival of Retrospective Clinical Data from Prostate Cancer Patients in Komfo Anokye Teaching Hospital, Ghana

Modeling the Survival of Retrospective Clinical Data from Prostate Cancer Patients in Komfo Anokye Teaching Hospital, Ghana Internatonal Journal of Appled Scence and Technology Vol. 5, No. 6; December 2015 Modelng the Survval of Retrospectve Clncal Data from Prostate Cancer Patents n Komfo Anokye Teachng Hosptal, Ghana Asedu-Addo,

More information

Using the Perpendicular Distance to the Nearest Fracture as a Proxy for Conventional Fracture Spacing Measures

Using the Perpendicular Distance to the Nearest Fracture as a Proxy for Conventional Fracture Spacing Measures Usng the Perpendcular Dstance to the Nearest Fracture as a Proxy for Conventonal Fracture Spacng Measures Erc B. Nven and Clayton V. Deutsch Dscrete fracture network smulaton ams to reproduce dstrbutons

More information

Incorrect Beliefs. Overconfidence. Types of Overconfidence. Outline. Overprecision 4/22/2015. Econ 1820: Behavioral Economics Mark Dean Spring 2015

Incorrect Beliefs. Overconfidence. Types of Overconfidence. Outline. Overprecision 4/22/2015. Econ 1820: Behavioral Economics Mark Dean Spring 2015 Incorrect Belefs Overconfdence Econ 1820: Behavoral Economcs Mark Dean Sprng 2015 In objectve EU we assumed that everyone agreed on what the probabltes of dfferent events were In subjectve expected utlty

More information

ARTICLE IN PRESS Neuropsychologia xxx (2010) xxx xxx

ARTICLE IN PRESS Neuropsychologia xxx (2010) xxx xxx Neuropsychologa xxx (200) xxx xxx Contents lsts avalable at ScenceDrect Neuropsychologa journal homepage: www.elsever.com/locate/neuropsychologa Storage and bndng of object features n vsual workng memory

More information

Prediction of Total Pressure Drop in Stenotic Coronary Arteries with Their Geometric Parameters

Prediction of Total Pressure Drop in Stenotic Coronary Arteries with Their Geometric Parameters Tenth Internatonal Conference on Computatonal Flud Dynamcs (ICCFD10), Barcelona, Span, July 9-13, 2018 ICCFD10-227 Predcton of Total Pressure Drop n Stenotc Coronary Arteres wth Ther Geometrc Parameters

More information

FAST DETECTION OF MASSES IN MAMMOGRAMS WITH DIFFICULT CASE EXCLUSION

FAST DETECTION OF MASSES IN MAMMOGRAMS WITH DIFFICULT CASE EXCLUSION computng@tanet.edu.te.ua www.tanet.edu.te.ua/computng ISSN 727-6209 Internatonal Scentfc Journal of Computng FAST DETECTION OF MASSES IN MAMMOGRAMS WITH DIFFICULT CASE EXCLUSION Gábor Takács ), Béla Patak

More information

EXAMINATION OF THE DENSITY OF SEMEN AND ANALYSIS OF SPERM CELL MOVEMENT. 1. INTRODUCTION

EXAMINATION OF THE DENSITY OF SEMEN AND ANALYSIS OF SPERM CELL MOVEMENT. 1. INTRODUCTION JOURNAL OF MEDICAL INFORMATICS & TECHNOLOGIES Vol.3/00, ISSN 64-6037 Łukasz WITKOWSKI * mage enhancement, mage analyss, semen, sperm cell, cell moblty EXAMINATION OF THE DENSITY OF SEMEN AND ANALYSIS OF

More information

Balanced Query Methods for Improving OCR-Based Retrieval

Balanced Query Methods for Improving OCR-Based Retrieval Balanced Query Methods for Improvng OCR-Based Retreval Kareem Darwsh Electrcal and Computer Engneerng Dept. Unversty of Maryland, College Park College Park, MD 20742 kareem@glue.umd.edu Douglas W. Oard

More information

Integration of sensory information within touch and across modalities

Integration of sensory information within touch and across modalities Integraton of sensory nformaton wthn touch and across modaltes Marc O. Ernst, Jean-Perre Brescan, Knut Drewng & Henrch H. Bülthoff Max Planck Insttute for Bologcal Cybernetcs 72076 Tübngen, Germany marc.ernst@tuebngen.mpg.de

More information

The Limits of Individual Identification from Sample Allele Frequencies: Theory and Statistical Analysis

The Limits of Individual Identification from Sample Allele Frequencies: Theory and Statistical Analysis The Lmts of Indvdual Identfcaton from Sample Allele Frequences: Theory and Statstcal Analyss Peter M. Vsscher 1 *, Wllam G. Hll 2 1 Queensland Insttute of Medcal Research, Brsbane, Australa, 2 Insttute

More information

Optimal Planning of Charging Station for Phased Electric Vehicle *

Optimal Planning of Charging Station for Phased Electric Vehicle * Energy and Power Engneerng, 2013, 5, 1393-1397 do:10.4236/epe.2013.54b264 Publshed Onlne July 2013 (http://www.scrp.org/ournal/epe) Optmal Plannng of Chargng Staton for Phased Electrc Vehcle * Yang Gao,

