Adaptive Fuzzy Color Interpolation
|
|
- Augustine Washington
- 5 years ago
- Views:
Transcription
1 Joural of Electroic Imagig Vol. (3), July 00. Adaptive Fuzzy Color Iterpolatio Pig-Sig Tsai, Tiku Acharya,, Ajay K. Ray 3 Itel Corporatio Desktop Architecture Lab 5000 West Chadler Boulevard Mail Stop: CH7- Chadler, Arizoa 856, USA Departmet of Electrical Egieerig Arizoa State Uiversity Tempe, Arizoa 8587, USA 3 Departmet of Electroics ad Electrical Commuicatio Egieerig Idia Istitute of Techology Kharagpur, Idia pstsai@ieee.org
2 Joural of Electroic Imagig Vol. (3), July 00. Abstract I a electroic color imagig device, such as a digital camera, usig a sigle CCD or CMOS sesor, the color iformatio is usually acquired i sub-sampled patters of Red (R), Gree (G) ad Blue (B) pixels. Full resolutio color is subsequetly geerated from this sub-sampled image. This is popularly called Color Iterpolatio or Color Demosaicig. I this paper, we preset a color iterpolatio algorithm usig a method of fuzzy membership assigmet alog with the cocept of smooth hue trasitio. The algorithm is adaptive i ature ad produces superior quality full resolutio color images compared to most of the popularly kow color iterpolatio algorithms i the literature. We preset the results of compariso with some challegig subsampled images for color iterpolatio. Keywords: Color Iterpolatio, Color Demosaicig, Fuzzy Membership, Hue, Digital Imagig. Itroductio Due to the cost ad packagig cosideratio, i digital imagig devices such as a digital camera, the image color is captured i a sub-sampled patter. Typically the raw image is captured with each "pixel'' locatio composed of oly oe of the three primary color compoets, R (Red), G (Gree), or B (Blue). This sub-sampled color image is geerated usig certai patter of a Color Filter Array (CFA). This CFA is realized by coatig the surface of the electroic sesor array usig some optical material that acts as bad-pass filter. This coatig o each pixel locatio of the sesor array permits the photos correspodig to oly oe color compoet (frequecy rage) to be trasmitted to the sesor ad the other two color compoets are blocked. A typical ad widely used CFA patter is called Bayer Patter [], as show i Figure. It should be oted that there are may other patters kow i the literature. However, we oly cosider the Bayer patter i this paper. m G R G R G B G B G B 3 G R G R G 4 B G B G B 5 G R G R G Figure : Bayer CFA patter A perfect, rather efficiet full color represetatio of a image eeds the iformatio of all the three colors i each pixel locatio. As a result, each pixel eeds to be represeted by 4 bit color, assumig 8 bits for each of R, G ad B. To achieve this, it is essetial to iterpolate the missig two colors i each pixel locatio usig the iformatio of the eighborig pixels. The methodology to recover or iterpolate these missig colors is popularly kow as Color Iterpolatio or Color Demosaicig. We used the termiology Color Iterpolatio i this paper, istead of Color Demosaicig. I this paper, we preset a brief review of the color iterpolatio algorithms existig i the literature i sectio. We proposed a ew color iterpolatio algorithm usig the cocept of Fuzzy membership assigmet based scheme ad smooth hue trasitio i sectio 3. We preset the experimetal results ad compariso of our
3 Joural of Electroic Imagig Vol. (3), July 00. proposed techique with other popular techiques i sectio 4. We coclude this paper i sectio 5.. Related Works I the past decade, may color iterpolatio algorithms have bee proposed i the literature. This area of research ad developmet is gettig more ad more attetio, thaks to the boomig market of digital imagig devices. These color iterpolatio algorithms ca be broadly classified ito o-adaptive algorithms ad adaptive algorithms, as described below. () No-adaptive algorithms: I o-adaptive color iterpolatio algorithms, a fixed patter of computatio is applied i every pixel locatio i the sub-sampled color image i order to recover the missig two color compoets. Usually, this type of algorithms is easy to implemet with low cost i terms of computatioal requiremets. () Adaptive algorithms: I adaptive color iterpolatio algorithms, itelliget processig is applied i every pixel locatio based o the characteristics of the image i order to recover the missig color compoets. This type of algorithms yield better results i terms of quality as compared with the o-adaptive algorithms. However, effective algorithms i this category are usually more computatioally itesive. We review some algorithms from both categories i order to itroduce some flavor ad characteristics of the color iterpolatio methodologies proposed i the literature. However, there are other algorithms that we did ot iclude i this paper, such as Cubic Covolutio type methods [, 3], Patter Recogitio Iterpolatio method [4], Patter Classificatio method [5], gradiet based method [6], ad so o. I this paper, we oly give a review to some basic methods that are related to the thought process of our proposed ew algorithm. Iterested readers are also suggested to visit the web-site at Staford Uiversity [7].. No-adaptive algorithms Nearest Neighbor Replicatio: I this simple color iterpolatio method [8, 9], each missig color i a pixel is allocated by the value of the earest pixel of the same color i the iput image. The earest eighbor ca be ay oe of the upper, lower, left ad right pixels. As discussed by Adams [8], the oly advatage of this approach is that computatioal requiremet is very small ad suitable for applicatios where speed is very crucial. However, the sigificat color errors make it uacceptable for still imagig syste such as high-resolutio digital cameras. Biliear Iterpolatio: Istead of replicatig the earest eighbors, biliear iterpolatio [8] allocates the missig color compoet with the liear average of the adjacet pixels with same color compoet. For example, the pixel locatio (, 3) i Figure cotais BLUE compoet oly. Hece the missig GREEN compoet ca be estimated as average of the left, right, top ad bottom GREEN pixel values. The missig RED compoet ca be estimated as average of the four diagoally adjacet corer eighbors cotaiig RED pixels. This method is very simple ad ca be easily implemeted. However, experimetal results show that ew kid of pixel artifacts, e.g. zipper effect, is itroduced i the eighborhood of the iterpolated pixels i the iterpolated full color image. This artifact may be acceptable i a video applicatio because the artifact may ot be visible by the huma eye due to effect of motio blur betwee video frames, but these artifacts are ot acceptable for a still imagig system. 3
4 Joural of Electroic Imagig Vol. (3), July 00. Media Iterpolatio: The media iterpolatio [0] allocates the missig color compoet with the media value of the adjacet pixels of same color compoet, as opposed to the liear average used i biliear iterpolatio. This provides a slightly better result i terms of visual quality as compared with the biliear iterpolatio. However, the resultat images are still blurry for images with high frequecy cotets, ad for high resolutio still imagig syste this is still ot acceptable. Smooth Hue Trasitio Iterpolatio: The key problem of the color artifacts i both biliear ad media iterpolatio is that the hue values of adjacet pixels chage suddely (o-smoothly). O the other had, the Bayer CFA patter ca be cosidered as combiatio of a lumiace chael (gree pixels) ad two chromiace chaels (red ad blue pixels). The smooth hue trasitio iterpolatio method [] iterpolates these chaels differetly. The missig GREEN compoet i every RED ad BLUE pixel locatios i the Bayer patter ca first be iterpolated usig biliear iterpolatio as discussed before. The idea of chromiace chael iterpolatio is to impose a smooth trasitio i hue value from pixel to pixel. I order to do so, it defies blue "hue value" as B/G, ad red "hue value" as R/G. For iterpolatio of the missig blue pixel values B m, i pixel locatio ( ) i the Bayer patter, the followig three differet cases may arise. Case : the pixel locatio ( ) cotais GREEN color compoet oly ad the adjacet left ad right pixel locatios cotai BLUE color compoet oly. For example, the pixel locatio (, ) i Figure cotais GREEN compoet oly ad its adjacet pixels i left ad right cotai oly BLUE iformatio. The BLUE iformatio i locatio ( ) ca be estimated as follows: B B,, * = G m G G B. m Case : the pixel locatio ( ) cotais GREEN color compoet oly ad the adjacet top ad bottom pixel locatios cotai BLUE color compoet oly. The pixel at locatio (3, 3) i Figure is such a example. The BLUE iformatio i locatio ( ) ca be estimated as follows: Bm, Bm,,, * = G m Gm, Gm, B. m Case 3: the pixel locatio ( ) cotais RED color compoet oly. Obviously, four diagoally eighborig corer pixels cotai BLUE color oly. For example, the pixel locatio (3, ) i Figure cotais RED color compoet oly. The BLUE iformatio i locatio ( ) ca be estimated as follows: Bm, Bm, Bm, Bm,,, * = G m Gm, Gm, Gm, Gm, B. m The iterpolatio of missig RED pixel values ca be carried out similarity. As metioed by Adams [8], depedig o where the iterpolatio step happes i the image processig chai the defiitio of "hue value" chages. For example, if the pixel value is trasformed ito logarithmic exposure space from liear space before iterpolatio, istead of B/G or R/G, oe ca ow defie the "hue value" as B-G or R-G. 4 4
