forcode想看的


forcode看过的


新技术
新趋势
奇思妙想
科学探索
科幻奇幻
资料搜集
网络研究
统计定量
社会学研究
书摘读后感
数码网络
软件评测
数据指标
实用信息
有趣的东西
房地产
网络赚钱
投资创业
新闻评论
网站经营
电影八卦
美景美人
人物朋友
情感回忆梦
forcode生活

2008-04-11

奇想录小专题:超高分辨率相机(组图)

forcode:昨天奇想录介绍了"超高速摄像技术",目前最快的摄像机每秒可以拍摄100万张照片,当然,由于拍摄速度很快,每张照片的分辨率不会很高,一般也就400×300左右,否则每秒拍摄的录像体积会大得超出了相机芯片之间的传输带宽。那么,目前像素分辨率最高的相机发展到什么水平了呢?40亿像素!!每张照片的体积是24G!!想想你手里的数码相机最多也就1000万像素吧?目前世界上最先进相机的分辨率是它的400倍,已经超出两个数量级了。像素在一定程度上可以弥补变焦能力的不足,因为单位面积内更多的像素意味着可以捕捉更多信息,在下面的组图中,我们可以看到,站在棒球场一侧看台顶部拍摄一张全场照片,放大之后,居然可以看清远方每个人的表情,这实在是不可思议,就那么一秒钟的拍摄,可以捕捉到如此丰富的信息。目前Gigapxl Project的人们计划拍摄美国海岸线的全景图片,拍摄全球各种艺术品的高清晰照片,这种超清晰成像技术短期内可能无法民用。一旦民用,将激发出不可想象的丰富创意:在照片上数数蚂蚁腿上的绒毛有多少根、搜集照片中戴眼镜的人镜片反射的高清晰图像、像观察月球环形山结构一样观察某个人脸上青春痘的色彩结构并判断其成熟度以及预测其下一次"火山爆发"的时间、利用遥控飞机携带相机绘制自己家附近的数平方公里的高清晰航拍图、利用高清晰图片的丰富性开发针对一幅图片的探索寻宝游戏(所有寻宝线索隐藏在一张图片中)……
本文全文地址:
http://www.qixianglu.cn/627655.html


奇想录小专题:超高速摄像技术(视频和组图)
http://www.qixianglu.cn/627628.html


世界上最高分辨率——40亿像素的相机照出的相片
两位设计师和他们的杰作

目 前民用数码相机,象素最高的也只有1600万,但是40亿象素的数码相机你见过吗?美国航天局为科研而特别研制的相机就使用了40亿象素的感光元件,40 亿象素可以拍摄出88000x44000的超大分辨率,这种尺寸的数码照片,如果不压缩的话,一张照片的容量将达到惊人的24GB!下面我们就去看一下这 40亿象素能带给我们什么样的感受。






海岸大桥全景,你注意到这还有人了吗?.G$P [ P o,G g
按此在新窗口浏览图片
放大 A
按此在新窗口浏览图片

最后一副图中的人你能从原图中找出来吗?此时照片只有原始照片画幅的0.05%!
按此在新窗口浏览图片


Gigapxl Project:
http://www.gigapxl.org/


Our Vision

Defining the upper limits of large-format film photography, digital scanning and image processing, custom-built Gigapxl™ cameras capture images with unprecedented resolution.

It would take a video wall of 10,000 television screens or 600 prints from a professional digital SLR camera to capture as much information as that contained in a single Gigapxl™ exposure.

The Project's near-term goal is to compile a coast-to-coast Portrait of America; photographing in exquisite detail the cities, parks and monuments of the USA and Canada.

A longer term goal is to create for future generations a world-wide archive of vanishing cultural and archaeological sites.

