【雙語(yǔ)閱讀】計(jì)算機(jī)圖像生成技術(shù).

　　電影《美國(guó)上尉》、《創(chuàng)戰(zhàn)記》和《本杰明?巴頓奇事》表明計(jì)算機(jī)圖像生成技術(shù)能以驚人的逼真手法使演員看上去比實(shí)際更年輕、老態(tài)或懦弱。 這種效果至少在那些被改造過(guò)的演員與觀眾保持合適距離時(shí)能夠?qū)崿F(xiàn)。 皮膚和面料紋理重建的困難在于它們被放大觀看時(shí)，效果并不理想。

　　Science and Technology

　　Computer-generated imagery

　　Fabricating fabric

　　How to generate more realistic images of clothes

　　FILMS like "Captain America", "Tron Legacy" and "The Curious Case of Benjamin Button" have shown that it is possible to use computer-generated imagery (CGI) to make actors look younger, older or wimpier than they actually are, in a surprisingly realistic manner. At least, it is possible if those altered actors are kept at a suitable distance from the viewer. The difficulty of recreating the textures of both skin and fabric means the fect is less convincing when seen close up.

　　The reason is that, whereas it is possible to simulate realistically the forces which make virtual skin and fabric hang, bend, flap and stretch, recreating the subtle ways they rlect light has so far proved extremely tricky. The shimmer and sheen of both fabric and skin depend on the geometry of their internal structures—the exact arrangement of threads or protein fibres. This is hard to model accurately. Steve Marschner and his colleagues at Cornell University have, though, come up with a way to get round that problem. Instead of modelling, they are copying. They are using computerised tomography (CT) to analyse the structures of fabrics at high resolution and then plugging the results into CGI. That, allied to the laws of optics and some heavy-duty computer power, seems to do the trick.

　　Computerised tomography is most familiar as a medical technique for examining people&aposs insides. Like classical radiology it uses X-rays. But because the image is constructed inside a computer using shots taken from many different directions, rather than being a single exposure recorded on photographic film, CT can capture fine detail and record soft tissues that are invisible to classical radiology.

　　Dr Marschner and his colleagues used a benchtop version of CT, developed for looking at the structure of materials rather than at human bodies, for their experiment. Employing doses of X-rays many times stronger than those used to study people, they obtained high-resolution information about small pieces of fabric. Computerised tomography allows the three-dimensional structure of the fibres in such scraps to be recorded, with all their kinks and imperfections. A number of small pieces can then be patched together into an entire garment inside a computer, in the same way that a handful of actors are turned into a CGI crowd. But because the internal structure of each bit of the garment matches that of a real piece of cloth, the way light will play on it can be calculated far more realistically than if it were just a computer model of what the interior of cloth is thought to look like.

　　Demonstrating the results of their technique at the SIGGRAPH computer-graphics conference in Vancouver this week, Dr Marschner and his colleagues showed realistic renderings of felt, gaberdine, silk and velvet. Moreover, their renderings remain realistic even when viewed close up. Sadly, skin is still beyond them. The high intensity of the X-rays involved would be too damaging for use on a living human being, and a corpse would probably not produce the right results. But once the rendering technique has been speeded up (at the moment it is still a bit slow and clunky), the swish of a virtual cloak or the doffing of a computerised hat should look far more realistic than it does now.

　　In the meantime, according to Dr Marschner&aposs colleague Kavita Bala, the technology might have an application in online retailing. At the moment, people buying clothes over the internet have only standard photographs to help them choose their purchases. Using CT-based computer graphics might, paradoxically, give a better idea of what the material an item of clothing is made from is really like than can be garnered from a boring, old photograph of the original.