5 Joural of Electroic Imagig Vol. (3), July 00. This is comig from the fact that log(x/y) = log(x) log(y) = X' Y'. Here X ad Y are the logarithmic values of X ad Y respectively. Sice the liear/oliear trasformatio ca be doe usig a simple table look-up ad all the divisio for calculatig hue value is replaced by subtractio, this helps reduce computatioal complexity for implemetatio.. Adaptive algorithms Patter Matchig based Iterpolatio Algorithm: I the Bayer patter, a BLUE or RED pixel has four eighborig GREEN pixels. A simple patter matchig techique for recostructig the missig color compoets based o the pixel cotexts was proposed by Wu, et. al. []. This patter matchig algorithm defies a gree patter for the pixel at locatio ( ) cotaiig a o-green color compoet as a four-dimesioal iteger-valued vector: g( ) = ( Gm,, Gm,, G, G ). The similarity (or differece) betwee two gree patters g ad g is defied as the vector -orm g g = g g. 0 i< 4 Whe the differece betwee two gree patters is small, it is likely that the two pixel locatios where the two gree patters are defied will have similar RED ad BLUE color compoets. A weighted average proportioal to degree of similarity of the gree patters is used to calculate the missig color compoet based o the gree patter cotexts. For example, the missig BLUE color value B m, i pixel locatio ( ) cotais oly RED color compoet is estimated by comparig the gree patter g ( ) with the four eighborig gree patters g ( m, ), g ( m, ), g ( m, ) ad g ( m, ). If all the differeces betwee g ( ) ad other four gree patters are uiformly small, the a simple average is used to estimated the missig BLUE color compoet, Bm, Bm, Bm, Bm, B =. 4 Otherwise, whe the largest differece is above certai threshold, oly the top two bestmatched gree patters iformatio are used. If g ( ) g( m, ) ad g ( ) g( m, ) are the two smallest differeces, the the missig BLUE color is estimated as follows. g( ) g( m, ) Bm, g( ) g( m, ) Bm, Bm, = g( ) g( m, ) g( ) g( m, ) The missig RED color values ca be carried out aalogously. This algorithm is simple ad efficiet. However, as poited out by Wu, et. al. [], the quality of recostructed images is still udesirable. Block Matchig Based Algorithm: Similar to the patter matchig algorithm described previously, the block matchig algorithm by Acharya, et. al. [3] defies the Color Block of a o-gree pixel as the vector formed by the four eighborig gree pixels. However, istead of usig vector orm to measure the similarity, a ew metrics Color i i 5
6 Joural of Electroic Imagig Vol. (3), July 00. Gravity is defied as the mea value of the four vector compoets for measurig the similarity betwee color blocks. Let x = ( x, x, x3, x4) be a vector of a Color Block, x x x3 x4 the the Color Gravity for this color block will be x = ( ). The 4 similarity betwee two color blocks is defied as the absolute differece of the two color gravity values. Also istead of usig the weighted average proportioal to the similarity of the gree patters for estimatig the missig color values, the block matchig algorithm is developed based o the selectio of a eighborig block whose Color Gravity is closest to the Color Gravity of the color block uder cosideratio. For a o-gree pixel i the Bayer patter image, there are four eighborig gree pixels G (the North eighbor), Gs (South), Ge (East), ad Gw (West) which form the Color Block g = ( G, Gs, Ge, Gw) whose Color Gravity is g. The missig GREEN value G is simply computed by the media of the four eighborig gree pixel value. If the pixel locatio cotais BLUE color value, it will have four diagoally RED pixels Re (North-East), Rse (South-East), Rsw (South-West), ad Rw (North-West) whose color blocks are g e, g se, g sw, ad g w. The correspodig color gravity are g e, g se, g sw, ad g w respectively. The missig RED value R is replaced usig oe of the four diagoally red pixels based o best match of their color gravity values. The best match or miimal differece mi is the miimum amogst the four absolute differeces = g ge, = g g se, 3 = g g sw, ad 4 = g gw respectively. Similarly, we ca estimate the missig BLUE color value, if the pixel locatio cotais RED color value due to the symmetry of red ad blue samplig positio i a Bayer patter image. For the gree pixel locatio, oly two color blocks (either up-bottom or left-right positios) are cosidered for the missig RED or BLUE color. Similar operatio ca be carried out aalogously. The algorithm ca be described as follows: Begi if the pixel locatio is ot GREEN the { G media{g, Gs, Ge, Gw }; compute = g ge, = g g se, 3 = g g sw, ad 4 = g gw ; Fid mi = mi{,, 3, 4} ; If ( mi = ) the R Re if Red is missig, B Be if Blue is missig; If ( = mi ) the R Rse if Red is missig, B Bse if Blue is missig; If ( = mi 3 ) the R Rsw if Red is missig, B Bsw if Blue is missig; If ( = mi 4 ) the R Rw if Red is missig, B Bw if Blue is missig; }; if the pixel locatio is GREEN the { compute u = G gu, b = G gb, l = G gl ad r = G gr ; If ( u < b ) the B Bu else B Bb ; 6
7 Joural of Electroic Imagig Vol. (3), July 00. } If ( l < r ) the R Rl else R Rr ; Ed. This method provides a much sharper image as compared with the simple media or biliear iterpolatio type methods ad the simple patter matchig method. However, sice this method does ot cosider smooth hue trasitio, the color bleedig artifacts may still be a problem for some images such as image of a Zebra. Edge Sesig Iterpolatio: Depedig o lumiace gradiets, differet predictors are used for the missig GREEN values i the edge sesig iterpolatio method [8, 4]. First, two gradiets are defied, oe i horizotal directio, the other i vertical directio, for each RED or BLUE oly pixel locatio. For istace, cosider the pixel "b8" as show i Figure. We defie two gradiets as H = g7 g9 ad V = g3 g3, where x deotes absolute value of x. Based o these gradiet values ad a certai threshold (T ), the iterpolatio algorithm the ca be described as follows if H < T ad V > T the G8 = (g7g9)/; else if H > T ad V < T the G8 = (g3g3)/; else G8 = (g3g7g9g3)/4; edif edif g r g3 r4 g5 b6 g7 b8 g9 b0 g r g3 r4 g5 b6 g7 b8 g9 b0 g r g3 r4 g5 Figure A slightly differet edge sesig iterpolatio algorithm is described i [5]. Istead of lumiace gradiets, chromiace gradiets are used. The two gradiets, refer to Figure 3 below, are defied as: b3 b7 H = b5, b b9 V = b5. 7
8 Joural of Electroic Imagig Vol. (3), July 00. b g b3 g4 b5 g6 b7 g8 b9 Figure 3 Liear Iterpolatio with Laplacia secod-order Correctio terms: This algorithm [6] is desiged for optimizig performace i terms of the visual quality of the iterpolated image whe applied o images with sharp edges. We illustrate this algorithm usig a example. Missig color compoets are estimated by followig steps. The first step i this algorithm is to estimate the missig GREEN color compoets at the pixel locatios cotaiig RED or BLUE color compoet oly. Let us cosider estimatig the gree value at a blue pixel locatio (usig Figure 3) as a example. Iterpolatio at a red pixel locatio ca be doe i the similar fashio. Let's fid out the missig GREEN compoet (G5) at pixel locatio b5. We defie horizotal ad vertical gradiets i this pixel locatio as follows: H = g4 g6 ( b5 b3) ( b7 b5), V = g g8 ( b5 b) ( b9 b5). ad V Ituitively, we ca cosider H above as combiatio of the lumiace gradiet ad the chromiace gradiet as described i edge sesig iterpolatio algorithm i the previous sectio. I the expressio of H above, as a example, the first term g4 g6 is the first-order differece of the eighborig gree values, cosidered to be the lumiace gradiet ad the secod term ( b5 b3) ( b7 b5) is the secod-order derivative of the eighborig blue values, cosidered as the chromiace gradiet. Usig these two gradiet values, the missig gree compoet G5 at pixel locatio b5 is estimated as follows. if H < V the G5 = (g4 g6)/ (-b3 *b5 - b7)/4; else if H > V the G5 = (gg8)/ (-b *b5 - b9)/4; else G5 = (gg4g6g8)/4 (-b - b3 4*b5 - b7 - b9)/8; edif edif The iterpolatio step for G5 has two parts. The first part is the liear average of the eighborig gree values, ad the secod part ca be cosidered as a secod-order correctio term based o the eighborig blue (red) values. 8