More information

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and Ths artcle appeared n a journal publshed by Elsever. The attached copy s furnshed to the author for nternal non-commercal research and educaton use, ncludng for nstructon at the authors nsttuton and sharng

More information

Physical Model for the Evolution of the Genetic Code

Physical Model for the Evolution of the Genetic Code Physcal Model for the Evoluton of the Genetc Code Tatsuro Yamashta Osamu Narkyo Department of Physcs, Kyushu Unversty, Fukuoka 8-856, Japan Abstract We propose a physcal model to descrbe the mechansms

More information

INITIAL ANALYSIS OF AWS-OBSERVED TEMPERATURE

INITIAL ANALYSIS OF AWS-OBSERVED TEMPERATURE INITIAL ANALYSIS OF AWS-OBSERVED TEMPERATURE Wang Yng, Lu Xaonng, Ren Zhhua, Natonal Meteorologcal Informaton Center, Bejng, Chna Tel.:+86 684755, E-mal:cdcsjk@cma.gov.cn Abstract From, n Chna meteorologcal

More information

NUMERICAL COMPARISONS OF BIOASSAY METHODS IN ESTIMATING LC50 TIANHONG ZHOU

NUMERICAL COMPARISONS OF BIOASSAY METHODS IN ESTIMATING LC50 TIANHONG ZHOU NUMERICAL COMPARISONS OF BIOASSAY METHODS IN ESTIMATING LC50 by TIANHONG ZHOU B.S., Chna Agrcultural Unversty, 2003 M.S., Chna Agrcultural Unversty, 2006 A THESIS submtted n partal fulfllment of the requrements

More information

ALMALAUREA WORKING PAPERS no. 9

ALMALAUREA WORKING PAPERS no. 9 Snce 1994 Inter-Unversty Consortum Connectng Unverstes, the Labour Market and Professonals AlmaLaurea Workng Papers ISSN 2239-9453 ALMALAUREA WORKING PAPERS no. 9 September 211 Propensty Score Methods

More information

Appendix F: The Grant Impact for SBIR Mills

Appendix F: The Grant Impact for SBIR Mills Appendx F: The Grant Impact for SBIR Mlls Asmallsubsetofthefrmsnmydataapplymorethanonce.Ofthe7,436applcant frms, 71% appled only once, and a further 14% appled twce. Wthn my data, seven companes each submtted

More information

Modeling Multi Layer Feed-forward Neural. Network Model on the Influence of Hypertension. and Diabetes Mellitus on Family History of

Modeling Multi Layer Feed-forward Neural. Network Model on the Influence of Hypertension. and Diabetes Mellitus on Family History of Appled Mathematcal Scences, Vol. 7, 2013, no. 41, 2047-2053 HIKARI Ltd, www.m-hkar.com Modelng Mult Layer Feed-forward Neural Network Model on the Influence of Hypertenson and Dabetes Melltus on Famly

More information

Gene Selection Based on Mutual Information for the Classification of Multi-class Cancer

Gene Selection Based on Mutual Information for the Classification of Multi-class Cancer Gene Selecton Based on Mutual Informaton for the Classfcaton of Mult-class Cancer Sheng-Bo Guo,, Mchael R. Lyu 3, and Tat-Mng Lok 4 Department of Automaton, Unversty of Scence and Technology of Chna, Hefe,

More information

Chapter 20. Aggregation and calibration. Betina Dimaranan, Thomas Hertel, Robert McDougall

Chapter 20. Aggregation and calibration. Betina Dimaranan, Thomas Hertel, Robert McDougall Chapter 20 Aggregaton and calbraton Betna Dmaranan, Thomas Hertel, Robert McDougall In the prevous chapter we dscussed how the fnal verson 3 GTAP data base was assembled. Ths data base s extremely large.

More information

Estimation for Pavement Performance Curve based on Kyoto Model : A Case Study for Highway in the State of Sao Paulo

Estimation for Pavement Performance Curve based on Kyoto Model : A Case Study for Highway in the State of Sao Paulo Estmaton for Pavement Performance Curve based on Kyoto Model : A Case Study for Kazuya AOKI, PASCO CORPORATION, Yokohama, JAPAN, Emal : kakzo603@pasco.co.jp Octávo de Souza Campos, Publc Servces Regulatory

More information

A Geometric Approach To Fully Automatic Chromosome Segmentation

A Geometric Approach To Fully Automatic Chromosome Segmentation A Geometrc Approach To Fully Automatc Chromosome Segmentaton Shervn Mnaee ECE Department New York Unversty Brooklyn, New York, USA shervn.mnaee@nyu.edu Mehran Fotouh Computer Engneerng Department Sharf

More information

CLUSTERING is always popular in modern technology

CLUSTERING is always popular in modern technology Max-Entropy Feed-Forward Clusterng Neural Network Han Xao, Xaoyan Zhu arxv:1506.03623v1 [cs.lg] 11 Jun 2015 Abstract The outputs of non-lnear feed-forward neural network are postve, whch could be treated

