1. Skip to navigation
  2. Skip to content
  3. Skip to sidebar

Shanghai Skyline - Stitched From 12,000 Pictures - 上海风景线 - 由12000张图片拼成 by Alfred Zhao

Want to add this gigapan to your favorites? Log In or Sign Up now.

Log In now to add this Gigapan to a group gallery.

About This Gigapan

Taken by
Alfred Zhao Alfred  Zhao
Explore score
272.31 Gigapixels
Date added
Dec 17, 2010
Date taken
May 25, 2010

EPIC Pro + Canon 7D + Canon 2X...


Shanghai, one of the most populous cities in the world, has never been pictured in such enormous detail. Let the world experience this fast growing city and understand China is always my dream. Through my telephoto lens, I wish you, the viewer, can blend into the daily life in Shanghai and enjoy your virtual visit.

Before being introduced to Gigapan, I used manual panorama head (Manfrotto 303SPH) and fisheye lens to take spherical panorama and build virtual tours for my clients. In late 2009, I learned about the Dresden 26 Gigapixel image (www.dresden-26-gigapixels.com/dresden26GP Will open in a new tab or window) created by the world renowned German photographer, Holger Schulze. It was then the world largest image. This compelling image absolutely fascinated me. Until today, I still think the Dresden image is one of the most important milestones in man’s pursuing of the world largest photograph. A few months later, a new world record was released; it further intrigued me, I decided to give it a try myself.

In March 2010, the first batch of Gigapan EPIC Pro was still on pre-order. Without any hesitation, I boldly placed a pre-order and received this precisely engineered yet inexpensive device in early April. Other necessary equipments were also ordered, by mid-April, all the hardware was ready.

To build the world largest image is not a trivial project; there are many challenges and barriers. To name a few, the density of the camera’s image sensor must be high; the optics must have long focal length yet still light enough to be maneuvered by a robotic panorama head; large amount of images must be transferred directly onto a laptop; the devices must be powered for extended hours; the photo location must have large amount of details to avoid empty pixels such as sky, water surface; the weather must be calm so that the camera is less likely to vibrate due to wind; the light condition must be consistent throughout shooting… I know this is an extremely difficult undertaking, but it is worth the effort.

My first attempt was on April 22nd, 2010, the day was partly sunny. I planned to take an image set of 9,000 pictures, but had to stop because of the change of weather condition. I ended up with a set of only 1,500 images. Spring in Shanghai is often cloudy or rainy; I have to wait for the perfect weather. The second time window appeared on April 29th, 2010. That day, I was able to complete a set of 9,000 images; however, I spent too much time to setup the equipment and didn’t start shooting until 10:30am. When I was done with the image set, it was almost dark. The image set suffered significant lighting change and only the first 6,000+ images could be used. Back in April 2010, once stitched, even 6,000 images should set a world record; however, I saw the possibility of acquiring even larger image set. Finally, the third time window appeared on May 25th, 2010. Learned from the previous experience, I started very early and got everything ready by 8:00am. The shooting started at 8:30am, with an average of 2.4 seconds per image, I took 12,000 images and finished shooting before dusk. Bingo, now I have the raw data.

Acquiring the images was just the first step towards the goal, to stitch it, post-process it and display it were a long way to go. Since this was a side project, I didn’t have much time to work on it. As months went by, I watched people setting new world records almost every month throughout 2010. This year is gigapixel image year. I never imagined such fierce competition. Finally, my image was completely stitched on Sept 20th, 2010. The raw stitch was a 1.24 TB image file. I wrote several tools for post-processing and finally generated the result image of 1,089,248,410,452 bytes (1.09 TB). The 1/1000 thumbnail of the un-cropped raw stitch is also on Gigapan at www.gigapan.org/gigapans/61595/ .

Post-processing and uploading took me another three months, now it is finally open to public. This is not the end of my panorama journey, it is a new start, challenging the limit is an infinite process. New records will appear in the future, it is only a matter of time.


