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


Comments on Gigapan: Sugar Loaf - 0.15 Terapix


Gigapan Comments (37)

Toggle Minimize gigapan_comment
  1. Gerald Donovan

    Gerald Donovan (October 06, 2010, 07:18AM )

    @nodalninja 1TP is a long way off. Rio-HK are shooting at 800mm on what I believe is the highest pixel density sensor currently on the market. Shoot a 360*60 and you're looking at around 0.3TP. You realistically can't increase the focal length - some would argue 800mm on a crop sensor is too long already, so to hit 1TP this way, you'll have to wait for the crop sensors to hit around 50GP, or FF sensors to hit 100GP. Some way off I reckon. Unless of course there's another way... ;)

  2. Dave Belcher

    Dave Belcher (October 05, 2010, 05:44AM )

    Quite the accomplishment! I'd like to try stitching this on my Mac laptop :-)) Although the image is great in terms of quantity, it certainly lacks overall quality. Try re-shooting it at a smaller scale and see if you can assemble a higher quality image...one worthy of printing.

  3. The Rio de Janeiro - Hong Kong Connection

    The Rio de Janeiro - Hong Kong Connection (October 03, 2010, 05:21AM )

    I agree! Perhaps we should measure the number of wavelet coefficients with high energy (disregarding noise).

  4. Brad Templeton

    Brad Templeton (October 02, 2010, 03:08PM )

    Actually, I don't think "original pixels" is all that relevant to these panos, all that matters is the stitched result, and not including any missing sections. But my view again is that total gigapixels is not the measurement. To do the proper measurement I think you would need to reduce any image until it qualifies as "sharp." That could be subjective but there are objective mans to do this, such as reaching a point where there is high contrast between adjacent or nearby pixels at many places in the image. Not that anybody here is doing this (except those who want to get past the 50mpix minimum) but clearly somebody trying to get maximum gigapix could just expand a picture as large as they like and declare it the largest. You really want to count the number of high-contrast gigapixels.

  5. The Rio de Janeiro - Hong Kong Connection

    The Rio de Janeiro - Hong Kong Connection (October 02, 2010, 02:48PM )

    Alfred, I am looking forward to test your application.

  6. Alfred  Zhao

    Alfred Zhao (October 02, 2010, 02:07PM )

    Just throwing some numbers: 12238 images means 58 rows and 211 columns. Considering a spherical model and the camera is sitting at the center. For a horizontal FOV of 221.5 degree, at the equator, each horizontal image covers 1.05 degree. For 400mm 2X tele-converter, the lens horizontal fov is 1.6113 degree. Therefore, minimum horizontal overlap and vertical overlap is 35%. The max horizontal overlap occur at the -32 degree and the overlap is 37%. Each image contribute from 41% to 42% optical pixels to the overall image. For 18M sensor (Canon T2i, Canon 7D...), the total optical pixel count is estimated to be 90 to 92 G. The raw KRO file size is 608GB, 152G pixels. If properly trimmed (average trimmed version retains 85% of the raw pixels), the estimated trimmed size is 520GB, 130G pixels. The estimated optical pixels are 92G pixels. I will create a web application for the community to get some standard measurements. Will post the link shortly. Alfred

  7. Brad Templeton

    Brad Templeton (October 02, 2010, 01:18PM )

    There are technical processing milestones to be had, but I agree with "Alfred" that we should be aiming for metrics that go beyond the raw number of pixels, just as people have finally figured out with digital cameras that more megapixels does not necessarily mean better. For example, I could put my camera on my 1200mm telescope on my Merlin, and deliver you a bigger image with sucky stitching of the blurry results. (Or even not so blurry results with a better telescope.) But I would not think that rated as a milestone. It is indeed right to consider the quality of the pixels, the quality of the stitching, the quality of the light, the clarity of the air and of course how interesting the scene is in judging a milestone. (gp.org is sort of trying to measure how interesting the photo is with the explore scores, but unfortunately they are highly biased towards images that are already in the most popular list and thus get more exposure to being explored. I am not sure of a good way to fix that, though.)

  8. Bill Bailey

    Bill Bailey (October 02, 2010, 08:08AM )

    I believe these "are" milestones - being the first to reach what others have not been able. This takes an incredible amount of work and shows us what is possible with given equipment. Once milestones are reached and benchmarks established many will seek to improve and enhance and the same levels. This is one area where technology is finding it difficult to meet the demands of the end users. It wasn't long ago that someone claimed BIG at 10GB. It won't be long now before we see someone break the 1TB milestone - RioHK is 15% of the way :-)

  9. Alfred  Zhao

    Alfred Zhao (October 02, 2010, 07:36AM )

    Exploring the limit of panorama photography is a very exciting experience. I am building a tool to theoretically estimate the true optical pixel amount of a super large panorama. There are at least five different quantitative measures of a panorama in terms of size. 1. Initial RAW file size 2. Initial RAW pixel count 3. Trimmed RAW file size. 4. Trimmed RAW pixel count 5. Estimated optical pixel count (depending on the image sensor size, overlapping amount, image array size, vertical starting angle and fov of the array). In addition, I think the community needs to start thinking about how to improve the quality of image as well. You will be surprised to see that an image of 10 times or even smaller may carry the same amount of information as the full size one. eg. an 100G pixel image may look no different than a 10G pixel version.

  10. The Rio de Janeiro - Hong Kong Connection

    The Rio de Janeiro - Hong Kong Connection (October 02, 2010, 07:11AM )

    It is sad to see that some people like to diminish the accomplishments of others:( But, when we make our work public, we have to be open to this kind of criticism...

  11. The Rio de Janeiro - Hong Kong Connection

    The Rio de Janeiro - Hong Kong Connection (October 02, 2010, 06:37AM )

    Alfred, we didn't experience that limitation, except that Autopano Giga took a long time to load all images. We stitched the panorama on a Dell workstation with the following configuration: 2x Pentium Xeon QC X5550 @ 2.66GHz; 24GB RAM; 1TB HD; Ubuntu 10.04.

  12. Jürgen Schrader

    Jürgen Schrader (October 02, 2010, 06:32AM )

    Well, yes, it's big and it's truely been a great effort to get there. But from the photographers point of view I doubt if one would want to call that a milestone.

  13. Alfred  Zhao

    Alfred Zhao (October 01, 2010, 07:30PM )

    Interesting, how did you load all images? Autopano Giga can only process up to 11000 images. Did you use all the images or there is some trick enable Autopano Giga to process more. What kind of computer do you use to make this happen? Thank you!

  14. The Rio de Janeiro - Hong Kong Connection

    The Rio de Janeiro - Hong Kong Connection (October 01, 2010, 06:09PM )

    We took 12238 pictures and used Autopano Giga.

  15. Raymond Goldfield

    Raymond Goldfield (October 01, 2010, 05:40PM )

    Wow, how many pictures did you take and what software do you use to stitch this panorama?

  16. Bill Bailey

    Bill Bailey (October 01, 2010, 03:46PM )

    Looks like you're the first to break 100BG!! Next Milestone = 1TB :-) Congrats for sure!

  17. Gerald Donovan

    Gerald Donovan (September 30, 2010, 09:51PM )

    Wow. Congrats!