More information

CONSTRUCTION OF STOCHASTIC MODEL FOR TIME TO DENGUE VIRUS TRANSMISSION WITH EXPONENTIAL DISTRIBUTION

CONSTRUCTION OF STOCHASTIC MODEL FOR TIME TO DENGUE VIRUS TRANSMISSION WITH EXPONENTIAL DISTRIBUTION Internatonal Journal of Pure and Appled Mathematcal Scences. ISSN 97-988 Volume, Number (7), pp. 3- Research Inda Publcatons http://www.rpublcaton.com ONSTRUTION OF STOHASTI MODEL FOR TIME TO DENGUE VIRUS

More information

EVALUATION OF BULK MODULUS AND RING DIAMETER OF SOME TELLURITE GLASS SYSTEMS

EVALUATION OF BULK MODULUS AND RING DIAMETER OF SOME TELLURITE GLASS SYSTEMS Chalcogende Letters Vol. 12, No. 2, February 2015, p. 67-74 EVALUATION OF BULK MODULUS AND RING DIAMETER OF SOME TELLURITE GLASS SYSTEMS R. EL-MALLAWANY a*, M.S. GAAFAR b, N. VEERAIAH c a Physcs Dept.,

More information

Fast Algorithm for Vectorcardiogram and Interbeat Intervals Analysis: Application for Premature Ventricular Contractions Classification

Fast Algorithm for Vectorcardiogram and Interbeat Intervals Analysis: Application for Premature Ventricular Contractions Classification Fast Algorthm for Vectorcardogram and Interbeat Intervals Analyss: Applcaton for Premature Ventrcular Contractons Classfcaton Irena Jekova, Vessela Krasteva Centre of Bomedcal Engneerng Prof. Ivan Daskalov

More information

Appendix for. Institutions and Behavior: Experimental Evidence on the Effects of Democracy

Appendix for. Institutions and Behavior: Experimental Evidence on the Effects of Democracy Appendx for Insttutons and Behavor: Expermental Evdence on the Effects of Democrac 1. Instructons 1.1 Orgnal sessons Welcome You are about to partcpate n a stud on decson-makng, and ou wll be pad for our

More information

Survival Rate of Patients of Ovarian Cancer: Rough Set Approach

Survival Rate of Patients of Ovarian Cancer: Rough Set Approach Internatonal OEN ACCESS Journal Of Modern Engneerng esearch (IJME) Survval ate of atents of Ovaran Cancer: ough Set Approach Kamn Agrawal 1, ragat Jan 1 Department of Appled Mathematcs, IET, Indore, Inda

More information

An Introduction to Modern Measurement Theory

An Introduction to Modern Measurement Theory An Introducton to Modern Measurement Theory Ths tutoral was wrtten as an ntroducton to the bascs of tem response theory (IRT) modelng and ts applcatons to health outcomes measurement for the Natonal Cancer

More information

The effect of salvage therapy on survival in a longitudinal study with treatment by indication

The effect of salvage therapy on survival in a longitudinal study with treatment by indication Research Artcle Receved 28 October 2009, Accepted 8 June 2010 Publshed onlne 30 August 2010 n Wley Onlne Lbrary (wleyonlnelbrary.com) DOI: 10.1002/sm.4017 The effect of salvage therapy on survval n a longtudnal

More information

Joint Modelling Approaches in diabetes research. Francisco Gude Clinical Epidemiology Unit, Hospital Clínico Universitario de Santiago

Joint Modelling Approaches in diabetes research. Francisco Gude Clinical Epidemiology Unit, Hospital Clínico Universitario de Santiago Jont Modellng Approaches n dabetes research Clncal Epdemology Unt, Hosptal Clínco Unverstaro de Santago Outlne 1 Dabetes 2 Our research 3 Some applcatons Dabetes melltus Is a serous lfe-long health condton

More information

Using a Wavelet Representation for Classification of Movement in Bed

Using a Wavelet Representation for Classification of Movement in Bed Usng a Wavelet Representaton for Classfcaton of Movement n Bed Adrana Morell Adam Depto. de Matemátca e Estatístca Unversdade de Caxas do Sul Caxas do Sul RS E-mal: amorell@ucs.br André Gustavo Adam Depto.

More information

Evaluation of the generalized gamma as a tool for treatment planning optimization

Evaluation of the generalized gamma as a tool for treatment planning optimization Internatonal Journal of Cancer Therapy and Oncology www.jcto.org Evaluaton of the generalzed gamma as a tool for treatment plannng optmzaton Emmanoul I Petrou 1,, Ganesh Narayanasamy 3, Eleftheros Lavdas

More information

A New Machine Learning Algorithm for Breast and Pectoral Muscle Segmentation

A New Machine Learning Algorithm for Breast and Pectoral Muscle Segmentation Avalable onlne www.ejaet.com European Journal of Advances n Engneerng and Technology, 2015, 2(1): 21-29 Research Artcle ISSN: 2394-658X A New Machne Learnng Algorthm for Breast and Pectoral Muscle Segmentaton

More information

Non-linear Multiple-Cue Judgment Tasks

Non-linear Multiple-Cue Judgment Tasks Non-lnear Multple-Cue Tasks Anna-Carn Olsson (anna-carn.olsson@psy.umu.se) Department of Psychology, Umeå Unversty SE-09 87, Umeå, Sweden Tommy Enqvst (tommy.enqvst@psyk.uu.se) Department of Psychology,