　　【中文對(duì)照翻譯】

　　科技

　　計(jì)算機(jī)圖像生成技術(shù)

　　虛擬面料

　　如何生成更逼真的衣物圖像

　　電影《美國(guó)上尉》、《創(chuàng)戰(zhàn)記》和《本杰明?巴頓奇事》表明計(jì)算機(jī)圖像生成技術(shù)能以驚人的逼真手法使演員看上去比實(shí)際更年輕、老態(tài)或懦弱。 這種效果至少在那些被改造過(guò)的演員與觀眾保持合適距離時(shí)能夠?qū)崿F(xiàn)。 皮膚和面料紋理重建的困難在于它們被放大觀看時(shí)，效果并不理想。

　　原因在于盡管計(jì)算機(jī)能夠真實(shí)地模擬使虛擬皮膚及面料產(chǎn)生懸掛、彎曲、飄動(dòng)和拉伸的外力，但迄今為止想要找到構(gòu)建反射光的精妙方法仍是極為困難的事情。 皮膚和面料的微光及光澤都取決于它們內(nèi)部的幾何結(jié)構(gòu)——織線或蛋白纖維的精確排列。 這種模型很難精確地建立。 不過(guò)康奈爾大學(xué)的史蒂夫?馬施納及其同事們發(fā)現(xiàn)了一個(gè)解決此問題的方法。 他們不是建立模型而是采用拷貝技術(shù)。 他們利用電腦斷層掃描術(shù)(簡(jiǎn)稱CT)以高分辨率分析面料的結(jié)構(gòu)，然后將結(jié)果輸入到計(jì)算機(jī)影像生成系統(tǒng)。 結(jié)合光學(xué)定律和強(qiáng)大的計(jì)算機(jī)處理能力，這種方法似乎能夠達(dá)到理想效果。

　　"電腦斷層掃描術(shù)"是最廣為人知的醫(yī)學(xué)技術(shù)，主要用于人體內(nèi)部檢查。 與傳統(tǒng)放射學(xué)相同，它也使用X射線。 但是因?yàn)橛?jì)算機(jī)構(gòu)建內(nèi)部圖像時(shí)采集的是多角度拍攝數(shù)據(jù)，而不是靠攝影膠片上的單一曝光影像，所以電腦斷層掃描術(shù)能夠捕捉細(xì)節(jié)并記錄下傳統(tǒng)放射學(xué)無(wú)法看到的軟組織。

　　馬施納博士及其同事們采用了臺(tái)式版本的CT進(jìn)行實(shí)驗(yàn)，這種CT是被研制用來(lái)觀察材料結(jié)構(gòu)而非人體的。 使用比人體研究高數(shù)倍的X射線劑量，他們獲得了小塊面料的高分辨率信息。 "電腦斷層掃描術(shù)"能夠記錄零碎面料的三維結(jié)構(gòu)和所有的扭結(jié)和缺陷。 然后計(jì)算機(jī)將大量細(xì)碎面料拼湊成一件完整的衣服，采用同樣方法，可將少數(shù)演員轉(zhuǎn)換成電腦合成人像。 但是因?yàn)檫@件服裝每片面料的內(nèi)部結(jié)構(gòu)都與真實(shí)布料匹配，因此與僅是看上去象布料的計(jì)算機(jī)衣服模型相比，光線在其上的表現(xiàn)效果能夠更加逼真。

　　本周，馬施納博士及其同事們將他們的技術(shù)成果在溫哥華舉行的計(jì)算機(jī)圖形會(huì)議上展示，并向人們演示了毛氈、華達(dá)呢、絲綢和天鵝絨的逼真效果圖。 此外，他們的效果圖即使被放大觀看依然很逼真。 但可惜的是，他們?nèi)匀粺o(wú)法讓皮膚實(shí)現(xiàn)這樣的效果。 因?yàn)楦邚?qiáng)度的X射線會(huì)對(duì)人體造成很大損傷，而采用尸體實(shí)驗(yàn)可能得不到正確的結(jié)果。 但是一旦繪圖技術(shù)突飛猛進(jìn)(目前技術(shù)仍有些緩慢和粗陋)，虛擬斗篷的抖動(dòng)或電腦生成帽子的脫落動(dòng)作會(huì)比現(xiàn)在看上去更為逼真。