Project Overview

The advent of digital technology has caused a revolution in the way we think of photography. Before this revolution, most of us thought of photography in terms of images captured on film that subsequently would be transformed into prints by way of photographic enlargement. However, rapid proliferation of digital cameras, scanners and printers has changed this perception. Even the prevailing jargon has changed. Where previously an image was described in terms of granularity and acutance, it now more often is defined by pixel count and dpi. Meanwhile, those of us who have spent much of our lives in the pursuit of film-based photography keep asking ourselves whether such photography can survive; and, if it does, what role will it play? When we debated this question in late 2000, it seemed reasonable to assume that digital cameras with resolutions in the 10-megapixel regime would become commonplace within a few years. This would put them in head-to-head competition with 35-mm film-based technology; perhaps even displacing that technology entirely within a decade or so. On the other hand, it seemed unlikely that digital cameras with resolutions much in excess of 10 megapixels would appear in the near term. Especially unlikely would be the emergence of digital cameras with resolutions approaching 100 megapixels. As a consequence, we felt it likely that film-based large-format photography would for the time being remain unchallenged. With this in mind, we have concentrated recent endeavors upon the application of ultra-high-resolution techniques to the field of large-format photography.

In defining the term "ultra-high-resolution," we have analyzed each factor that bears upon the image forming process; especially taking into account the effects of atmospheric blurring, lens aberrations and film granularity. When it became apparent that the sought-for resolution could not be preserved via conventional photographic enlargement, the scope of our analysis was widened to include film scanners and digital printers. We concluded that, consistent with the largest practicable roll film format (9"×18"), we could expect to achieve a resolution equivalent to 1000 megapixels. Hence, came the name Gigapxl™. With recent developments, this figure approaches 4000 megapixels, but the name remains unchanged.

Subject to the limitations of human vision, a minimum of 8 square inches of print area is needed to convey the information contained in a 1-megapixel image. When scaled to 1000 megapixels, the minimum print area becomes 50 square feet. For prints made from our 9"×18" format, this equates to a print which has a height of 5 feet and a width of 10 feet. Likewise, a 4000-megapixel print has dimensions of 10 feet by 20 feet. Meanwhile, close-up sharpness matches that of a 4"×6" print from a 3-megapixel digital camera. The information content of a Gigapxl™ print can be compared to that available in a real-world scene which is viewed through a pair of binoculars. In the case of 1000-megapixel images, one would require 6X binoculars; twice this power at 4000 megapixels.

The first Gigapxl™ cameras were completed and ready for test in February 2001; the first color landscapes being produced a month later. Early images had a pixel count of 260 megapixels (20-micron scan resolution) and were printed on photographic paper. Within a year, however, the count had increased to 670 megapixels (12.5-micron scan resolution). At which level, although substantially higher resolution was being achieved on film, the pixel count temporarily became constrained by issues related to scanner resolution and the file size limits of Adobe Photoshop. Meanwhile, with second-generation cameras (combining superior lenses and a variety of focal lengths) nearing completion, we switched from photographic printing to pigment ink printing. Working closely with Adobe, issues related to Photoshop file size have slowly but surely been resolved. Meantime, collaboration with Leica Geosystems (manufacturer of the DSW500 digital scanner) is about to yield scans with a resolution of 6 microns. At which time, numerous existing negatives will be redigitized at 2,900 megapixels. By year end (2004), we expect to push scan resolution to the 5-4 micron range; the corresponding pixel counts being 4,180 megapixels and 6,530 megapixels, respectively.

Extensive viewer response to Gigapxl™ imagery was first obtained in 2003. In March of that year, a 21-foot panorama of San Francisco was exhibited at The Albuquerque Museum. Four months later, a similar image was exhibited at the Palace of Fine Arts/Exploratorium in San Francisco. Aside from general expressions of awe, feedback mainly has centered around the extent to which ultra-high-resolution adds a humanizing touch to subject material which otherwise tends to be dominated by its monumental scale. Especially it has been noted that the ability to capture the minutiae of everyday life provides a level of interest not found in conventional cityscapes.