More information

NHS Outcomes Framework

NHS Outcomes Framework NHS Outcomes Framework Doman 1 Preventng people from dyng prematurely Indcator Specfcatons Verson: 1.21 Date: May 2018 Author: Clncal Indcators Team NHS Outcomes Framework: Doman 1 Preventng people from

More information

A Linear Regression Model to Detect User Emotion for Touch Input Interactive Systems

A Linear Regression Model to Detect User Emotion for Touch Input Interactive Systems 2015 Internatonal Conference on Affectve Computng and Intellgent Interacton (ACII) A Lnear Regresson Model to Detect User Emoton for Touch Input Interactve Systems Samt Bhattacharya Dept of Computer Scence

More information

A comparison of statistical methods in interrupted time series analysis to estimate an intervention effect

A comparison of statistical methods in interrupted time series analysis to estimate an intervention effect Peer revew stream A comparson of statstcal methods n nterrupted tme seres analyss to estmate an nterventon effect a,b, J.J.J., Walter c, S., Grzebeta a, R. & Olver b, J. a Transport and Road Safety, Unversty

More information

*VALLIAPPAN Raman 1, PUTRA Sumari 2 and MANDAVA Rajeswari 3. George town, Penang 11800, Malaysia. George town, Penang 11800, Malaysia

*VALLIAPPAN Raman 1, PUTRA Sumari 2 and MANDAVA Rajeswari 3. George town, Penang 11800, Malaysia. George town, Penang 11800, Malaysia 38 A Theoretcal Methodology and Prototype Implementaton for Detecton Segmentaton Classfcaton of Dgtal Mammogram Tumor by Machne Learnng and Problem Solvng *VALLIAPPA Raman, PUTRA Sumar 2 and MADAVA Rajeswar

More information

Computing and Using Reputations for Internet Ratings

Computing and Using Reputations for Internet Ratings Computng and Usng Reputatons for Internet Ratngs Mao Chen Department of Computer Scence Prnceton Unversty Prnceton, J 8 (69)-8-797 maoch@cs.prnceton.edu Jaswnder Pal Sngh Department of Computer Scence

More information

Estimation of Relative Survival Based on Cancer Registry Data

Estimation of Relative Survival Based on Cancer Registry Data Revew of Bonformatcs and Bometrcs (RBB) Volume 2 Issue 4, December 203 www.sepub.org/rbb Estmaton of Relatve Based on Cancer Regstry Data Olaf Schoffer *, Ante Nedostate 2, Stefane J. Klug,2 Cancer Epdemology,

More information

A MIXTURE OF EXPERTS FOR CATARACT DIAGNOSIS IN HOSPITAL SCREENING DATA

A MIXTURE OF EXPERTS FOR CATARACT DIAGNOSIS IN HOSPITAL SCREENING DATA Journal of Theoretcal and Appled Informaton Technology 2005 ongong JATIT & LLS ISSN: 1992-8645 www.jatt.org E-ISSN: 1817-3195 A MIXTURE OF EXPERTS FOR CATARACT DIAGNOSIS IN HOSPITAL SCREENING DATA 1 SUNGMIN

More information

Price linkages in value chains: methodology

Price linkages in value chains: methodology Prce lnkages n value chans: methodology Prof. Trond Bjorndal, CEMARE. Unversty of Portsmouth, UK. and Prof. José Fernández-Polanco Unversty of Cantabra, Span. FAO INFOSAMAK Tangers, Morocco 14 March 2012

More information

Prototypes in the Mist: The Early Epochs of Category Learning

Prototypes in the Mist: The Early Epochs of Category Learning Journal of Expermental Psychology: Learnng, Memory, and Cognton 1998, Vol. 24, No. 6, 1411-1436 Copyrght 1998 by the Amercan Psychologcal Assocaton, Inc. 0278-7393/98/S3.00 Prototypes n the Mst: The Early

More information

Biomarker Selection from Gene Expression Data for Tumour Categorization Using Bat Algorithm

Biomarker Selection from Gene Expression Data for Tumour Categorization Using Bat Algorithm Receved: March 20, 2017 401 Bomarker Selecton from Gene Expresson Data for Tumour Categorzaton Usng Bat Algorthm Gunavath Chellamuthu 1 *, Premalatha Kandasamy 2, Svasubramanan Kanagaraj 3 1 School of

More information

Introduction ORIGINAL RESEARCH

Introduction ORIGINAL RESEARCH ORIGINAL RESEARCH Assessng the Statstcal Sgnfcance of the Acheved Classfcaton Error of Classfers Constructed usng Serum Peptde Profles, and a Prescrpton for Random Samplng Repeated Studes for Massve Hgh-Throughput

More information

Investigation of zinc oxide thin film by spectroscopic ellipsometry

Investigation of zinc oxide thin film by spectroscopic ellipsometry VNU Journal of Scence, Mathematcs - Physcs 24 (2008) 16-23 Investgaton of znc oxde thn flm by spectroscopc ellpsometry Nguyen Nang Dnh 1, Tran Quang Trung 2, Le Khac Bnh 2, Nguyen Dang Khoa 2, Vo Th Ma

