Emerging Technology

The ability to stitch two photos together to make a panorama has been with us for many years. I can remember using a hand scanner and aligning and stitching photos back in the early 90's with relative success. There are also some wonderful online panoramas that you can find if you go searching and MS Encarta had some pretty neat ones too. So why would I be so excited by the release of Photosynth which is a free online photo stitching site. Sure you can put together a remarkable number of photos (it complains that it might get slow if you try to upload more than 300!), and it does do a remarkable job of putting them together in layers and in sequences. You can dive into the picture to hit a different perspective or more details close up. It has been released to the public for us to play with and has a generous file host capacity of 20gb. You can already see a number of interesting photosynths appearing.

The exciting part is not here yet, but the potential is extraordinary. It was alluded to by Blaise Aguera Y Arcas in his presentation at TED 2007 where he introduced Photosynth. At the moment we are collecting our own pictures either from the internet or from our own personal collections to create our synths. But what the programme is doing is creating visual meta data and interlinking images to one another. This portends a visual search engine of enormous significance. Imagine taking an old picture from your travel photo album, which you have forgotten to write on the back of. Where was it again? You just can't remember. Scan it, and upload into a visual search engine and it stitches it into a Synth. The resultant page that would pop out is a synth of the surrounding photos of the same location in differing perspectives, taken by different people, no doubt linked to a google earth location and map!

Photosyth at the moment shows the remarkable ability for the software to piece together a visual jigsaw, it is only the dimensions that are daunting. And that is truly only momentary given the rapid pace of processing and storage evolution. And where does that lead to, well to a time line version mapping the past seems feasible, and to sketched visual searches based on a hand drawing of what you are looking for… draw a quick sketch of something or maybe put together template objects, for example a pyramid (hand drawn or computer graphic) which becomes the search "keypix" (keyword) to find buildings with triangular fixtures. Pictures would become tagged with visual features and rich in meta data content. Multiple keypix would refine the search and a new breed of engine would be created. Look out text, there is a visual revolution on the doorstep!