While technical issues which relate to scanning and digital processing continue to be addressed, current efforts are focused upon the expansion of an image portfolio. At this point, we have stockpiled some 500 images; a fair proportion of which already have been scanned at the 10-micron level. Subject material for the most part typifies the American landscape. To date, photographic forays have been made to all Provinces and States, with the exception of Hawaii. Notable urban subjects include cities such as San Francisco, Los Angeles, San Diego, Calgary, Colorado Springs, Dallas, Seattle, and Denver. Work in the National and State Parks/Monuments extends to Yosemite, Point Lobos, Mount Tamalpais, Mono Lake, Monument Valley, Canyonlands, Grand Canyon, White Sands, Mesa Verde, Canyon de Chelly, the Grand Tetons, Yellowstone, Devil's Tower, Mount Rushmore, the Badlands of South Dakota, Denali, Jasper, and Banff.

In terms of the future, we have been much encouraged by the diversity of applications which continue to emerge. One of particular appeal relates to the documentation of cultural and archaeological sites which cannot be preserved and which inevitably will deteriorate with the passage of time. Many thousands of these sites are present around the world. Prime examples include entire cities such as Rome, Italy. In this instance, limestone structures which have stood for thousands of years have become the victims of acid rain. Stonemason's chisel marks, until recently clear to see, have all but vanished. Only through a massive program of ultra-high-resolution documentary photography can such details be preserved for enjoyment and study by future generations.

Technology

It began with a seemingly simple question. As a designer and builder of cameras, what could one do that has not been done before? The question came on the heels of an earlier project which involved the design and construction of an astrocamera having a resolution of 0.01 millimeter across photographic plates that are 356 millimeters square. Recognizing this to be the equivalent of 1250-megapixel imagery prompted the question of whether or not anything comparable had been achieved in the context of landscape photography. Some back-of-the-envelope calculations concerning the performance of conventional large-format cameras indicated that it had not. Thus began a quest which has become ever more fascinating; namely the pursuit of full-color panoramic landscapes which contain prodigious amounts of information. Early on, the goal was set at 1,000 megapixels. However, as technology has advanced, the bar has been raised to 4,000 megapixels; a figure that we expect to reach within the next several months. At this level, a real-world 90-degree panorama would need to be searched with 12X tripod-mounted binoculars before one could hope to accumulate an equivalent amount of information. Unsurprisingly, a host of problems have been encountered along the way. But the results have been spectacular. And so, the fascination continues.

Since the equivalent of 1000 megapixels had been exceeded with the astrocamera, it might seem that one simply could point a similar camera toward the horizon and the original goal would be achieved. But landscape photography is very different from astrophotography. When photographing through a quiescent night sky, it is not unusual to create imagery wherein each pixel equates to no more than a couple of arc seconds. Hence, a photograph encompassing one square degree of sky can have a pixel count on the order of 3.2 million. By extension, a night-sky exposure having an angular diameter of only 20 degrees can, in principle, contain 1000 megapixels. As we shall see, however, near-horizontal paths through the daytime atmosphere are vastly different from near-vertical paths at night. We find, for example, that, particularly in the context of landscape photography, the limitations imposed by atmospheric transmission drive us toward very large fields of view. Likewise, the limited resolution of color-negative film calls for very large formats. Meanwhile, in bringing these large fields and formats together, we require photographic lenses of extraordinary performance.

When issues related to lenses, formats and film have been resolved; it still remains to transform images on film into prints that maximize the information content which can be assimilated by a viewer. For this purpose, one cannot resort to contact printing, insofar as the fineness of detail in a contact print far exceeds the acuity of the unaided human eye. In fact, when working with a sharply focused negative, a 1000-megapixel print which is ideally sized for close inspection requires an enlargement ratio of at least 12X. Because the negative is large to start with, this ideal print becomes very large indeed; typically in the 50 square feet regime. Prints of this size can, in principle, be produced either by conventional photographic enlargement or by a combination of digital scanning and printing. However, even the best enlarging lenses cannot provide imagery in the 1000-megapixel range; at least not without loss of contrast relative to the original negative. Hence, in practice, there is little choice but to use high-resolution digital scanning together with some form of digital printing.