More information

AUTOMATED CHARACTERIZATION OF ESOPHAGEAL AND SEVERELY INJURED VOICES BY MEANS OF ACOUSTIC PARAMETERS

AUTOMATED CHARACTERIZATION OF ESOPHAGEAL AND SEVERELY INJURED VOICES BY MEANS OF ACOUSTIC PARAMETERS AUTOMATED CHARACTERIZATIO OF ESOPHAGEAL AD SEVERELY IJURED VOICES BY MEAS OF ACOUSTIC PARAMETERS B. García, I. Ruz, A. Méndez, J. Vcente, and M. Mendezona Department of Telecommuncaton, Unversty of Deusto

More information

Arithmetic Average: Sum of all precipitation values divided by the number of stations 1 n

Arithmetic Average: Sum of all precipitation values divided by the number of stations 1 n Char of ssgnment Suggested soluton PRCIPITTION Task (Charactersaton of the study area) rea: 43.3 km 2 Rver length: 0.296 km Hghest pont: 346 m a.s.l. Lowest pont (staton elevaton): 668 m a.s.l. Domnant

More information

Machine Understanding - a new area of research aimed at building thinking/understanding machines

Machine Understanding - a new area of research aimed at building thinking/understanding machines achne Understandng - a new area of research amed at buldng thnkng/understandng machnes Zbgnew Les and agdalena Les St. Queen Jadwga Research Insttute of Understandng, elbourne, Australa sqru@outlook.com

More information

AN ENHANCED GAGS BASED MTSVSL LEARNING TECHNIQUE FOR CANCER MOLECULAR PATTERN PREDICTION OF CANCER CLASSIFICATION

AN ENHANCED GAGS BASED MTSVSL LEARNING TECHNIQUE FOR CANCER MOLECULAR PATTERN PREDICTION OF CANCER CLASSIFICATION www.arpapress.com/volumes/vol8issue2/ijrras_8_2_02.pdf AN ENHANCED GAGS BASED MTSVSL LEARNING TECHNIQUE FOR CANCER MOLECULAR PATTERN PREDICTION OF CANCER CLASSIFICATION I. Jule 1 & E. Krubakaran 2 1 Department

More information

Perceptual image quality: Effects of tone characteristics

Perceptual image quality: Effects of tone characteristics Journal of Electronc Imagng 14(2), 023003 (Apr Jun 2005) Perceptual mage qualty: Effects of tone characterstcs Peter B. Delahunt Exponent Inc. 149 Commonwealth Drve Menlo Park, Calforna 94025 Xueme Zhang

More information

An expressive three-mode principal components model for gender recognition

An expressive three-mode principal components model for gender recognition Journal of Vson (4) 4, 36-377 http://journalofvson.org/4/5// 36 An expressve three-mode prncpal components model for gender recognton James W. Davs Hu Gao Department of Computer and Informaton Scence,

More information

Sparse Representation of HCP Grayordinate Data Reveals. Novel Functional Architecture of Cerebral Cortex

Sparse Representation of HCP Grayordinate Data Reveals. Novel Functional Architecture of Cerebral Cortex 1 Sparse Representaton of HCP Grayordnate Data Reveals Novel Functonal Archtecture of Cerebral Cortex X Jang 1, Xang L 1, Jngle Lv 2,1, Tuo Zhang 2,1, Shu Zhang 1, Le Guo 2, Tanmng Lu 1* 1 Cortcal Archtecture

More information

Research Article Statistical Segmentation of Regions of Interest on a Mammographic Image

Research Article Statistical Segmentation of Regions of Interest on a Mammographic Image Hndaw Publshng Corporaton EURASIP Journal on Advances n Sgnal Processng Volume 2007, Artcle ID 49482, 8 pages do:10.1155/2007/49482 Research Artcle Statstcal Segmentaton of Regons of Interest on a Mammographc

More information

A New Diagnosis Loseless Compression Method for Digital Mammography Based on Multiple Arbitrary Shape ROIs Coding Framework

A New Diagnosis Loseless Compression Method for Digital Mammography Based on Multiple Arbitrary Shape ROIs Coding Framework I.J.Modern Educaton and Computer Scence, 2011, 5, 33-39 Publshed Onlne August 2011 n MECS (http://www.mecs-press.org/) A New Dagnoss Loseless Compresson Method for Dgtal Mammography Based on Multple Arbtrary

More information

Research Article Statistical Analysis of Haralick Texture Features to Discriminate Lung Abnormalities

Research Article Statistical Analysis of Haralick Texture Features to Discriminate Lung Abnormalities Hndaw Publshng Corporaton Internatonal Journal of Bomedcal Imagng Volume 2015, Artcle ID 267807, 7 pages http://dx.do.org/10.1155/2015/267807 Research Artcle Statstcal Analyss of Haralck Texture Features

More information

INTEGRATIVE NETWORK ANALYSIS TO IDENTIFY ABERRANT PATHWAY NETWORKS IN OVARIAN CANCER

INTEGRATIVE NETWORK ANALYSIS TO IDENTIFY ABERRANT PATHWAY NETWORKS IN OVARIAN CANCER INTEGRATIVE NETWORK ANALYSIS TO IDENTIFY ABERRANT PATHWAY NETWORKS IN OVARIAN CANCER LI CHEN 1,2, JIANHUA XUAN 1,*, JINGHUA GU 1, YUE WANG 1, ZHEN ZHANG 2, TIAN LI WANG 2, IE MING SHIH 2 1The Bradley Department