9 Joural of Electroic Imagig Vol. (3), July 00. The missig red (or blue) color compoets are estimated i every pixel locatio after estimatio of the missig gree compoets i every pixel locatio. Depedig o the positio, refer to Figure 4 below, we have three cases: r g r3 g4 b5 g6 r7 g8 r9 Figure 4. Estimate red (blue) color compoet at a gree pixel where earest eighbors of red (blue) pixels are i the same colum, e.g. pixel locatio g4 as show i Figure 4 above. We estimate the red compoet R4 at pixel locatio g4 as follows. R4 = (r r7) / (g4 - G g4 - G7) / 4. Estimate red (blue) color compoet at a gree pixel where earest eighbors of red (blue) pixels are i the same row, e.g. pixel locatio g as show i Figure 4. We estimate the red compoet R at pixel locatio g as follows. R = (r r3) / (g - G g - G3) / 4 3. Estimate red (blue) color compoet at a blue (red) pixel. For istace, estimate red compoet R5 at pixel locatio b5 as show i Figure 4. Here we first defie two diagoal gradiets as follows: N = r r9 G5 g G5 g9, P = r3 r7 G5 g3 G5 g7, Usig these diagoal gradiets, the algorithm for estimatig the missig color compoets is described as: if N < P the R5 = (rr9)/ (-G *G5 G9)/; else if H > V the R5 = (r3r7)/ (-G3 *G5 G7)/; else R5 = (rr3r7r9)/4 (-G - G3 4*G5 - G7 - G9)/4; edif edif. This method provides much better visual quality of the recostructed image cotaiig a lot of sharp edges. However, the secod-order derivative for calculatig the gradiets makes the algorithm quite sesitive to oise. Sice oly the color iformatio i the same directio (vertical, horizotal, or oe of the diagoal directios based o the gradiet iformatio) is used for iterpolatio, we believe that it is still possible to further improve the visual quality of the recostructed image. 9
10 Joural of Electroic Imagig Vol. (3), July Proposed Fuzzy Assigmet Based Adaptive Method 3. Fuzzy membership assigmet strategy I our proposed approach, we have assiged a fuzzy membership [7, 8, 9] to all the surroudig four coected eighborig pixels ( G, G, G3, G4 as show i Figure 5, as a example) i order to estimate the missig color compoet (say gree) i a pixel locatio (R i Figure 5). The membership assigmet strategy differs depedig upo characteristic of a possible edge at the pixel locatio. For example, let us cosider estimatio of the missig gree color compoet at a pixel locatio, say cotaiig R as show i Figure 5. B G 3 B G R G B G 4 B Figure 5 Depedig upo the correlatio amogst the surroudig pixels, we have formulated a strategy to assig membership grades to the surroudig horizotal ad vertical pixels. The membership grades have bee experimetally derived through exhaustive subjective visual ispectio, takig ito cosideratio the exhaustive set of images havig possible edges i the horizotal ad vertical directios. We have cosidered the followig four cases, where there is a possible edge alog the horizotal directio. I a likewise maer we have also cosidered the cases where there are possible edges i the vertical directio. Case : G G is small while 3 G4 G is arbitrarily large, subject to the coditio that G 3 G4 G G >> 0. Here we have assumed the existece of a horizotal edge while the horizotal eighborig pixels G ad G have approximately the same itesity. Case : G G is small ad 3 G4 G is arbitrary ad G G G4. I this case also there is clearly a possible edge at the pixel locatio R, ad the itesity of this edge depeds upo the surroudig pixel values G 3 ad G 4. Case 3: This case is similar to case with a differece that here the pixels G, G ad G 3 to be approximately of similar pixel itesity, i.e. G G is small ad G3 G4 is arbitrary ad G G. G3 0
11 Joural of Electroic Imagig Vol. (3), July 00. Case 4: I this case we have cosidered that all the four coectig eighborig pixels G, G, G 3 ad G 4 that are all differet subject to the coditio that G 3 G4 G G >> 0. I each of the above cases we have computed the horizotal ad vertical membership fuctios ad a similar logic has bee applied to the images havig a edge i the vertical directio. While computig the membership assigmets, we have estimated the likelihood of the occurreces of each of the four cases as described above usig a large umber of images. We have also observed that the assumptio of equal a- priori probabilities of occurrece of differet types of edges, viz., horizotal ad vertical edges is ot quite true. Thus we have also foud out the likelihood of occurrece of each of the types of edges from our database images. Fially from the average of these membership grades, the fuzzy membership value of 0.5 has bee assiged to the horizotal gree pixel values G ad G. Similarly the two vertical gree pixel value G 3 ad G 4 have bee assiged membership grade of 0.. We have observed that the above assigmet of membership values have resulted i very good color demosaicig, yieldig visually pleasig color recostructio. The membership assigmet has bee decided after performig exhaustive experimets o a large umber of images. O the basis of above discussio, the missig gree value at pixel locatio R ca be iterpolated as follows: Missig 0.5* G 0.5* G 0.* G3 0.* G4 G = = ( G G ) ( G * * G4 ) Usig this fuzzy membership assigmet as a weighted-average tool for missig color iterpolatio, we ca fully utilize all the eighborig iformatio for estimatig the missig color iformatio. 3. Proposed three-steps iterpolatio algorithm The proposed iterpolatio algorithm is a three-step algorithm as summarized below. Step : Estimatio of all missig Gree values. After completio of this step, each ad every pixel locatio has a value for the gree color compoet. Step : Estimatio of missig Blue (Red) color compoet at each pixel locatio cotaiig Red (Blue) color compoet oly. The gree values estimated i the previous step are used i this step. The decisio is based o the chage of hue values. Step 3: Estimatio of missig Red ad Blue at gree pixels. The estimated Red/Blue at blue/red pixels i the previous step have bee utilized for iterpolatio of the missig Red ad Blue at gree pixels. The details of above steps i the proposed color iterpolatio algorithm have bee described below. It should be oted that by lower case r, g or b, we represet the red (r), gree (g) or blue
12 Joural of Electroic Imagig Vol. (3), July 00. (b) values preset i each pixel i the sub-sampled Bayer patter color image. Ad by upper case letters R, G ad B, we deote the estimated values of Red (R), Gree (G) ad Blue (B) i each pixel locatio. Step : Estimatio of all missig Gree values From the Bayer patter, the arragemet of the eighborig pixels i a 5x5 widow with a red pixel at the ceter of the widow is show below. We estimate the missig gree color compoet at this pixel locatio as follows. r m-,- g m-,- r m-, g m-, r m-, g m-,- b m-,- g m-, b m-, g m-, r - g - r g r g m,- b m,- g m, b m, g m, r m,- g m,- r m, g m, r m, First, we estimate two parameters i terms of chages i the Hue values, oe i horizotal directio ad the other i vertical directio, usig the followig two equatios. C hor ( R G ) ( R G ) ( r r ) ( r r ) g ( r g g r ) = ( Rm, Gm, ) ( Rm, Gm ) ( r r ) ( r r ) C ver, m, = m g m, m, ( g g r ) r, m, m, m, g g m, As metioed earlier i the smooth hue trasitio iterpolatio, if the pixel value is already trasformed from liear exposure space ito logarithmic exposure space, oe ca defie the RED hue value as R G. Whe R is ot available, we used a simple average of eighborig R values to approximate it. As show i the above two equatios, the two chages of hue values, C hor ad C ver, ca be easily obtaied usig the filterig operatio at the pixel locatio ( ) i the iput Bayer patter image with a five taps filter, (-0.5,, 0, -, 0.5). The, depedig upo the values of these two parameters, differet fuzzy memberships umbers are used as weightig factors to estimate the missig Gree value (as described i the followig "if the- else" statemet).
13 Joural of Electroic Imagig Vol. (3), July 00. if ( else edif C < C ) the hor ver * I hor G = * I ; if ( Cver < Chor ) G = * I * I ; else edif hor hor ver G = 0.5* I 0.5* I ; Where the two variables I hor ad I ver are estimated as show i equatios () ad () below based o the assumptio of smooth hue trasitio as discussed earlier with a weightig factor (0.5) i the secod term to reduce the sesitivity of oise i the image. ( gm, gm, ) 0.5 ( rm, rm, rm, ) 4 ( g g ) 0.5 ( r r r ) 4 I hor ver = = () I () The two scalig factors, ad 0.667, i above expressios have bee derived experimetally based o the fuzzy membership assigmet strategy. A similar strategy is applied for estimatio of the missig Gree value at the blue pixels i Bayer patter image. Step : Estimatio of the missig Blue/Red value at red/blue pixel From the Bayer patter, the arragemet of the eighborig pixels i a 3x3 widow with a red pixel at the ceter of the widow is show below. ver ver b m-,- g m-, b m-, g - r g b m,- g m, b m, The hue values of the four corer pixels i the widow ad the differece of hues alog the diagoals are estimated usig followig equatios. hue b G ; w = m, m, sw = bm, Gm, e = bm, Gm, se = bm, Gm, hue ; hue ; hue ; hue md = hue hue ; w se 3
14 Joural of Electroic Imagig Vol. (3), July 00. hue sd = hue hue ; e sw Where each G mi,j at r has bee estimated i step. Here hue_md ad hue_sd idicate the differece of hues alog the diagoals. The procedure for estimatio of the missig Blue value i the red pixel is show below. if ( else hue < hue ) the md B B sd ( b G ) ( b G ).8333*( ) = G 0 m, m, m, m, 0 ( bm, Gm, ) ( b Gm, ) m,.667 * ( ).667 *(( b G ) ( b G )) = G 0 m, m, m, m, 0 ( bm, Gm, ) ( b Gm, ) m,.8333* ( ) ; ; edif Estimatio of the missig Red value at a blue pixel may similarly be obtaied. Step 3: Estimatio of the missig Blue/Red value at gree pixel From the Bayer patter, the arragemet of the eighborig pixels i a 3x3 widow with a gree pixel at the ceter of the widow is show below. g m-,- b m-, g m-, r - g r g m,- b m, g m, Here the four hue values adjacet to the ceter pixel i the above widow ad the differece of hues alog the horizotal ad vertical directios are estimated usig followig equatios. hue = b G, ; m, m e = bm, G w = bm, G s bm, Gm hue ; hue ; hue =, ; hue = hue hue ; hue hor ver e w = hue hue ; s Both the color values G mi,j ad B mi,j at the pixel locatio g i above equatios have bee estimated i the previous steps. Here hue_hor ad hue_ver idicate the differece of hues alog 4