Image dimension: 887276 (w) x 306908 (h)
Date taken: May 25, 2010
Shooting time: 8:30 am – 4:30 pm
Date stitched: Sept 20, 2010
Date open to public: Dec 20, 2010
Raw image size: 1.24 TB
Raw cropped image size: 1.09 TB
Total pixels: 272.31 G
Horizontal viewing angle: 175 degree
Vertical viewing angle: 65 degree (+5 degree to -60 degree)
Total shots: 12,000 (150 columns and 80 rows)
Lens: Canon 400mm F5.6 and 2X tele-converter
Camera: Canon 7D (18M pixels per image)
Estimated overlap: 28% (set to 25%)
Average time per image: 2.4s
Estimated optical pixels: 112G pixels
Projection: spherical
Related images
This is the man cave (office) where I stitched this image. www.gigapan.org/gigapans/68015/
My other Gigapan images ordered by size

www.gigapan.org/gigapans/66287/ (Artist Point – Yellow Stone National Park – 110G pixels)
www.gigapan.org/gigapans/59267/ (Park Avenue – Arches National Park – 77G pixels)
www.gigapan.org/gigapans/57941/ (My Backyard – Round Lake, IL – 43G pixels)
www.gigapan.org/gigapans/58980/ (Lower Delicate Arch Viewpoint – Arches National Park – 15G pixels)
www.gigapan.org/gigapans/59095/ (Balanced Rock – Arches National Park – 9G pixels)
My favorite images taken in China

www.gigapan.org/gigapans/66837/ (Guilin)
www.gigapan.org/gigapans/66909/ (Guilin)
www.gigapan.org/gigapans/67296/ (Lujiazui’s night – Shanghai)
www.gigapan.org/gigapans/66841/ (Liu Garden – Suzhou)
www.gigapan.org/gigapans/66834/ (Zhuo Zheng Garden – Suzhou)

Due to the enormous size of this panorama image, uploading onto Gigapan is also a non-trivial task. I received great help from the Gigapan team, especially Paul Heckbert, Vanessa Constanti, and Randy Sargent.

Gigapixel photography is expensive and time consuming, without the un-conditioned support from my family, this wouldn’t be possible.
More Information

For more details about the creation of this image as well as my panorama related research, please visit www.shanghai-272-gigapixels.com Will open in a new tab or window/ . In addition, I will post the updates on twitter twitter.com/alf168 Will open in a new tab or window , for those who are interested in my panorama journey; you are welcome to follow me.

The creation of this panorama is intended to explore the limit of photo equipment, computer hardware, network resource and various other contributing factors that limit the size of a panorama image. This is by no means a perfect image; there are many aspects to be improved.



上海是全世界人口最多的城市之一,之前没有任何一张照片以如此多的细节来描绘这个城市。这是一张关于上海的“清明上河图”。我拍摄此图的目的是让世界来感受这个快速发展的城市,让世界来更多地了解中国。通过我的超长焦镜头,我希望您作为一个访客,能够融入到上海的都市生活中,来体验一个普通上海人的一天。 我作为一名久居海外的华人,仅以此图献给我的故乡-上海。

在接触Gigapan之前,我使用Manfrotto 303SPH手动云台和鱼眼镜头来拍摄球形全景并且为我的客户制作虚拟旅游图。2009年末,一个偶然的机会,当我在浏览网站时,我看到了由德国著名摄影师Holger Schulze拍摄的26G像素的德莱斯顿(Dresden)全景 (www.dresden-26-gigapixels.com/dresden26GP)。 Will open in a new tab or window 当时这是世界上最大的图片。这幅震撼人心的图片让我很久不能平静下来。直到现在,尽管很多摄影师已经突破了26G像素的图片,我仍然认为这是具有里程碑意义的重要图片之一。 几个月后,又有新的世界纪录问世。 自此,我决定,我要为中国也做一幅这样的图片,希望能为中国人争一口气,在国际竞争中占一席之地。

2010年3月, Gigapan EPIC Pro云台还没有正式上市,在没有看到和试用过这个设备的时候,我毫不犹豫地预定了一台。四月上旬,第一批EPIC Pro上市了,我得到了其中的一台。这是一部非常精密的摄影设备,我爱不释手。在四月中旬我回上海出差之前,全部设备都准备就绪了。