More information

Reconstruction of gene regulatory network of colon cancer using information theoretic approach

Reconstruction of gene regulatory network of colon cancer using information theoretic approach Reconstructon of gene regulatory network of colon cancer usng nformaton theoretc approach Khald Raza #1, Rafat Parveen * # Department of Computer Scence Jama Mlla Islama (Central Unverst, New Delh-11005,

More information

Unobserved Heterogeneity and the Statistical Analysis of Highway Accident Data

Unobserved Heterogeneity and the Statistical Analysis of Highway Accident Data Unobserved Heterogenety and the Statstcal Analyss of Hghway Accdent Data Fred L. Mannerng Professor of Cvl and Envronmental Engneerng Courtesy Department of Economcs Unversty of South Florda 4202 E. Fowler

More information

Automatic Labelling and BI-RADS Characterisation of Mammogram Densities

Automatic Labelling and BI-RADS Characterisation of Mammogram Densities Lmted crculaton. For revew only. Automatc Labellng and BI-RADS Charactersaton of ammogram Denstes K. aras,. G. Lnguraru 3,. G. Ballester 4, S. Petroud,. Tsknaks and Sr. Brady Insttute of Computer Scence,

More information

A Classification Model for Imbalanced Medical Data based on PCA and Farther Distance based Synthetic Minority Oversampling Technique

A Classification Model for Imbalanced Medical Data based on PCA and Farther Distance based Synthetic Minority Oversampling Technique A Classfcaton Model for Imbalanced Medcal Data based on PCA and Farther Dstance based Synthetc Mnorty Oversamplng Technque NADIR MUSTAFA School of Computer Scence and Engneerng Unversty of Electronc Scence

More information

STAGE-STRUCTURED POPULATION DYNAMICS OF AEDES AEGYPTI

STAGE-STRUCTURED POPULATION DYNAMICS OF AEDES AEGYPTI Internatonal Conference Mathematcal and Computatonal Bology 211 Internatonal Journal of Modern Physcs: Conference Seres Vol. 9 (212) 364 372 World Scentfc Publshng Company DOI: 1.1142/S21194512543 STAGE-STRUCTURED

More information

Shape-based Retrieval of Heart Sounds for Disease Similarity Detection Tanveer Syeda-Mahmood, Fei Wang

Shape-based Retrieval of Heart Sounds for Disease Similarity Detection Tanveer Syeda-Mahmood, Fei Wang Shape-based Retreval of Heart Sounds for Dsease Smlarty Detecton Tanveer Syeda-Mahmood, Fe Wang 1 IBM Almaden Research Center, 650 Harry Road, San Jose, CA 95120. {stf,wangfe}@almaden.bm.com Abstract.

More information

We analyze the effect of tumor repopulation on optimal dose delivery in radiation therapy. We are primarily

We analyze the effect of tumor repopulation on optimal dose delivery in radiation therapy. We are primarily INFORMS Journal on Computng Vol. 27, No. 4, Fall 215, pp. 788 83 ISSN 191-9856 (prnt) ó ISSN 1526-5528 (onlne) http://dx.do.org/1.1287/joc.215.659 215 INFORMS Optmzaton of Radaton Therapy Fractonaton Schedules

More information

Richard Williams Notre Dame Sociology Meetings of the European Survey Research Association Ljubljana,

Richard Williams Notre Dame Sociology   Meetings of the European Survey Research Association Ljubljana, Rchard Wllams Notre Dame Socology rwllam@nd.edu http://www.nd.edu/~rwllam Meetngs of the European Survey Research Assocaton Ljubljana, Slovena July 19, 2013 Comparng Logt and Probt Coeffcents across groups

More information

HIV/AIDS-related Expectations and Risky Sexual Behavior in Malawi

HIV/AIDS-related Expectations and Risky Sexual Behavior in Malawi HIV/AIDS-related Expectatons and Rsky Sexual Behavor n Malaw Adelne Delavande Unversty of Essex and RAND Corporaton Hans-Peter Kohler Unversty of Pennsylvanna January 202 Abstract We use probablstc expectatons

More information

Optimal probability weights for estimating causal effects of time-varying treatments with marginal structural Cox models

Optimal probability weights for estimating causal effects of time-varying treatments with marginal structural Cox models Optmal probablty weghts for estmatng causal effects of tme-varyng treatments wth margnal structural Cox models Mchele Santacatterna, Cela García-Pareja Rno Bellocco, Anders Sönnerborg, Anna Ma Ekström

More information

WHO S ASSESSMENT OF HEALTH CARE INDUSTRY PERFORMANCE: RATING THE RANKINGS

WHO S ASSESSMENT OF HEALTH CARE INDUSTRY PERFORMANCE: RATING THE RANKINGS WHO S ASSESSMENT OF HEALTH CARE INDUSTRY PERFORMANCE: RATING THE RANKINGS ELLIOTT PARKER and JEANNE WENDEL * Department of Economcs, Unversty of Nevada, Reno, NV, USA SUMMARY Ths paper examnes the econometrc

More information

Semantics and image content integration for pulmonary nodule interpretation in thoracic computed tomography

