GigaPixel Panorama


[Ewout] sent us some info on this Automated Gigapixel Panorama Acquisition system.  The system automates the process of taking the large amounts of images required to do gigapixel panoramics. You tell it key information, like what lens, and what percent overlap you want and the system will calculate how many images it will take, as well as the gigapixel count. The results are quite stunning, no visible seams with fantastic detail.  Interestingly, this was created for a class in embedded system design (ECE4180) at Georgia Institute of Technology and so was our post earlier today on Digitally Assisted Billiards. Is Hack a Day part of the class curriculum? It should be.

27 thoughts on “GigaPixel Panorama

  1. We planned to create our own platform to rotate around the nodal point, since this would eliminate Parallax error. However, due to budgetary reasons we used a CCTV platform. As long as you’re not making a panorama of a distance far away such as a city scape the distortion is minimized.

  2. Now just invent a system which can do this with long exposures whilst dynamically tracking a star field at night. I don’t know what purpose it would serve, but I’m sure it would be cool.

  3. “Is Hack a Day part of the class curriculum?”

    I guess you could say it is since the class curriculum is right on topic with this website. A number of 4180 project have been on this website in the past. Also, I know a lot of current and former students (such as myself) love Hack-A-Day.

    Go hack-a-day and go jackets!

  4. Just buy a Red camera and stitch the stills together later…. i think that would be faster as well. Some of the projects on hack a day are completely impractical – this is one of them.

  5. pditty, Mounting a camera nodally will eliminate any parallax issues. I built a very similar camera head a few weeks ago. Mine is nodally mounted though it sounds like the one in this article has better software in that it figures out how many pictures are needed. Mine just takes an array of pictures that always have the same number of pictures per row. That isn’t necessary for rows looking close to straight up or straight down.

    Next step for my camera is to make it a digital slit scan… I’m not quite sure how to deal with all of the data that I would generate. Each picture will be about 300 megabytes without compression. I’m looking at embedded Linux computers because those processors can probably deal with that amount of info in a reasonable time.

  6. This is great. I don’t know what the critics are on about ‘impractical’. Its fully automated, how isn’t it practical?!

    Source code is available, all that there needs to be now is a build log!

    Nicely done.

  7. Ive done this with some success making a bracket from aluminum bar stock to allow camera to rotate around the “nodal point”. Biggest problem I’ve encountered is exposure. Brightness often varies over a tremendous range when you are outside on a clear day using sunlight.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.