15 Joural of Electroic Imagig Vol. (3), July 00. the horizotal ad vertical directios. The procedure for estimatio of the missig Blue value i the gree pixel is show below. if ( hue < hue ) the hor B ver ( B G ) ( B G ).8333*( ) = g 0 0 ( bm, Gm, ) ( b Gm, ) m,.667 * ( ) ; else B.667 *(( B G ) ( B G )) = g 0 0 ( bm, Gm, ) ( b Gm, ) m,.8333* ( ) ; edif Similarly, the missig Red color compoet is estimated i each pixel locatio cotaiig oly the gree color compoet i the Bayer patter image. 4. Experimetal Results Sice it is difficult to grab the sub-sampled Bayer patter images directly from the digital cameras available i the market, we have sythetically geerated the sub-sampled Bayer patter images from 4-bit full color RGB high quality images by simply droppig two color values i each pixel locatio. We have geerated more tha 5 such test images of differet types ad characteristics of the cotets. However i this paper, we have reported results of performace of the iterpolatio algorithms applied o four Bayer patter images that have bee sythetically geerated from four origial full color test images show i Figure 6. We carefully have chose these challegig images of differet characteristics to demostrate performace of differet color iterpolatio algorithms. As show i Figure 6(a), the STAR image cotais lots of high frequecy patters i the form of black ad white sharp edges i differet agles. As show i Figure 6(b), ZEBRA is a image with black ad white strips i the scee. As show i Figure 6(c), the TOWN image cotais a lot of sharp ad colorful edges, ad the NEWENG image i Figure 6(d) is a atural outdoor scee. We compared performace of the proposed color iterpolatio algorithm with three methods, biliear iterpolatio, block matchig based algorithm by Acharya, et. al. [3] ad combiatio of Smooth Hue Trasitio ad Edge Sesig iterpolatio as described i sectio. The recostructed full color images usig these algorithms are show i Figures 7-. I all of these images, our proposed algorithm produces much better result i terms of subjective quality ad objective measure of the peak sigal-to-oise ratio (PSNR). Table shows the PSNR results for these test images. As we ca see, we ca get up to about 0 db improvemet by fuzzy color iterpolatio algorithm as compared with the simple biliear iterpolatio o the STAR image, ad o the average, our ew method ca achieve more tha 5 db improvemet as compared with the other three methods we tested. This STAR image is very useful for testig color iterpolatio algorithms. We ca easily see the blurrig ad color bleedig artifacts, as show i Figure 7(b), (c) ad (d), produced by the biliear iterpolatio, Block Matchig method, ad combiatio of Smooth Hue Trasitio ad Edge Sesig iterpolatio. The proposed fuzzy color iterpolatio algorithm produces the best result i terms of visual quality, as show i Figure 7(a), i term of those blurrig ad color bleedig 5
16 Joural of Electroic Imagig Vol. (3), July 00. artifacts. Figure 8 shows a zoomed versio of those results i Figure 7. I Figures 9, we show three sets of recostructed ZEBRA, TOWN, NEWENG images usig the same algorithms. Image STAR Table : PSNR compariso of differet algorithms PSNR Color Chael Biliear Block Smooth Hue iterpolatio Matchig Edge Sesig Proposed method R G B R ZEBRA G B R TOWN G B R NEWENG G B
17 Joural of Electroic Imagig Vol. (3), July Coclusio ad Summary Due to the cost ad packagig cosideratio, i digital imagig device such as a digital camera, oly a sigle electroic sesor is used ad the eed for color iterpolatio will remai critical util other techologies such as multi-chael color moiré free sesor [0] is mature. I this paper, we preseted a ew color iterpolatio algorithm for Bayer patter sub-sampled color images. The proposed algorithm utilizes the fuzzy membership assigmet as a weightig factor alog with the cocept of smooth hue trasitio for estimatig the missig colors i each pixel. This algorithm sigificatly improves the overall visual quality of the recostructed color images. The experimetal results show that the algorithm preserves colors o the edges with miimal or o visual artifacts. We have preseted the objective quality metrics i terms of PSNR to show the performace of the algorithm with four challegig images for color iterpolatio. Ackowledgemet The authors solemly ackowledge the cotributio of Late A. K. V. Subba Rao for his dedicated work i the area of Color Iterpolatio ad his collaboratio i early phase of this research. Refereces. Bryce E. Bayer, "Color imagig array," U.S. Patet 3,97,065, Eastma Kodak Compay, Do P. Mitchell et. al, "Recostructio Filters i Computer Graphics," Computer Graphics, (SIGGRAPH'88 Proceedigs), Vol., No.4, P.-8, August Hsieh S. Hou et. al., "Cubic Splies for Image Iterpolatio ad Digital Filterig," IEEE Trasactios o Acoustic, Speech ad Sigal Processig, Vol. ASSP-6, pp , David R. Cok, "Sigal processig method ad apparatus for sampled image sigals," U.S. Patet 4,630,307, Eastma Kodak Compay, Eiichi Shimizu et. al., "The Digital Camera Usig New Compressio ad Iterpolatio Algorith" IS&T 49th Aual Coferece, pp:68-7, Tmoasz A. Matraszek, David R. Cok, ad Robert T. Gray, "Gradiet based method for providig values for ukow pixels i a digital image," U.S. Patet 5,875,040, Eastma Kodak Compay, James E. Adams, Jr., Iteractios betwee color plae iterpolatio ad other image processig fuctios i electroic photography, Proceedigs of the SPIE Electroic Imagig Coferece, Vol. 46, pp:44-5, Tadashi Sakamoto, Chikako Nakaishi ad Tomohiro Hase, "Software Pixel Iterpolatio for Digital Still Cameras Suitable for A 3-bit MCU," IEEE Trasactios o Cosumer Electroics, Vol. 44, No. 4, pp 34-35, Nov
18 Joural of Electroic Imagig Vol. (3), July William T Freema, "Method ad apparatus for recostructig missig color samples," U.S. Patet 4,663,655, Polaroid Corporatio, David R Cok, "Sigal processig method ad apparatus for producig iterpolated chromiace values i a sampled color image sigal," U.S. Patet 4,64,678, Eastma Kodak Compay, X. Wu, W. K. Choi, ad P. Bao, Color Restoratio from Digital Camera Data by Patter Matchig, Proceedigs of the SPIE s Electroic Imagig Coferece, Color Imagig: Device- Idepedet Color, Color Hardcopy, ad Graphic Arts II, Vol. 308, pp. -7, Tiku Acharya ad Pig-Sig Tsai, "A New Block Matchig Based Color Iterpolatio Algorith'' Proceedigs of the SPIE Electroic Imagig Coferece, Color Imagig: Device- Idepedet Color, Color Hardcopy, ad Graphic Arts IV, Vol. 3648, pp , Robert H. Hibbard, "Apparatus ad method for adaptively iterpolatig a full color image utilize lumiace gradiets," U.S. Patet 5,38,976, Eastma Kodak Compay, Glaude A. Laroche ad Mark A. Prescott, Apparatus ad method for adaptively iterpolatig a full color image utilize chromiace gradiets," U.S. Patet 5,373,3, Eastma Kodak Compay, Joh F. Hamilto, Jr. ad James E. Adams, Jr., "Adaptive color pla iterpolatio i sigle sesor color electroic camera," U.S. Patet 5,69,734, Eastma Kodak Compay, Zadeh, L. A. Fuzzy Sets, Iformatio ad Cotrol 8; , Zimmerma, H. J. Fuzzy Set Theory ad Its Applicatios, d ed. Norwell, MA: Kluwer Academic, Kazuo, T: traslated by Niimura, T. A Itroductio to Fuzzy Logic for Practical Applicatios, Spriger-Verlag New York, Ic P. G. Herzog, D. Kipp, H. Stiebig ad F. Koeig, "Characterizatio of ovel three ad six chael color moire free sesors,'' Proceedigs of the SPIE Electroic Imagig Coferece, Color Imagig: Device-Idepedet Color, Color Hardcopy, ad Graphic Arts IV, Vol. 3648, pp 48-59,
19 Joural of Electroic Imagig Vol. (3), July 00. (a) (b) (c) (d) Figure 6: Origial 4 bits full color images: (a) Star (b) Zebra (c) Tow (d) Neweg. 9
20 Joural of Electroic Imagig Vol. (3), July 00. (a) (b) (c) (d) Figure 7: STAR image: (a) Recostructed image usig proposed algorith (b) Recostructed image usig Biliear color iterpolatio, (c) Recostructed image usig Block Matchig based method, ad (d) Recostructed image usig combiatio of Smooth Hue ad Edge Sesig iterpolatio. 0
21 Joural of Electroic Imagig Vol. (3), July 00. (a) (b) (c) (d) Figure 8: STAR image (ceter portio oly): (a) Recostructed image usig proposed algorith (b) Recostructed image usig Biliear color iterpolatio, (c) Recostructed image usig Block Matchig based method, ad (d) Recostructed image usig combiatio of Smooth Hue ad Edge Sesig iterpolatio.
22 Joural of Electroic Imagig Vol. (3), July 00. (a) (b) (c) (d) Figure 9: Zebra image: (a) Recostructed image usig proposed algorith (b) Recostructed image usig Biliear color iterpolatio, (c) Recostructed image usig Block Matchig based method, ad (d) Recostructed image usig combiatio of Smooth Hue ad Edge Sesig iterpolatio.