Semantics and image content integration for pulmonary nodule interpretation in thoracic computed tomography Semantcs and mage content ntegraton for pulmonary nodule nterpretaton n thoracc computed tomography Danela S. Racu a, Ekarn Varutbangkul a, Jane G. Csneros a, Jacob D. Furst a, Davd S. Channn b, Samuel

More information

The Influence of the Isomerization Reactions on the Soybean Oil Hydrogenation Process

The Influence of the Isomerization Reactions on the Soybean Oil Hydrogenation Process Unversty of Belgrade From the SelectedWorks of Zeljko D Cupc 2000 The Influence of the Isomerzaton Reactons on the Soybean Ol Hydrogenaton Process Zeljko D Cupc, Insttute of Chemstry, Technology and Metallurgy

More information

IMPROVING THE EFFICIENCY OF BIOMARKER IDENTIFICATION USING BIOLOGICAL KNOWLEDGE

IMPROVING THE EFFICIENCY OF BIOMARKER IDENTIFICATION USING BIOLOGICAL KNOWLEDGE IMPROVING THE EFFICIENCY OF BIOMARKER IDENTIFICATION USING BIOLOGICAL KNOWLEDGE JOHN H. PHAN The Wallace H. Coulter Department of Bomedcal Engneerng, Georga Insttute of Technology, 313 Ferst Drve Atlanta,

More information

HIV/AIDS-related Expectations and Risky Sexual Behavior in Malawi

HIV/AIDS-related Expectations and Risky Sexual Behavior in Malawi Unversty of Pennsylvana ScholarlyCommons PSC Workng Paper Seres 7-29-20 HIV/AIDS-related Expectatons and Rsky Sexual Behavor n Malaw Adelne Delavande RAND Corporaton, Nova School of Busness and Economcs

More information

Resampling Methods for the Area Under the ROC Curve

Resampling Methods for the Area Under the ROC Curve Resamplng ethods for the Area Under the ROC Curve Andry I. Bandos AB6@PITT.EDU Howard E. Rockette HERBST@PITT.EDU Department of Bostatstcs, Graduate School of Publc Health, Unversty of Pttsburgh, Pttsburgh,

More information

A Computer-aided System for Discriminating Normal from Cancerous Regions in IHC Liver Cancer Tissue Images Using K-means Clustering*

A Computer-aided System for Discriminating Normal from Cancerous Regions in IHC Liver Cancer Tissue Images Using K-means Clustering* A Computer-aded System for Dscrmnatng Normal from Cancerous Regons n IHC Lver Cancer Tssue Images Usng K-means Clusterng* R. M. CHEN 1, Y. J. WU, S. R. JHUANG, M. H. HSIEH, C. L. KUO, Y. L. MA Department

More information

A GEOGRAPHICAL AND STATISTICAL ANALYSIS OF LEUKEMIA DEATHS RELATING TO NUCLEAR POWER PLANTS. Whitney Thompson, Sarah McGinnis, Darius McDaniel,

A GEOGRAPHICAL AND STATISTICAL ANALYSIS OF LEUKEMIA DEATHS RELATING TO NUCLEAR POWER PLANTS. Whitney Thompson, Sarah McGinnis, Darius McDaniel, A GEOGRAPHICAL AD STATISTICAL AALYSIS OF LEUKEMIA DEATHS RELATIG TO UCLEAR POWER PLATS Whtney Thompson, Sarah McGnns, Darus McDanel, Jean Sexton, Rebecca Pettt, Sarah Anderson, Monca Jackson ABSTRACT:

More information

AUTOMATED DETECTION OF HARD EXUDATES IN FUNDUS IMAGES USING IMPROVED OTSU THRESHOLDING AND SVM

AUTOMATED DETECTION OF HARD EXUDATES IN FUNDUS IMAGES USING IMPROVED OTSU THRESHOLDING AND SVM AUTOMATED DETECTION OF HARD EXUDATES IN FUNDUS IMAGES USING IMPROVED OTSU THRESHOLDING AND SVM Wewe Gao 1 and Jng Zuo 2 1 College of Mechancal Engneerng, Shangha Unversty of Engneerng Scence, Shangha,

More information

Encoding processes, in memory scanning tasks

Encoding processes, in memory scanning tasks vlemory & Cognton 1976,4 (5), 501 506 Encodng processes, n memory scannng tasks JEFFREY O. MILLER and ROBERT G. PACHELLA Unversty of Mchgan, Ann Arbor, Mchgan 48101, Three experments are presented that

More information

Desperation or Desire? The Role of Risk Aversion in Marriage. Christy Spivey, Ph.D. * forthcoming, Economic Inquiry. Abstract

Desperation or Desire? The Role of Risk Aversion in Marriage. Christy Spivey, Ph.D. * forthcoming, Economic Inquiry. Abstract Desperaton or Desre? The Role of Rsk Averson n Marrage Chrsty Spvey, Ph.D. * forthcomng, Economc Inury Abstract Because of the uncertanty nherent n searchng for a spouse and the uncertanty of the future

More information

Active Affective State Detection and User Assistance with Dynamic Bayesian Networks. Xiangyang Li, Qiang Ji