想要创造新的世界纪录并不是一件容易的事情,有很多的限制因素。这包括,相机必须有最高的像素密度;所使用的镜头必须是在云台承载范围内焦距最长的;图片必须直接存储到计算机上;拍摄现场必须有办法为设备长时间供电;拍摄场景必须有丰富的画面,要避免大量天空或水面的空白像素;天气必须平静,这样才不会有风带来的震动;光线条件必须在整个拍摄过程中保持稳定… 尽管有很多的困难,我觉得这是值得去挑战的事情。


可以说,拍摄还只是万里长征的第一步,拼接、后期处理、和显示还有很多困难。由于是一个附属项目,我的时间投入很有限,所以进展很慢。而与此同时,2010年,超大图 的世界纪录在不断地被打破,几乎每个月都有新的记录诞生,我只好望洋兴叹。好在我的图片还足够多,但是时不待我。我完全没有想象到竞争会如此激烈。9月20日,图片终于拼完了。未经后处理的图片是一个1.24 TB的文件。我写了一些工具软件进行后期处理,最终图片为1.09 TB。未经后处理的图片的 1/1000缩略图的网址是http://www.gigapan.org/gigapans/61595/ 。



图片尺寸: 887276 (w) x 306908 (h)
拍摄日期: 2010年5月25日
拍摄时间: 上午8:30 – 下午4:30
拼接完成日期: 2010年9月20日
正式发布日期: 2010年12月20日
未经后处理的图片文件尺寸: 1.24 TB
剪裁后的图片文件尺寸: 1.09 TB
总像素量: 272.31 G
水平拍摄角度: 175度
垂直拍摄角度: 65度 (+5 度到 -60度)
全部图片数: 12000张 (150 列, 80行)
镜头: 佳能400mm F5.6超长焦镜头和 2X长焦增倍镜
相机: 佳能 7D (18M 像素)
重叠量: 28% (设置为25%)
每幅图片平均拍摄时间: 2.4秒
光学像素量: 112G像素
投影方式: 球形
这是我拼图时所使用的办公室的图片: www.gigapan.org/gigapans/68015/

www.gigapan.org/gigapans/66287/ (美国黄石国家公园 – 110G 像素)
www.gigapan.org/gigapans/59267/ (Park Avenue – Arches National Park – 77G像素)
www.gigapan.org/gigapans/57941/ (我的后院 – Round Lake, IL – 43G 像素)
www.gigapan.org/gigapans/58980/ (Lower Delicate Arch Viewpoint – Arches National Park – 15G 像素)
www.gigapan.org/gigapans/59095/ (Balanced Rock – Arches National Park – 9G 像素)

www.gigapan.org/gigapans/66837/ (中国桂林- 两江四湖)
www.gigapan.org/gigapans/66909/ (中国桂林-世外桃源)
www.gigapan.org/gigapans/67296/ (上海外滩远眺陆家嘴)
www.gigapan.org/gigapans/66841/ (苏州留园)
www.gigapan.org/gigapans/66834/ (苏州拙政园)

由于图片尺寸极大,上传至Gigapan网站并非易事,感谢 Gigapan团队给予的巨大帮助,尤其是Paul Heckbert, Vanessa Constanti, 和 Randy Sargent。


更多关于制作这幅图片的信息和我在全景摄影方面的研究成果,我会逐渐在我的网站上发布 ( www.shanghai-272-gigapixels.com Will open in a new tab or window/ ) 。 我会把更新信息发布在twitter上( twitter.com/alf168 Will open in a new tab or window ) , 欢迎加入我的twitter群。


Gigapan Comments (84)

Toggle Minimize gigapan_comment
  1. Alfred  Zhao

    Alfred Zhao (January 05, 2011, 06:27AM )

    地图标注的问题,我看了一下,发现谷歌的卫星图和地图之间有相当大的差别,看来谷歌中国地图的精度不是太理 想。也就是说,如果按地址找到的位置与按卫星图找到的位置是不同的。

  2. abalone

    abalone (January 04, 2011, 05:02PM )