　　同時(shí)，馬施納博士的同事卡維塔?巴拉表示這項(xiàng)技術(shù)可能會(huì)在網(wǎng)上零售業(yè)得到應(yīng)用。 目前，人們?cè)诰W(wǎng)上購(gòu)買衣物僅能通過(guò)瀏覽標(biāo)準(zhǔn)圖片幫助選購(gòu)。 相反，基于電腦斷層掃描術(shù)的計(jì)算機(jī)圖形技術(shù)可能提供了一個(gè)更好的辦法，人們能夠看到一件衣服所使用材料的真實(shí)質(zhì)地，而無(wú)須將令人生厭，老舊的原物照片收集來(lái)觀看。

　　電影《美國(guó)上尉》、《創(chuàng)戰(zhàn)記》和《本杰明?巴頓奇事》表明計(jì)算機(jī)圖像生成技術(shù)能以驚人的逼真手法使演員看上去比實(shí)際更年輕、老態(tài)或懦弱。 這種效果至少在那些被改造過(guò)的演員與觀眾保持合適距離時(shí)能夠?qū)崿F(xiàn)。 皮膚和面料紋理重建的困難在于它們被放大觀看時(shí)，效果并不理想。

　　Science and Technology

　　Computer-generated imagery

　　Fabricating fabric

　　How to generate more realistic images of clothes

　　FILMS like "Captain America", "Tron Legacy" and "The Curious Case of Benjamin Button" have shown that it is possible to use computer-generated imagery (CGI) to make actors look younger, older or wimpier than they actually are, in a surprisingly realistic manner. At least, it is possible if those altered actors are kept at a suitable distance from the viewer. The difficulty of recreating the textures of both skin and fabric means the fect is less convincing when seen close up.

　　The reason is that, whereas it is possible to simulate realistically the forces which make virtual skin and fabric hang, bend, flap and stretch, recreating the subtle ways they rlect light has so far proved extremely tricky. The shimmer and sheen of both fabric and skin depend on the geometry of their internal structures—the exact arrangement of threads or protein fibres. This is hard to model accurately. Steve Marschner and his colleagues at Cornell University have, though, come up with a way to get round that problem. Instead of modelling, they are copying. They are using computerised tomography (CT) to analyse the structures of fabrics at high resolution and then plugging the results into CGI. That, allied to the laws of optics and some heavy-duty computer power, seems to do the trick.

　　Computerised tomography is most familiar as a medical technique for examining people&aposs insides. Like classical radiology it uses X-rays. But because the image is constructed inside a computer using shots taken from many different directions, rather than being a single exposure recorded on photographic film, CT can capture fine detail and record soft tissues that are invisible to classical radiology.

　　Dr Marschner and his colleagues used a benchtop version of CT, developed for looking at the structure of materials rather than at human bodies, for their experiment. Employing doses of X-rays many times stronger than those used to study people, they obtained high-resolution information about small pieces of fabric. Computerised tomography allows the three-dimensional structure of the fibres in such scraps to be recorded, with all their kinks and imperfections. A number of small pieces can then be patched together into an entire garment inside a computer, in the same way that a handful of actors are turned into a CGI crowd. But because the internal structure of each bit of the garment matches that of a real piece of cloth, the way light will play on it can be calculated far more realistically than if it were just a computer model of what the interior of cloth is thought to look like.

　　Demonstrating the results of their technique at the SIGGRAPH computer-graphics conference in Vancouver this week, Dr Marschner and his colleagues showed realistic renderings of felt, gaberdine, silk and velvet. Moreover, their renderings remain realistic even when viewed close up. Sadly, skin is still beyond them. The high intensity of the X-rays involved would be too damaging for use on a living human being, and a corpse would probably not produce the right results. But once the rendering technique has been speeded up (at the moment it is still a bit slow and clunky), the swish of a virtual cloak or the doffing of a computerised hat should look far more realistic than it does now.

　　In the meantime, according to Dr Marschner&aposs colleague Kavita Bala, the technology might have an application in online retailing. At the moment, people buying clothes over the internet have only standard photographs to help them choose their purchases. Using CT-based computer graphics might, paradoxically, give a better idea of what the material an item of clothing is made from is really like than can be garnered from a boring, old photograph of the original.

上12下

共2頁(yè)