During the course of discussion, we follow the Gigapxl™ image-forming process from real-world subject to final print; explaining along the way how, at each step of the process, we strive to minimize the loss of information content. Of importance in this regard is the manner in which the losses associated with various steps are balanced such that no single step dominates the overall loss. Specific topics include the information content of images, the choice of format, the choice of film, the transmission of images through typical atmospheres, the requirements of lenses, the design of a camera which is consistent with the selected lens/format combination, the response of viewers to high-resolution imagery, and the various means for producing large digital prints.

Continue with Information Content...
http://www.gigapxl.org/technology-content.htm
超高分辨率图片集:
http://www.gigapxl.org/gallery.htm


世界上最高分辨率的显示系统

 

您想想过,在一个比HDTV的质量还要好100倍的显示屏上看一个人脸会是什么效果吗?圣迭戈的加利福尼亚大学的工程师做成了一条目前世界上最高分辨率的计算机显示系统,其像素数达到了220M像素(2.2亿像素),是全高清电视(1920*1080207万像素)的100倍还多,55个横向排列和50个纵向排列的全高清液晶面板通过光纤以太网络连接构成了这套交互式超高分辨率的显示系统。

 

这套系统将提供给地球科学、生物工程、气象预报、大脑图像等相关科学领域的科学家使用,高分辨率可以为以上学科的科学家提供一个更加高效的科研手段。

世界上分辨率最高的商业人造卫星将使用SGI技术

只有SGI Altix系统才能提供符合要求的64位高性能计算,满足国家安全和商业应用的几十亿字节大小的数据集

美国佛罗里达州奥兰多,2006年地理空间情报年会(GEOINT 2006),Booth 207,(2006年11月15日)—— 计划在2007年春发射的超高分辨率地球成像卫星,GeoEye(Nasdaq:GEOY),世界最大的商用卫星遥感公司,购买了来自SGI的高带宽、高 性能计算技术。在弗吉尼亚州杜勒斯市,新的GeoEye-1卫星的地面站,四个SGI® Altix®系统在第一个季度内交货,它们将对.41米全色(对所有可见色彩都敏感)和1.65米多 光谱(感应并记录不可见以及电磁波频谱的可见部分的放射线)的成像进行核心卫星图像的处理。GeoEye-1预计每天可以搜集700,000多平方公里的 ——数十亿字节——的高分辨率图像。作为长期为美国政府、国际政府和数量不断增长的商业客户提供图像的SGI用户,GeoEye选中SGI系统是因为来自 这颗新卫星的图像数据量将需要4倍的处理能力。

"我们目前有三颗卫星在轨运行——OrbView-2和3,以及IKONOS——但是当 GeoEye-1完全运行起来时,它将成为世界上最强大最精确的高分辨率商用成像卫星。和目前在轨的任何商用系统相比,它将在给定时间内收集更多的图 像," GeoEye的高级首席测量工程师Don Koboldt说。"我之所以选择SGI Altix是因为它拥有64位而不是32位的处理能力,并且我们所作的每个工作都需要很多的计算性能,包括传感器建模和将所有数据重新取样到地理学系统 上。更不用说需要被处理的象素数据的实际数量了。这既是一个I/O密集型(I/O-bound)问题,也是一个计算问题,而SGI在为这样的需求设计系统 方面独领风骚。"

GeoEye-1卫星拥有达到.41米分辨率的能力,简单来说这意味着,从轨道采集并由 SGI® Altix® 350系统处理的高分辨率图像将能够辨识地面上16英寸或者更大尺寸的物体。以这个分辨率,人们将能够识别出位于棒球场里放着的一个盘子或者数出城市街道 内的下水道出入孔的个数。

总体来说,GeoEye的产品被应用于范围广泛的应用中,这包括用于国防和情报部门的大型地区绘图、州和当地政府的城市规划和绘图、保险和风险管理、环境监控和灾难救援。这些图像还是在线地图搜索引擎的理想选择。

GeoEye最近获得了美国政府提供的1960万美元的合同,向几个联邦机构提供图像、增值 产品和服务,包括美国农业部(USDA)、国家公园管理局以及国家地理空间情报局(NGA)。GeoEye使用多种通用软件,并在圣路易斯办公室编写了很 多内部专用代码。为美国政府进行的大部分高端增值工作也是在这里完成的。所有的代码都是被编写为多平台的,这就使得SGI Altix系统的开放系统Linux®环境能够与公司的需求集合达到完美一致。