    Alrfed, 看了你的图片,觉得非常震撼,你是怎么拍出来的?我们用傻瓜相机也能拍吗?还有就是,你在地图上的标注好像 与图片对不太上,我英语能看懂,但是写不好,就直接用中文了。

  3. Alfred  Zhao

    Alfred Zhao (January 03, 2011, 05:57AM )

    Taaplari, to answer your question regarding the zoom level. I think Gigapan's max zoom level is 1:9, meaning, a pixel is zoomed to 3x3. It is easy to setup a simple experiment. Open up your gigapan image side by side with your source image. Zoom your source image to 1:1 in your preferred image viewer and then zoom gigapan to its max level, measure the difference. You should be able to find out the max zoom level.

  4. Giga Works

    Giga Works (January 03, 2011, 12:28AM )

    Since all the Gurus are here, I would like to share with you my latest Gigapan of my beloved city and Hometown Beirut. Awaiting for your comments also! www.gigapan.org/gigapans/67894/

  5. Taaplari

    Taaplari (January 02, 2011, 10:48PM )

    Almost every picture here is "too near" when zoomed to maximum. So in that level there is usually 5-10 times too many pixels in the pictures, because there is no added informarion after certain zoom level. I'm interested to hear in what zoom level is pixels calculated? If resolution is calculated in maximum zoom level, this picture would be near correct: www.gigapan.org/gigapans/50243/ Sorry my bad English.

  6. Alfred  Zhao

    Alfred Zhao (January 02, 2011, 04:22PM )

    This is the 360 degree panorama ( www.gigapan.org/gigapans/68015/ ) of my office. I focused on the screen, therefore, the bookshelf is slightly out of focus. For higher quality large interior panorama, focus stacking became a must. Does anyone know how to automatically shoot panorama with focus stacking.

  7. Alfred  Zhao

    Alfred Zhao (December 29, 2010, 12:09PM )

    Dave, it will be a fun project for you. According to my calculation, with your telescope lens and a Hasselblad H3DII 50, you will get around 700G optical pixel for the same field of view as in this picture. The total amount of image pixels will definitely into the Terapixel range. However, there will be other constraints, with the same shooting speed, 2.4 seconds per picture, you will need 16 hours to complete a similar fov. Even 1.6 seconds per picture, you will need 11 hours. I won't recommend any speed faster than 2 seconds per picture. Another constraint is the robotic head, will it be able to handle the weight? Also, DOF will be very short for a telescope. You can try, I will be eager to see a sample image set. This is about exploring the limit, right? Should be very fun!

  8. Dave Belcher

    Dave Belcher (December 29, 2010, 04:32AM )

    Congratulations, Alfred. A mighty accomplishment. I'm having thoughts of creating an image of around 1 terapixel or even larger. I have an optical instrument of 2540mm focal length @ f/14 (Questar Seven) which when mounted on it's altazimuth/equatorial mount can be coupled to a medium format camera such as a Phase One, Pentax or Hasselblad. Field of view is 1.2 degrees with an image circle of nearly 75mm. I shall be running tests in the spring. Nice thing about the Questar is it's very hi resolving power of less than .2 arc seconds and flat field with no vignetting.

  9. Raymond Goldfield

    Raymond Goldfield (December 26, 2010, 08:28PM )

    Funny emperor analogy, when emperors find out they don't have cloth, for sure, they are going chase after you.

  10. Alfred  Zhao

    Alfred Zhao (December 26, 2010, 12:03PM )

    It is a scalar in vector space, not a direction in vector space. These are different.

  11. Ronnie Miranda

    Ronnie Miranda (December 26, 2010, 11:58AM )

    Alfred, I'm not arguing but rather seeking clarification. And I think you made it clear. Using your very own logic and argument: if a compass, for example, is giving us a wrong reading, then "no one should alter the instrument they use if that's the only instrument they have" in order to remain consistent and loyal to the incorrect data even if the compass is pointing us in the wrong direction.

