A Whole Earth Photo Alblum
by
Randy Mears
If you could stitch together all of the photographs on the Internet, how much of the world would you be able to view? Several years ago I was thinking about this very topic and what metadata you would need to store with each photo. I came up with latitude, longitude, altitude, heading and angle of elevation. I like to think of it as extended geotagging. Other useful information such as lens specification, date and time are already in place on many photographs. While most of the photographs on the Internet currently lack the detailed location information there are already enough pictures out there to cover popular destinations fairly well.
If you want to see what I mean, check out Google Images. Just type in the name of a popular destination and look at all the pictures that you get. The "Taj Mahal" has about 210,000 images, "Giza Pyramid" has 23,000 and "Café Du Monde" has 310,000. Now think about all of the nearby images not counted here, different points-of-view that were taken in the general vicinity but not tagged as such, and you get an idea of how well covered the area would be. If properly geotagged these images could be geographically connected to the adjacent, more popular, images. Of course, many of those pictures will have unwanted items like people and cars but digital magic could ultimately handle that as well.
While the necessary metadata may be missing from the Internet’s chaotic image collection at this point, that may change as GPS chips and electronic compass chips begin to gain popularity in cell phones and digital cameras. Once we have a large collection of geotagged images we will still have lots of holes, but that’s ok. There will always be places people don’t photograph or don’t go; but the popular places will get filled in fairly quickly. With thorough geotagging as a first step we can move on to the second hurdle.
That second hurtle is the software. The key to successfully putting together a patchwork of photos is scaling, color matching and seamless photo-stitching software. In order to be useful in this grand photo stitching scheme, the solution would also need to be completely automated and operate on-the-fly. I found this article about software that can erase unwanted elements from photos and discovered that it may have many of the inner workings necessary to create a smooth looking photo from such a patchwork collection (although that isn’t exactly its intended purpose). The kinds of techniques used by this software are the kinds of things that could make stitched images look seamless, even when cobbled together from collections of Internet photo snippets. An added bonus is the software’s ability to remove unwanted elements from the resulting image (like cars, buses and recognizable people). In the end you do have to ask how real this cobbled together software enhanced image is, but does that really matter here?
While this Whole Earth Photo Album may seem a little far fetched now, it could happen; and some of your photos could be part of it - sometimes.