GeoEye-1预定在加利福尼亚的范登堡空军基地发射,借此机会公司期望能够有来自政府商业和当地政府集团的更大销售。GeoEye还特别期望能够将他们的产品扩展到其他需要高分辨率精度地图(map-accurate)图像的市场中。

"今天图像都是"软"拷贝的,也就是说,是数字格式的。你可以在任何一个图像集上精确定 位," Koboldt说,他以前在NGA里工作了十年,帮助设计了GeoEye-1地面系统的组件。"换句话说,我们将精确地知道每个东西在哪里——该目标在地 球表面地真实位置地几米范围内。如果你想要知道一些东西的位置,你可以通过使用我们的图像产品在屏幕上进行测量就能完成。除了精确性方面的改进之外,图像 变得更大,数据的数量也更大,Altix将让我们在图像大小和文件系统大小方面提高了不止一点。"

GeoEye购买了四个SGI Altix 350系统,每个系统上都配备了16颗Intel® Itanium® 2处理器。SGI Altix系统运行Novell® SUSE® Linux企业版服务器9,这就确保了Altix部署能够与现有的GeoEye应用软件相吻合。这四个SGI Altix系统将被连接到一个存储区域网络中,该网络还包括先前购买的SGI® Altix® 3700服务器。当发射GeoEye-1时,SGI® Altix® 3700将被用于图像重建操作。

"64位SGI Altix系统的性能是高分辨率地球成像卫星数据所需的大量数据处理的理想选择,"SGI的政府和防御市场部门经理Gene Gray说。"商用卫星数据公司,例如世界领先的GeoEye,政府实验室,例如NASA和NOAA,私有天气预报公司,以及天气研究和灾难救援领域所涉 及的一流大学,都依赖SGI技术的处理性能来提供保护和服务美国和世界所需的数据集。"

Seitz 6x17 Digital:神级的一亿六千万像素相机

按此在新窗口浏览图片

我们第一次看到这台相机时,下巴立刻掉到了地上 -- 还弹了两下。不,它不是我们见过像素数最多的相机,但这台是你真的可以买到的,不是远在某某不知名实验室里的超级计划。

当 然,像素数(换成熟悉的数字,160mp)不是它唯一疯狂的地方 -- 规格到了这个境界任何数字都是疯狂的。首先它的感光器大小有 6x17 公分大、 最高分辨率达 7500x21,250px、未压缩的 RAW 档一张就有 950MB 的大小、ISO 值由 500 一直到 10,000、最短快门  1/20,000 秒、还有个「可拆卸」的触控式 VGA 屏幕。近 1G 的档案大小当然不是 8GB SDHC 或甚至 64GB CF 卡所能顶 下来的,因此 Seitz 让你透过 Gigabit 以太网络将照片传到他们提供的「尖端科技电脑系统」 -- 实际上也就是台 Mac mini。

所 以这些疯狂的规格要花多少银子才能入手?较高端的「机动版」要价 45,500 瑞士法郎(30 万人民币),而(相对)简易的「工作室版」则是要价  42,300 瑞士法郎(27.7 万人民币)。这个价格还不包括镜头,但不包括那台 Mac mini 的话也太不公道了。

这台相机不消说当然会出现在 Photokina 上,但要正式上市大概要等到 2007 年初了。

按此在新窗口浏览图片

按此在新窗口浏览图片

按此在新窗口浏览图片

按此在新窗口浏览图片

按此在新窗口浏览图片

按此在新窗口浏览图片

按此在新窗口浏览图片

按此在新窗口浏览图片

按此在新窗口浏览图片

按此在新窗口浏览图片

按此在新窗口浏览图片

按此在新窗口浏览图片

--
未来新技术和新趋势的种种可能:
奇想录 http://www.qixianglu.cn
订   阅 http://feed.feedsky.com/woodphone
非主流的科幻界的非主流论坛——奇想家园:
http://www.douban.com/group/forcode/