23 Joural of Electroic Imagig Vol. (3), July 00. (a) (b) (c) (d) Figure 0: Tow image: (a) Recostructed image usig proposed algorith (b) Recostructed image usig Biliear color iterpolatio, (c) Recostructed image usig Block Matchig based method, ad (d) Recostructed image usig combiatio of Smooth Hue ad Edge Sesig iterpolatio. 3
24 Joural of Electroic Imagig Vol. (3), July 00. (a) (b) (c) (d) Figure : Neweg image: (a) Recostructed image usig proposed algorith (b) Recostructed image usig Biliear color iterpolatio, (c) Recostructed image usig Block Matchig based method, ad (d) Recostructed image usig combiatio of Smooth Hue ad Edge Sesig iterpolatio. 4
CHAPTER 5 A NEAR-LOSSLESS RUN-LENGTH CODER
95 CHAPTER 5 A NEAR-LOSSLESS RUN-LENGTH CODER 5.1 GENERAL Ru-legth codig is a lossless image compressio techique, which produces modest compressio ratios. Oe way of icreasig the compressio ratio of a ru-legth
More informationA New Space-Repetition Code Based on One Bit Feedback Compared to Alamouti Space-Time Code
Proceedigs of the 4th WSEAS It. Coferece o Electromagetics, Wireless ad Optical Commuicatios, Veice, Italy, November 0-, 006 107 A New Space-Repetitio Code Based o Oe Bit Feedback Compared to Alamouti
More informationFingerprint Classification Based on Directional Image Constructed Using Wavelet Transform Domains
7 Figerprit Classificatio Based o Directioal Image Costructed Usig Wavelet Trasform Domais Musa Mohd Mokji, Syed Abd. Rahma Syed Abu Bakar, Zuwairie Ibrahim 3 Departmet of Microelectroic ad Computer Egieerig
More informationAn Adaptive Image Denoising Method based on Thresholding
A Adaptive Image Deoisig Method based o Thresholdig HARI OM AND MANTOSH BISWAS Departmet of Computer Sciece & Egieerig Idia School of Mies, Dhabad Jharkad-86004 INDIA {hariom4idia, matoshb}@gmail.com Abstract
More informationApplication of Image Fusion to Wireless Image Transmission
Applicatio of Image Fusio to Wireless Image Trasmissio Liae C. Ramac ad Pramod K. Varshey EECS Departmet, Syracuse Uiversity 121 Lik Hall, Syracuse, NY 13244 Email: lcramac@mailbox.syr.edu, varshey@cat.syr.edu
More informationAPPLICATION NOTE UNDERSTANDING EFFECTIVE BITS
APPLICATION NOTE AN95091 INTRODUCTION UNDERSTANDING EFFECTIVE BITS Toy Girard, Sigatec, Desig ad Applicatios Egieer Oe criteria ofte used to evaluate a Aalog to Digital Coverter (ADC) or data acquisitio
More informationLossless image compression Using Hashing (using collision resolution) Amritpal Singh 1 and Rachna rajpoot 2
Lossless image compressio Usig Hashig (usig collisio resolutio) Amritpal Sigh 1 ad Racha rajpoot 2 1 M.Tech.* CSE Departmet, 2 Departmet of iformatio techology Guru Kashi UiversityTalwadi Sabo, Bathida
More informationMeasurement of Equivalent Input Distortion AN 20
Measuremet of Equivalet Iput Distortio AN 2 Applicatio Note to the R&D SYSTEM Traditioal measuremets of harmoic distortio performed o loudspeakers reveal ot oly the symptoms of the oliearities but also
More informationLogarithms APPENDIX IV. 265 Appendix
APPENDIX IV Logarithms Sometimes, a umerical expressio may ivolve multiplicatio, divisio or ratioal powers of large umbers. For such calculatios, logarithms are very useful. They help us i makig difficult
More informationWavelet Transform. CSEP 590 Data Compression Autumn Wavelet Transformed Barbara (Enhanced) Wavelet Transformed Barbara (Actual)
Wavelet Trasform CSEP 59 Data Compressio Autum 7 Wavelet Trasform Codig PACW Wavelet Trasform A family of atios that filters the data ito low resolutio data plus detail data high pass filter low pass filter
More informationImage Denoising using Fast Bilateral Filter
Iteratioal Joural of Curret Egieerig ad Techology E-ISS 77 4106, P-ISS 347 5161 015IPRESSCO, All Rights Reserved Available at http://ipressco.com/category/ijcet Research Article Riku A Joseph * ad Maju
More informationComparison of Frequency Offset Estimation Methods for OFDM Burst Transmission in the Selective Fading Channels
Compariso of Frequecy Offset Estimatio Methods for OFDM Burst Trasmissio i the Selective Fadig Chaels Zbigiew Długaszewski Istitute of Electroics ad Telecommuicatios Pozań Uiversity of Techology 60-965
More informationColor gamut of SOCS and its comparison to Pointer's gamut
Color gamut of SOCS ad its compariso to Poiter's gamut Masao Iui*, Tomotaka Hirokaw, oshihiko Azum, ad Johji Tajim* * Tokyo Polytechic Uiversity 58 Iiyama, Atsugi, Kaagawa 24-297, Japa ** Nagoya City Uiversity
More informationDesign of FPGA- Based SPWM Single Phase Full-Bridge Inverter
Desig of FPGA- Based SPWM Sigle Phase Full-Bridge Iverter Afarulrazi Abu Bakar 1, *,Md Zarafi Ahmad 1 ad Farrah Salwai Abdullah 1 1 Faculty of Electrical ad Electroic Egieerig, UTHM *Email:afarul@uthm.edu.my
More informationINCREASE OF STRAIN GAGE OUTPUT VOLTAGE SIGNALS ACCURACY USING VIRTUAL INSTRUMENT WITH HARMONIC EXCITATION
XIX IMEKO World Cogress Fudametal ad Applied Metrology September 6, 9, Lisbo, Portugal INCREASE OF STRAIN GAGE OUTPUT VOLTAGE SIGNALS ACCURACY USING VIRTUAL INSTRUMENT WITH HARMONIC EXCITATION Dalibor
More informationRadar emitter recognition method based on AdaBoost and decision tree Tang Xiaojing1, a, Chen Weigao1 and Zhu Weigang1 1
Advaces i Egieerig Research, volume 8 d Iteratioal Coferece o Automatio, Mechaical Cotrol ad Computatioal Egieerig (AMCCE 7) Radar emitter recogitio method based o AdaBoost ad decisio tree Tag Xiaojig,
More informationPROJECT #2 GENERIC ROBOT SIMULATOR
Uiversity of Missouri-Columbia Departmet of Electrical ad Computer Egieerig ECE 7330 Itroductio to Mechatroics ad Robotic Visio Fall, 2010 PROJECT #2 GENERIC ROBOT SIMULATOR Luis Alberto Rivera Estrada
More informationX-Bar and S-Squared Charts
STATGRAPHICS Rev. 7/4/009 X-Bar ad S-Squared Charts Summary The X-Bar ad S-Squared Charts procedure creates cotrol charts for a sigle umeric variable where the data have bee collected i subgroups. It creates
More informationHistory and Advancement of the Family of Log Periodic Toothed Planer Microstrip Antenna
Joural of Electromagetic Aalysis ad Applicatios, 2011, 3, 242-247 doi:10.4236/jemaa.2011.36039 Published Olie Jue 2011 (http://www.scirp.org/joural/jemaa) History ad Advacemet of the Family of Log Periodic
More informationDesign of FPGA Based SPWM Single Phase Inverter
Proceedigs of MUCEET2009 Malaysia Techical Uiversities Coferece o Egieerig ad Techology Jue 20-22, 2009, MS Garde,Kuata, Pahag, Malaysia MUCEET2009 Desig of FPGA Based SPWM Sigle Phase Iverter Afarulrazi
More informationRecovering low frequencies for impedance inversion by frequency domain deconvolution
Recoverig low frequecies for impedace iversio by frequecy domai decovolutio Sia. Esmaeili*, CREWES, Uiversity of Calgary, sesmaeil@ucalgary.ca Gary. Frak. Margrave, CREWES, Uiversity of Calgary, margrave@ucalgary.ca
More informationLecture 4: Frequency Reuse Concepts
EE 499: Wireless & Mobile Commuicatios (8) Lecture 4: Frequecy euse Cocepts Distace betwee Co-Chael Cell Ceters Kowig the relatio betwee,, ad, we ca easily fid distace betwee the ceter poits of two co
More informationCOMPRESSION OF TRANSMULTIPLEXED ACOUSTIC SIGNALS
COMPRESSION OF TRANSMULTIPLEXED ACOUSTIC SIGNALS Mariusz Ziółko, Przemysław Sypka ad Bartosz Ziółko Departmet of Electroics, AGH Uiversity of Sciece ad Techology, al. Mickiewicza 3, 3-59 Kraków, Polad,
More informationAC : USING ELLIPTIC INTEGRALS AND FUNCTIONS TO STUDY LARGE-AMPLITUDE OSCILLATIONS OF A PENDULUM
AC 007-7: USING ELLIPTIC INTEGRALS AND FUNCTIONS TO STUDY LARGE-AMPLITUDE OSCILLATIONS OF A PENDULUM Josue Njock-Libii, Idiaa Uiversity-Purdue Uiversity-Fort Waye Josué Njock Libii is Associate Professor
More informationThe Detection of Abrupt Changes in Fatigue Data by Using Cumulative Sum (CUSUM) Method
Proceedigs of the th WSEAS Iteratioal Coferece o APPLIED ad THEORETICAL MECHANICS (MECHANICS '8) The Detectio of Abrupt Chages i Fatigue Data by Usig Cumulative Sum (CUSUM) Method Z. M. NOPIAH, M.N.BAHARIN,
More informationIntermediate Information Structures
Modified from Maria s lectures CPSC 335 Itermediate Iformatio Structures LECTURE 11 Compressio ad Huffma Codig Jo Roke Computer Sciece Uiversity of Calgary Caada Lecture Overview Codes ad Optimal Codes
More informationA New Design of Log-Periodic Dipole Array (LPDA) Antenna
Joural of Commuicatio Egieerig, Vol., No., Ja.-Jue 0 67 A New Desig of Log-Periodic Dipole Array (LPDA) Atea Javad Ghalibafa, Seyed Mohammad Hashemi, ad Seyed Hassa Sedighy Departmet of Electrical Egieerig,
More informationAnalysis of SDR GNSS Using MATLAB
Iteratioal Joural of Computer Techology ad Electroics Egieerig (IJCTEE) Volume 5, Issue 3, Jue 2015 Aalysis of SDR GNSS Usig MATLAB Abstract This paper explais a software defied radio global avigatio satellite
More informationMEASUREMENT AND CONTORL OF TOTAL HARMONIC DISTORTION IN FREQUENCY RANGE 0,02-10KHZ.