Active Affective State Detection and User Assistance with Dynamic Bayesian Networks. Xiangyang Li, Qiang Ji Actve Affectve State Detecton and User Assstance wth Dynamc Bayesan Networks Xangyang L, Qang J Electrcal, Computer, and Systems Engneerng Department Rensselaer Polytechnc Insttute, 110 8th Street, Troy,

More information

Insights in Genetics and Genomics

Insights in Genetics and Genomics Insghts n Genetcs and Genomcs Research Artcle Open Access New Score Tests for Equalty of Varances n the Applcaton of DNA Methylaton Data Analyss [Verson ] Welang Qu Xuan L Jarrett Morrow Dawn L DeMeo Scott

More information

Proceedings of the 6th WSEAS Int. Conf. on EVOLUTIONARY COMPUTING, Lisbon, Portugal, June 16-18, 2005 (pp )

Proceedings of the 6th WSEAS Int. Conf. on EVOLUTIONARY COMPUTING, Lisbon, Portugal, June 16-18, 2005 (pp ) Proceedngs of the 6th WSEAS Int. Conf. on EVOLUTIONARY COMPUTING, Lsbon, Portugal, June 6-8, 2005 (pp285-20) Novel Intellgent Edge Detector for Sonographcal Images Al Rafee *, Mohammad Hasan Morad **,

More information

Estimating the distribution of the window period for recent HIV infections: A comparison of statistical methods

Estimating the distribution of the window period for recent HIV infections: A comparison of statistical methods Research Artcle Receved 30 September 2009, Accepted 15 March 2010 Publshed onlne n Wley Onlne Lbrary (wleyonlnelbrary.com) DOI: 10.1002/sm.3941 Estmatng the dstrbuton of the wndow perod for recent HIV

More information

A-UNIFAC Modeling of Binary and Multicomponent Phase Equilibria of Fatty Esters+Water+Methanol+Glycerol

A-UNIFAC Modeling of Binary and Multicomponent Phase Equilibria of Fatty Esters+Water+Methanol+Glycerol -UNIFC Modelng of Bnary and Multcomponent Phase Equlbra of Fatty Esters+Water+Methanol+Glycerol N. Garrdo a, O. Ferrera b, R. Lugo c, J.-C. de Hemptnne c, M. E. Macedo a, S.B. Bottn d,* a Department of

More information

STOCHASTIC MODELS OF PITCH JITTER A D AMPLITUDE SHIMMER FOR VOICE MODIFICATIO

STOCHASTIC MODELS OF PITCH JITTER A D AMPLITUDE SHIMMER FOR VOICE MODIFICATIO STOCHASTIC MODELS OF PITCH JITTER A D AMPLITUDE SHIMMER FOR VOICE MODIFICATIO Dma Runsky,2 and Yzhar Lavner Department of Computer Scence, Tel-Ha Academc College, Israel 2 Israel Development Center, Intel

More information

A Novel artifact for evaluating accuracies of gear profile and pitch measurements of gear measuring instruments

A Novel artifact for evaluating accuracies of gear profile and pitch measurements of gear measuring instruments A Novel artfact for evaluatng accuraces of gear profle and ptch measurements of gear measurng nstruments Sonko Osawa, Osamu Sato, Yohan Kondo, Toshyuk Takatsuj (NMIJ/AIST) Masaharu Komor (Kyoto Unversty)

More information

An Approach to Discover Dependencies between Service Operations*

An Approach to Discover Dependencies between Service Operations* 36 JOURNAL OF SOFTWARE VOL. 3 NO. 9 DECEMBER 2008 An Approach to Dscover Dependences between Servce Operatons* Shuyng Yan Research Center for Grd and Servce Computng Insttute of Computng Technology Chnese

More information

Journal of Engineering Science and Technology Review 11 (2) (2018) Research Article

Journal of Engineering Science and Technology Review 11 (2) (2018) Research Article Jestr Journal of Engneerng Scence and Technology Revew 11 (2) (2018) 8-12 Research Artcle Detecton Lung Cancer Usng Gray Level Co-Occurrence Matrx (GLCM) and Back Propagaton Neural Network Classfcaton

More information

THE NORMAL DISTRIBUTION AND Z-SCORES COMMON CORE ALGEBRA II

THE NORMAL DISTRIBUTION AND Z-SCORES COMMON CORE ALGEBRA II Name: Date: THE NORMAL DISTRIBUTION AND Z-SCORES COMMON CORE ALGEBRA II The normal dstrbuton can be used n ncrements other than half-standard devatons. In fact, we can use ether our calculators or tables

More information

The High way code. the guide to safer, more enjoyable drug use [GHB] Who developed it?

The High way code. the guide to safer, more enjoyable drug use [GHB] Who developed it? The Hgh way code the gude to safer, more enjoyable drug use [] Who developed t? What s t? The frst gude to safer drug use voted for by people who take drugs. How was t was developed? GDS asked loads of

More information

Project title: Mathematical Models of Fish Populations in Marine Reserves

Project title: Mathematical Models of Fish Populations in Marine Reserves Applcaton for Fundng (Malaspna Research Fund) Date: November 0, 2005 Project ttle: Mathematcal Models of Fsh Populatons n Marne Reserves Dr. Lev V. Idels Unversty College Professor Mathematcs Department

More information