MIT研制高清晰度多视角裸眼3D

最近美国麻省理工学院媒体实验室描述了一种用于呈现较高分辨率3D图象的新算法。据发明者表示，这种高分辨率3D技术无须佩带眼镜，也不会减少亮度或限制观众观看的方位，与传统的自动立体显示技术类似。

Ramesh Raskar教授的相机文化团队设计了该技术的原型。该技术是将立体相机拍摄的图象分成分别适于左右眼的两幅图象，然后分别计算每一个像素垂直和水平两个方向的视差。接着，该新技术淘汰了垂直条缝视差屏障方法将原始图象传送至正确的眼睛，而是使用一种适应当前3D每一像素信息数量的自定义多角度的视差屏障。（注：祼眼3D主要工作方式是这样的：对于每帧画像，需要生成两幅画面，一个适用于左眼，一个用于右眼。这些画面被细分为很细的垂直分段并交替排列。而后用图像通过视差屏障来观察，其实主要就是一个划有很多垂直狭缝的观察屏。视差屏障被调节为可在平行方向区别人的两个眼睛。所以摆放好屏，并且如果观察者在相对于屏的正确空间方向，那么左眼将只能看到适于左眼的画面，同样右眼只看到适于它的画面。如施展魔法一般，一幅3D图象出现了。但其局限性不少。）

MIT的原型使用了两层分别计算每幅图象视差屏障的液晶显示屏，结果使得自动立体显示的图象更明亮，因为它不像传统的裸眼3D显示屏那样把每个像素的亮度都减半了。同时这一计算方法使观众变换方位时也完全能感受到图象的3D效果。（注：麻省理工小组利用双LCD方法，但不用垂直条形屏障，他们屏障是由数以千计的、与下方图像相匹配的小孔隙构成的。当图像改变，屏障也跟着变。这需要大量的数学计算，辅助以独特的算法，专为持续改变屏障取向和多种方式通光而设计。）

由于视差屏障模块低亮度耗能少，显示器本身耗能也比传统裸眼3D方法更少。但是，每个像素创建一个自定义的视差屏障使得计算复杂性大大增加，因而需要额外的电源。

为了补救这一缺陷，研究者目前正在努力简化算法的复杂性，同时设计加速芯片来抵消创建自定义视差屏障的额外功耗，使高分辨率3D显示屏得以实现。

相机文化团队成员包括Ramesh Raskar和Doug Lanman教授，博士后研究员Yun Hee Kim以及博士研究生Matthew Hirsch。

放弃使用双镜头相机未剪辑的左右图象，MIT使用如上图所示的双层LCD内容自适应视差屏障

编译：Aileen Zhu

点击参考原文：MIT prep high-def, glasses-free 3-D

《电子工程专辑》网站版权所有，谢绝转载

{pagination}

MIT prep high-def, glasses-free 3-D

R. Colin Johnson

A new algorithm for rendering higher resolution 3-D images was recently described by the Massachusetts Institute of Technology (MIT) Media Lab. The high-resolution 3-D technique is glasses-free, but does not reduce brightness or restrict viewer orientation as with conventional auto-stereoscopic techniques, according to its inventors.

Prototyped by professor Ramesh Raskar’s Camera Culture group, the technique performs calculations for both vertical and horizontal parallax on every pixel in separate left and right images from a stereo camera. Then, instead of converting the images into vertical slits that route the original images to the correct eye with a parallax barrier, the new technique uses a customized multi-perspective parallax barrier that adapts to the amount of 3-D information present at each pixel.

MIT's prototype uses two stacked liquid crystal displays with a parallax barrier that is calculated separately for each image, resulting in an auto-stereoscopic image that is brighter—since it does not half the light from each pixel as with conventional glasses-free displays—plus can be properly perceived even if the viewers head tilts from side-to-side.

Because the parallax barrier blocks less light, the display itself also consumes less energy than conventional glass-free 3-D methods. However the computational complexity of creating a custom parallax barrier for each image will require extra power.

To remedy, the researchers are currently working to simplify the algorithm's complexity, as well as create custom acceleration chips that offset the extra computational effort expended to create the customized parallax barriers that enable the higher resolution 3-D display. Details are available on the MIT Camera Culture website.

Camera Culture members include professor Ramesh Raskar and Doug Lanman, post-doctoral researcher Yun Hee Kim and doctoral candicate Matthew Hirsch.

Instead of using the unedited left and right images from a twin-lens camera, MIT's dual-stacked LCD displays uses content-adaptive parallax barriers as displayed here.

编译：Aileen Zhu