ELECTRONICS 00 September, Sozopol, BLGARIA MEASREMENT AND CONTORL OF TOTAL HARMONIC DISTORTION IN FREQENCY RANGE 0,0-0KHZ. Plame Agelov Agelov Faculty for Computer Sciece, Egieerig ad Natural Studies,
More informationA study on the efficient compression algorithm of the voice/data integrated multiplexer
A study o the efficiet compressio algorithm of the voice/data itegrated multiplexer Gyou-Yo CHO' ad Dog-Ho CHO' * Dept. of Computer Egieerig. KyiigHee Uiv. Kiheugup Yogiku Kyuggido, KOREA 449-71 PHONE
More informationA Novel Small Signal Power Line Quality Measurement System
IMTC 3 - Istrumetatio ad Measuremet Techology Coferece Vail, CO, USA, - May 3 A ovel Small Sigal Power Lie Quality Measuremet System Paul B. Crilly, Erik Leadro Boaldi, Levy Ely de Lacarda de Oliveira,
More informationHigh Speed Area Efficient Modulo 2 1
High Speed Area Efficiet Modulo 2 1 1-Soali Sigh (PG Scholar VLSI, RKDF Ist Bhopal M.P) 2- Mr. Maish Trivedi (HOD EC Departmet, RKDF Ist Bhopal M.P) Adder Abstract Modular adder is oe of the key compoets
More informationCombined Scheme for Fast PN Code Acquisition
13 th Iteratioal Coferece o AEROSPACE SCIENCES & AVIATION TECHNOLOGY, ASAT- 13, May 6 8, 009, E-Mail: asat@mtc.edu.eg Military Techical College, Kobry Elkobbah, Cairo, Egypt Tel : +(0) 4059 4036138, Fax:
More informationIntroduction to Wireless Communication Systems ECE 476/ECE 501C/CS 513 Winter 2003
troductio to Wireless Commuicatio ystems ECE 476/ECE 501C/C 513 Witer 2003 eview for Exam #1 March 4, 2003 Exam Details Must follow seatig chart - Posted 30 miutes before exam. Cheatig will be treated
More informationApplication of Improved Genetic Algorithm to Two-side Assembly Line Balancing
206 3 rd Iteratioal Coferece o Mechaical, Idustrial, ad Maufacturig Egieerig (MIME 206) ISBN: 978--60595-33-7 Applicatio of Improved Geetic Algorithm to Two-side Assembly Lie Balacig Ximi Zhag, Qia Wag,
More informationFault Diagnosis in Rolling Element Bearing Using Filtered Vibration and Acoustic Signal
Volume 8 o. 8 208, 95-02 ISS: 3-8080 (prited versio); ISS: 34-3395 (o-lie versio) url: http://www.ijpam.eu ijpam.eu Fault Diagosis i Rollig Elemet Usig Filtered Vibratio ad Acoustic Sigal Sudarsa Sahoo,
More informationHigh-Order CCII-Based Mixed-Mode Universal Filter
High-Order CCII-Based Mixed-Mode Uiversal Filter Che-Nog Lee Departmet of Computer ad Commuicatio Egieerig, Taipei Chegshih Uiversity of Sciece ad Techology, Taipei, Taiwa, R. O. C. Abstract This paper
More informationUsing Color Histograms to Recognize People in Real Time Visual Surveillance
Usig Color Histograms to Recogize People i Real Time Visual Surveillace DANIEL WOJTASZEK, ROBERT LAGANIERE S.I.T.E. Uiversity of Ottawa, Ottawa, Otario CANADA daielw@site.uottawa.ca, lagaier@site.uottawa.ca
More informationMid-level representations for audio content analysis *Slides for this lecture were created by Anssi Klapuri
Mid-level represetatios for audio cotet aalysis *Slides for this lecture were created by Assi Klapuri Represetatios 1 Sources: Ellis, Rosethal, Mid-level represetatios for Computatioal Auditory Scee Aalysis,
More informationDensity Slicing Reference Manual
Desity Slicig Referece Maual Improvisio, Viscout Cetre II, Uiversity of Warwick Sciece Park, Millbur Hill Road, Covetry. CV4 7HS Tel: 0044 (0) 24 7669 2229 Fax: 0044 (0) 24 7669 0091 e-mail: admi@improvisio.com
More informationSubband Coding of Speech Signals Using Decimation and Interpolation
3 th Iteratioal Coferece o AEROSPACE SCIENCES & AVIATION TECHNOLOGY, ASAT- 3, May 6 8, 9, E-Mail: asat@mtc.edu.eg Military Techical College, Kobry Elkobbah, Cairo, Egypt Tel : +() 459 43638, Fax: +() 698
More informationCompound Controller for DC Motor Servo System Based on Inner-Loop Extended State Observer
BULGARIAN ACADEMY OF SCIENCES CYBERNETICS AND INFORMATION TECHNOLOGIES Volume 6, No 5 Special Issue o Applicatio of Advaced Computig ad Simulatio i Iformatio Systems Sofia 06 Prit ISSN: 3-970; Olie ISSN:
More informationCOS 126 Atomic Theory of Matter
COS 126 Atomic Theory of Matter 1 Goal of the Assigmet Video Calculate Avogadro s umber Usig Eistei s equatios Usig fluorescet imagig Iput data Output Frames Blobs/Beads Estimate of Avogadro s umber 7.1833
More informationEncode Decode Sample Quantize [ ] [ ]
Referece Audio Sigal Processig I Shyh-Kag Jeg Departmet of Electrical Egieerig/ Graduate Istitute of Commuicatio Egieerig M. Bosi ad R. E. Goldberg, Itroductio to Digital Audio Codig ad Stadards, Kluwer
More informationProblem of calculating time delay between pulse arrivals
America Joural of Egieerig Research (AJER) 5 America Joural of Egieerig Research (AJER) e-issn: 3-847 p-issn : 3-936 Volume-4, Issue-4, pp-3-4 www.ajer.org Research Paper Problem of calculatig time delay
More informationNovel Steganography System using Lucas Sequence
Novel Stegaography System usig ucas Sequece Fahd Alharbi Faculty of Egieerig Kig Abdulaziz Uiversity Rabigh, KSA Abstract Stegaography is the process of embeddig data ito a media form such as image, voice,
More informationSHORT-TERM TRAVEL TIME PREDICTION USING A NEURAL NETWORK
SHORT-TERM TRAVEL TIME PREDICTION USING A NEURAL NETWORK Giovai Huiske ad Eric va Berkum Dept. of Civil Egieerig - Uiversity of Twete - 7500 AE Eschede - The Netherlads E-mail: g.huiske@ctw.utwete.l ad
More informationSingle Bit DACs in a Nutshell. Part I DAC Basics
Sigle Bit DACs i a Nutshell Part I DAC Basics By Dave Va Ess, Pricipal Applicatio Egieer, Cypress Semicoductor May embedded applicatios require geeratig aalog outputs uder digital cotrol. It may be a DC
More informationA SELECTIVE POINTER FORWARDING STRATEGY FOR LOCATION TRACKING IN PERSONAL COMMUNICATION SYSTEMS
A SELETIVE POINTE FOWADING STATEGY FO LOATION TAKING IN PESONAL OUNIATION SYSTES Seo G. hag ad hae Y. Lee Departmet of Idustrial Egieerig, KAIST 373-, Kusug-Dog, Taejo, Korea, 305-70 cylee@heuristic.kaist.ac.kr
More informationData Acquisition System for Electric Vehicle s Driving Motor Test Bench Based on VC++ *
Available olie at www.sciecedirect.com Physics Procedia 33 (0 ) 75 73 0 Iteratioal Coferece o Medical Physics ad Biomedical Egieerig Data Acquisitio System for Electric Vehicle s Drivig Motor Test Bech
More informationSurvey of Low Power Techniques for ROMs
Survey of Low Power Techiques for ROMs Edwi de Agel Crystal Semicoductor Corporatio P.O Box 17847 Austi, TX 78744 Earl E. Swartzlader, Jr. Departmet of Electrical ad Computer Egieerig Uiversity of Texas
More informationDisplacement estimation
Displacemet estimatio Displacemet estimatio by block matchig l Search strategies l Subpixel estimatio Digital Image Processig: Berd Girod, Gordo Wetzstei 2013-2016 Staford Uiversity -- Displacemet Estimatio
More informationSelection of the basic parameters of the lens for the optic-electronic target recognition system
Proceedigs of the 5th WSEAS It. Cof. o COMPUTATIONAL INTELLIGENCE, MAN-MACHINE SYSTEMS AND CYBERNETICS, Veice, Italy, November 0-, 006 317 Selectio of the basic parameters of the les for the optic-electroic
More informationOn Parity based Divide and Conquer Recursive Functions
O Parity based Divide ad Coquer Recursive Fuctios Sug-Hyu Cha Abstract The parity based divide ad coquer recursio trees are itroduced where the sizes of the tree do ot grow mootoically as grows. These
More informationEstimation of non Distortion Audio Signal Compression
Estimatio of o Distortio Audio Sigal Compressio M. BAK, S. PODOXI, V. TSIGOUZ Departmet of Commuicatio Egieerig Ceter for Techological Educatio Holo 5 Golomb street, POB 35 Holo 58, Tel: +97-3-5669 ISRAEL