  12. Alfred  Zhao

    Alfred Zhao (December 26, 2010, 07:28AM )

    One more comment, I only stated the fact on the wikipedia entry, the image size in terms of pixel count is accurate, I also stated the optical pixel count, it is to the best understanding of the estimation model. Therefore, the entry is listed according to fact, no imagination, no data manipulation. It is the world record in terms of both image size as well as optical size estimation, there is no dispute at all. I think someone early in the thread has already acknowledge that.

  13. Alfred  Zhao

    Alfred Zhao (December 26, 2010, 07:16AM )

    Ronmir, the threads conversation is very clear. The reasoning is very simple. No one should alter the instrument they use. Otherwise, how does people know if this person is loyal to the data or not? Technology discussion trying to find out why the instrument does not produce data accurately is a completely separate issue. What's the model used to estimate the theoretical result is yet another topic. Do not mix the concepts together. When all the "emperors" do not wear cloth, why someone wear cloth (clearly state the fact) becomes the target? Now, every "emperor" should dress up, tell the world how much data they have manipulated. Regarding the wikipedia entry, I do not know about anything before 2007, but after 2007, every single one are produced by Autopano. Go ahead argue with all those guys now.

  14. Ronnie Miranda

    Ronnie Miranda (December 26, 2010, 06:40AM )

    The Google PTGUI group thread on this topic is here groups.google.com/group/ptgui/brow se_thread/thread/67924e4e4a3373d4 Will open in
a new tab or window

  15. Ronnie Miranda

    Ronnie Miranda (December 26, 2010, 06:20AM )

    First of all, this is not meant to be critical or negative. I can't seem to get my head around some of the reasons and arguments being presented here. Alfred, you have categorically stated that the goal of this image was to achieve a world record - pure and simple. No one is questioning all the hard work, effort, time, and even money involved in creating this goal. In fact, everyone here can appreciate all that. What bothers some people is, if it is a known fact that Autpano outputs the wrong resolution which you have personally acknowledged in this post , quote "Autopano Giga does inflate the image" and also in the Google PTGUI group , why then did you say that "there is really nothing wrong with the Autopano tool?" This is a contradiction. You also said "since we all use the same tool, we should use the same baseline." If a tool is known to be flawed, why insist on using it much so even as a baseline? Example, if a compass is giving us a wrong reading and pointing us in the wrong direction, should we still continue using it because everyone else does so or because that's how the manufacturer designed it??? That's totally ridiculous. It doesn't make any sense especially for "scientific reasons." You argue that "Autopano Giga has its reason for interpolation" and insist that we should stick to it. At the same time, you complain about the GigaPan website lacking a random image rotation mechanism for the most popular image. But doesn't GigaPan have its reasons as well for computing the explore score and ranking the most popular Gigapan? If you can accept the flawed interpolation of Autopano, why can't you accept the sensible ranking system of GigaPan? Is it because it doesn't serve your purpose? You also said "the community really should put an end to the arm race." Aside from yourself, who ever said that this was a competition? It seems that you have been obsessed with this non-existent competition from the start which resulted in this disputed 272GP. You further emphasized this by declaring "I feel no obligation to make any additional changes. I also should not make revision to their (Wikipedia) entry." Is it because you already "achieved" the world record that now you are calling an end to the "race" and yet, ironically, are already planning your next bigger project with a Mark 1Ds III with 800mm f/5.6? In fairness to Gerald, he clearly stated that Dubai 45 was only intended as a technical test. Therefore, he wasn't out to "compete" with anyone. In contrast, you categorically stated that the Shanghai image was created for the sole purpose of competing and breaking the world record. There seems to be a lot of contradictions, inconsistencies and self-serving reasons. Yes, Sevilla, Sugar Loaf, Kolor, etc. should correct their calculations too - but that's a separate issue. By insisting to use the results of a flawed tool, aren't you just perpetuating the problem - all for the sake of claiming a world record??? Again, everyone here including myself can appreciate and recognize all your hard work in creating this gigapan. There is no doubt about that. What most people are trying to understand is the real motivation to all these. And it seems to be for the fame, the glory and the prestige - which has probably now backfired since there is a dispute going on. There is nothing wrong with achieving a world record. But the issue is about providing accurate representation of your data and results. It is not a question about being fair but rather about doing what is right. If you do so, you'll probably earn more credibility and respect. If fairness is the issue, then it is also not fair for us who do the right thing if you misrepresent your work. As stated by Greg, Wikipedia "is a resource that many people depend on from academics to reporters." Therefore, any entry in Wikipedia must be and should be accurate and legit. I suggest that all so-called "world record" entries that have used Autopano, should be removed until their calculations have been corrected and properly represented. Sorry for the long post. Again, I don't mean to be critical. Just trying to understand and make sense of all these.