More informationTehrani N Journal of Scientific and Engineering Research, 2018, 5(7):1-7
Available olie www.jsaer.com, 2018, 5(7):1-7 Research Article ISSN: 2394-2630 CODEN(USA): JSERBR 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38
More informationHigh-resolution speech signal reconstruction in Wireless Sensor Networks
High-resolutio speech sigal recostructio i Wireless Sesor Networks Adria Pazarloglou, Radu Stoleru, Ricardo Gutierrez-Osua Departmet of Computer Sciece, Texas A&M Uiversity {adria, stoleru, rgutier}@cs.tamu.edu
More informationPHY-MAC dialogue with Multi-Packet Reception
PHY-AC dialogue with ulti-packet Receptio arc Realp 1 ad Aa I. Pérez-Neira 1 CTTC-Cetre Tecològic de Telecomuicacios de Cataluya Edifici Nexus C/Gra Capità, - 0803-Barceloa (Cataluya-Spai) marc.realp@cttc.es
More informationAssessment of Soil Parameter Estimation Errors for Fusion of Multichannel Radar Measurements
Assessmet of Soil Parameter Estimatio Errors for Fusio of Multichael Radar Measuremets A. urei, D. Marshall, D. Radford,. Lever Cardiff Uiversity 5 The Parade, Roath, CF4 0YF, Cardiff, U A.urei@cs.cf.ac.u
More informationIllumination Compensation for 2-D Barcode Recognition Basing Morphologic
Research Joural of Applied Scieces, Egieerig ad Techology 5(12): 3273-3280, 2013 ISSN: 2040-7459; e-issn: 2040-7467 Maxwell Scietific Orgaizatio, 2013 Submitted: March, 09, 2012 Accepted: August 15, 2012
More informationAcquisition of GPS Software Receiver Using Split-Radix FFT
006 IEEE Coferece o Systems, Ma, ad Cyberetics October -, 006, Taipei, Taiwa Acquisitio of GPS Software Receiver Usig Split-Radix FFT W. H. Li, W. L. Mao, H. W. Tsao, F. R. Chag, ad W. H. Huag Abstract
More informationBOUNDS FOR OUT DEGREE EQUITABLE DOMINATION NUMBERS IN GRAPHS
BULLETIN OF THE INTERNATIONAL MATHEMATICAL VIRTUAL INSTITUTE ISSN 2303-4874 (p), ISSN (o) 2303-4955 www.imvibl.org/bulletin Vol. 3(2013), 149-154 Former BULLETIN OF THE SOCIETY OF MATHEMATICIANS BANJA
More informationx y z HD(x, y) + HD(y, z) HD(x, z)
Massachusetts Istitute of Techology Departmet of Electrical Egieerig ad Computer Sciece 6.02 Solutios to Chapter 5 Updated: February 16, 2012 Please sed iformatio about errors or omissios to hari; questios
More informationThe Fast Haar Wavelet Transform for Signal & Image Processing
Vol. 7, No., The Fast Haar Wavelet Trasform for Sigal & Image Processig V.Ashok T.Balakumara C.Gowrishakar epartmet of BE, epartmet of ECE epartmet of EEE Velalar College of Egg.&Tech. Velalar College
More informationA New Basic Unit for Cascaded Multilevel Inverters with the Capability of Reducing the Number of Switches
Joural of Power Electroics, ol, o, pp 67-677, July 67 JPE --6 http://dxdoiorg/6/jpe67 I(Prit: 98-9 / I(Olie: 9-78 A ew Basic Uit for Cascaded Multi Iverters with the Capability of Reducig the umber of
More informationA New FDTD Method for the Study of MRI Pulsed Field Gradient- Induced Fields in the Human Body
A New FDTD Method for the Study of MRI Pulsed Field Gradiet- Iduced Fields i the Huma Body Stuart Crozier, Huawei Zhao ad Liu Feg Cetre For Magetic Resoace, The Uiversity of Queeslad, St. Lucia, Qld 4072,
More informationNovel Modeling Techniques for RTL Power Estimation
Novel Modelig Techiques for RTL Power Estimatio Michael Eierma Walter Stechele Istitute for Itegrated Circuits Istitute for Itegrated Circuits Techical Uiversity of Muich Techical Uiversity of Muich Arcisstr.
More informationBy: Pinank Shah. Date : 03/22/2006
By: Piak Shah Date : 03/22/2006 What is Strai? What is Strai Gauge? Operatio of Strai Gauge Grid Patters Strai Gauge Istallatio Wheatstoe bridge Istrumetatio Amplifier Embedded system ad Strai Gauge Strai
More informationTesting of methods for blind estimation of noise variance on large image database
Testig of methods for blid estimatio of oise variace o large image database Vladimir V. Luki a, Sergey K. Abramov a, Mikhail L. Uss a,b, Igor Marusiy a, Nikolay N. Poomareko a, Alexader A. Zelesky a, Beoit
More informationLaboratory Exercise 3: Dynamic System Response Laboratory Handout AME 250: Fundamentals of Measurements and Data Analysis
Laboratory Exercise 3: Dyamic System Respose Laboratory Hadout AME 50: Fudametals of Measuremets ad Data Aalysis Prepared by: Matthew Beigto Date exercises to be performed: Deliverables: Part I 1) Usig
More informationMarkov Modulated Punctured Autoregressive Processes for Traffic and Channel Modeling *
Markov Modulated Puctured Autoregressive Processes for Traffic ad Chael Modelig * Trista Pei-chu Che ad Tsuha Che + Departmet of Electrical ad Computer Egieerig Caregie Mello Uiversity, PA 1513 {peichu,
More informationALGORITHM FOR COLOR RECOGNITION, ADAPTED FOR PLATFORMS WITH LIMITED RESOURCES
АЛГОРИТЪМ ЗА РАЗПОЗНАВАНЕ НА ЦВЕТОВЕ, АДАПТИРАН ЗА ПЛАТФОРМИ С ОГРАНИЧЕНИ РЕСУРСИ ALGORITHM FOR COLOR RECOGNITION, ADAPTED FOR PLATFORMS WITH LIMITED RESOURCES R. Ivaov Techical Uiversity of Gabrovo, 5300
More informationAdaptive Resource Allocation in Multiuser OFDM Systems
Adaptive Resource Allocatio i Multiuser OFDM Systems Fial Report Multidimesioal Digital Sigal Processig Malik Meherali Saleh The Uiversity of Texas at Austi malikmsaleh@mail.utexas.edu Sprig 005 Abstract
More informationImage Contrast Enhancement Using Histogram Modification Technique
Image Cotrast Ehacemet Usig Histogram Modificatio Techique Shekhar R. Suralkar 1, Atul H. Karode 2, Maali S. Rathi 3 1 Asso. Prof. (E&TC Dept.), SSBT s COET, Jalgao, Idia 2 Asst. Prof. (E&TC Dept.), SSBT
More informationCooperative Diversity Based on Code Superposition
1 Cooperative Diversity Based o Code Superpositio Lei Xiao, Thomas E. Fuja, Jörg Kliewer, Daiel J. Costello, Jr. Departmet of Electrical Egieerig, Uiversity of Notre Dame, Notre Dame, IN 46556, USA Email:
More informationDIGITALLY TUNED SINUSOIDAL OSCILLATOR USING MULTIPLE- OUTPUT CURRENT OPERATIONAL AMPLIFIER FOR APPLICATIONS IN HIGH STABLE ACOUSTICAL GENERATORS
Molecular ad Quatum Acoustics vol. 7, (6) 95 DGTALL TUNED SNUSODAL OSCLLATOR USNG MULTPLE- OUTPUT CURRENT OPERATONAL AMPLFER FOR APPLCATONS N HGH STABLE ACOUSTCAL GENERATORS Lesław TOPÓR-KAMŃSK Faculty
More informationOptimal Geolocation Updating for Location Aware Service Provisioning in Wireless Networks
Optimal Geolocatio Updatig for Locatio Aware Service Provisioig i Wireless Networks Siri Tekiay Amer Catovic tekiay@adm.jit.edu axc4466@jit.edu New Jersey Istitute of Techology Uiversity Heights, Newark,
More informationNovel image enhancement technique using shunting inhibitory cellular neural networks
Edith Cowa Uiversity Research Olie ECU Publicatios Pre. 2011 2001 Novel image ehacemet techique usig shutig ihibitory cellular eural etworks Tarik Hammadou Motorola Labs Abdesselam Bouzerdoum Edith Cowa
More informationVoice Command Recognition System Based on MFCC and VQ Algorithms
Voice Commad Recogitio System Based o MFCC ad VQ Algorithms Mahdi Shaeh, ad Azizollah Taheri Abstract The goal of this project is to desig a system to recogitio voice commads. Most of voice recogitio systems
More informationROBUST RADIO RESOURCE ALLOCATION IN LTE NETWORKS BY CHANNEL AND RELAY ASSIGNMENT
Joural of Egieerig Sciece ad Techology Vol. 12, No. 7 (2017) 1845-1854 School of Egieerig, Taylor s Uiversity ROBUST RADIO RESOURCE ALLOCATION IN LTE NETWORKS BY CHANNEL AND RELAY ASSIGNMENT R. SANTHAKUMAR*,
More informationMassachusetts Institute of Technology Dept. of Electrical Engineering and Computer Science Fall Semester, Introduction to EECS 2.
Massachusetts Istitute of Techology Dept. of Electrical Egieerig ad Computer Sciece Fall Semester, 006 6.08 Itroductio to EECS Prelab Exercises Pre-Lab#3 Modulatio, demodulatio, ad filterig are itegral
More informationThe Fundamental Capacity-Delay Tradeoff in Large Mobile Ad Hoc Networks
The Fudametal Capacity-Delay Tradeoff i Large Mobile Ad Hoc Networks Xiaoju Li ad Ness B. Shroff School of Electrical ad Computer Egieerig, Purdue Uiversity West Lafayette, IN 47907, U.S.A. {lix, shroff}@ec.purdue.edu
More informationEMPIRICAL MODE DECOMPOSITION IN AUDIO WATERMARKING BY USING WAVELET METHOD
EMPIRICAL MODE DECOMPOSITION IN AUDIO WATERMARKING BY USING WAVELET METHOD G.PAVAN KUMAR (PG Scholor) S.RAHUL (M.Tech) Departmet of ECE, S.V.College of Egieerig ad Techology, JNTUH Assistat Professor,
More informationUnit 5: Estimating with Confidence
Uit 5: Estimatig with Cofidece Sectio 8.2 The Practice of Statistics, 4 th editio For AP* STARNES, YATES, MOORE Uit 5 Estimatig with Cofidece 8.1 8.2 8.3 Cofidece Itervals: The Basics Estimatig a Populatio
More informationTHE LUCAS TRIANGLE RECOUNTED. Arthur T. Benjamin Dept. of Mathematics, Harvey Mudd College, Claremont, CA Introduction
THE LUCAS TRIANLE RECOUNTED Arthur T Bejami Dept of Mathematics, Harvey Mudd College, Claremot, CA 91711 bejami@hmcedu 1 Itroductio I 2], Neville Robbis explores may properties of the Lucas triagle, a
More informationImplementation of Fuzzy Multiple Objective Decision Making Algorithm in a Heterogeneous Mobile Environment
Implemetatio of Fuzzy Multiple Objective Decisio Makig Algorithm i a Heterogeeous Mobile Eviromet P.M.L. ha, Y.F. Hu, R.E. Sheriff, Departmet of Electroics ad Telecommuicatios Departmet of yberetics, Iteret
More informationPerformance Analysis of Channel Switching with Various Bandwidths in Cognitive Radio
Performace Aalysis of Chael Switchig with Various Badwidths i Cogitive Radio Po-Hao Chag, Keg-Fu Chag, Yu-Che Che, ad Li-Kai Ye Departmet of Electrical Egieerig, Natioal Dog Hwa Uiversity, 1,Sec.2, Da-Hsueh
More informationReducing Power Dissipation in Complex Digital Filters by using the Quadratic Residue Number System Λ
Reducig Power Dissipatio i Complex Digital Filters by usig the Quadratic Residue Number System Λ Agelo D Amora, Alberto Naarelli, Marco Re ad Gia Carlo Cardarilli Departmet of Electrical Egieerig Uiversity
More informationON THE FUNDAMENTAL RELATIONSHIP BETWEEN THE ACHIEVABLE CAPACITY AND DELAY IN MOBILE WIRELESS NETWORKS
Chapter ON THE FUNDAMENTAL RELATIONSHIP BETWEEN THE ACHIEVABLE CAPACITY AND DELAY IN MOBILE WIRELESS NETWORKS Xiaoju Li ad Ness B. Shroff School of Electrical ad Computer Egieerig, Purdue Uiversity West
More informationsible number of wavelengths. The wave~~ngt~ ~ ~ ~ c ~ n b~dwidth is set low eno~gh to interfax One of the most im
sible umber of wavelegths. The wave~~gt~ ~ ~ ~ c ~ b~dwidth is set low eo~gh to iterfax vices. Oe of the most im ed trasmitters ad ysis much more CO "The author is also f Cumputer sciece Departmet, Uiversity
More informationBANDWIDTH AND GAIN ENHANCEMENT OF MULTIBAND FRACTAL ANTENNA BASED ON THE SIERPINSKI CARPET GEOMETRY
ISSN: 2229-6948(ONLINE) DOI: 10.21917/ijct.2013.0095 ICTACT JOURNAL ON COMMUNICATION TECHNOLOGY, MARCH 2013, VOLUME: 04, ISSUE: 01 BANDWIDTH AND GAIN ENHANCEMENT OF MULTIBAND FRACTAL ANTENNA BASED ON THE
More informationExperimental Noise Analysis of Reed Switch Sensor Signal under Environmental Vibration
Computer Techology ad Applicatio 7 (16) 96-1 doi: 1.1765/1934-733/16..4 D DAVID PUBLISHING Experimetal Noise Aalysis of Reed Switch Sesor Sigal uder Evirometal Vibratio Odgerel Ayurzaa 1 ad Hiesik Kim
More informationInternational Power, Electronics and Materials Engineering Conference (IPEMEC 2015)
Iteratioal Power, Electroics ad Materials Egieerig Coferece (IPEMEC 205) etwork Mode based o Multi-commuicatio Mechaism Fa Yibi, Liu Zhifeg, Zhag Sheg, Li Yig Departmet of Military Fiace, Military Ecoomy
More informationTotal Harmonics Distortion Reduction Using Adaptive, Weiner, and Kalman Filters
Wester Michiga Uiversity ScholarWorks at WMU Master's Theses Graduate College 6-2016 Total Harmoics Distortio Reductio Usig Adaptive, Weier, ad Kalma Filters Liqaa Alhafadhi Wester Michiga Uiversity, liquaa.alhafadhi@yahoo.com
More informationIntegrated Detection Method of Underwater Acoustic Fuze Based on IEMD, VIFD and ED
Advaces i Applied Acoustics Vol. Iss., November Itegrated Detectio Method of Uderwater Acoustic Fuze Based o IEMD, VIFD ad ED Qiao Hu Sciece ad Techology o Uderwater Iformatio ad Cotrol Laboratory, Xi
More informationICM7213. One Second/One Minute Timebase Generator. Features. Description. Ordering Information. Pinout. August 1997
August 997 Features Guarateed V Operatio Very Low Curret Cosumptio (Typ).... µa at V All Outputs TTL Compatible O Chip Oscillator Feedback Resistor Oscillator Requires Oly Exteral compoets: Fixed Capacitor,
More informationAntenna Diversity Techniques for a Single Carrier System with Frequency Domain Equalization An Overview
Atea Diversity Techiques for a Sigle Carrier System with Frequecy Domai Equalizatio A Overview. Witschig, G. Strasser,. Weigel, A. Spriger Istitute for Commuicatios ad Istitute for Techical Electroics
More informationWAVE-BASED TRANSIENT ANALYSIS USING BLOCK NEWTON-JACOBI
WAVE-BASED TRANSIENT ANALYSIS USING BLOCK NEWTON-JACOBI Muhammad Kabir McGill Uiversity Departmet of Electrical ad Computer Egieerig Motreal, QC H3A 2A7 Email: muhammad.kabir@mail.mcgill.ca Carlos Christofferse
More informationCAEN Tools for Discovery
Applicatio Note AN2506 Digital Gamma Neutro discrimiatio with Liquid Scitillators Viareggio 19 November 2012 Itroductio I recet years CAEN has developed a complete family of digitizers that cosists of
More informationMeasurements of weld geometry using image processing technology
Measuremets of weld geometry usig image processig techology Chia-Lug Chag a, Ye-Hug Che b a Departmet of Mechaical Egieerig, Natioal Yuli Uiversity of Sciece ad Techology, Douliu, Yuli, Taiwa, ROC b Departmet
More information