  16. Alfred  Zhao

    Alfred Zhao (December 25, 2010, 08:35PM )


  17. Alfred  Zhao

    Alfred Zhao (December 25, 2010, 08:32PM )

    I went to a party tonight and one of my friends suggested a fun project, Can anyone find out how many AC units are in this picture and what's the distribution across all brands. I will post this on other forums and see if this is going to be interesting or not. Will also translate this into Chinese and see if someone will respond.

  18. Chris Zhang

    Chris Zhang (December 25, 2010, 06:20PM )

    Great Job! Alfred. I visited Shanghai about two months ago and I was really impressed by how Shanghai has been changed in last 20 years. Your project provides a very unique view of what a visitor will never be able to see by him or her self. Thanks.

  19. Alfred  Zhao

    Alfred Zhao (December 24, 2010, 09:18PM )

    An interesting calculation to share with the community. The theoretical upper limit for 7D 800mm lens and full sphere panorama is 682G pixels. No chance to get into tera pixel range. Even with newer higher density sensor being introduced, the lens probably won't have the resolving power. Besides the number, there is not much additional value. Probably the community really should put an end to the arm race. Instead, concentrate on improving the image quality or getting into more interesting photo subjects.

  20. Alfred  Zhao

    Alfred Zhao (December 24, 2010, 08:18PM )

    Here is the discussion on PTGUI google group -------------------------------------------------- -------------- The real pixel count of a pano can be computed directly from the angular size of the pano and the angular size of one pixel, without any guesswork about overlaps or projections. Here is an example. ----- Your pano covers 175 x 65 degrees and was taken with an 800mm lens on an EOS 7D. Canon's specs give the width of the sensor on the 7D as 22.3mm, and the horizontal pixel count as 5184, so the width of one pixel is 22.3 / 5184 mm. With your 800mm lens, that covers an angle of arcsin ( 800 / pixel width ) which comes out to 0.000311 degrees.----- There are 562,232 pixels in 175 degrees, and 208829 pixels in 65 degrees; product 117,410,395,104 pixels - 117.4 Gpixels. ----- Regards, Tom (note: Tom Sharpless) ----- On Dec 23, 8:14 pm, Sacha wrote: > Fascinating mess in the resolution argument. I've thought about it for > awhile. I think perhaps it should be based on total images subtracting > overlap and using the mtf of the lens. Final pixel size is irrelevant > for obvious or numerous reasons. > > Unsure about generating super size 360s buy I think a giga is wildly > used for a reason. a good area for joost to adress. > > On Dec 23, 5:44 pm, "Alfred" wrote: > > > > > > > > > Hi, there, > > > First of all, happy holidays to every one! I wish all of you the > > best in the coming 2011. > > > I recently released a large panorama on Gigapan > > athttp://www.gigapan.org/gigapans/6 6626/. This panorama received > > many serious debate about the true pixel count. Since I used > > Autopano, my argument was, when there is no standard way to calculate the total pixel count, we should not manipulate the software output, rather, everyone use the same baseline. > > Then everyone should also provide theoretical estimation of the > > optical pixel count. In this case, 112G pixels. > > > The questions here are: > > > 1. Is there a way to stitch super size panorama in PTGUI? Can > > we emit raw files so that it is not subjected to image format limit? > > > 2. Is there a way to import Autopano config file and process in PTGUI? > > > 3. Can PTGUI accurately report the optical pixel count for > > super size panorama? > > > Thanks! > > > Cheers, > > > Alfred

The GigaPan EPIC Series, Purchase an GigaPan EPIC model and receive GigaPan Stitch complimentary

Where in the World is this GigaPan?


Member Log In