Take a photograph, marking 4 (ish) control points on it. Upload the picture, the marked points, and some data value, possibly completely unrelated with the picture, to be associated with it. The picture and points serve as a key into this key-value database. Anyone coming along later can take approximately the same picture, mark approximately the same control points, and fetch the value. The server does approximate image matching, aided by aligning the control points (perhaps in 3D) of the Insertion photograph with the points in the Query.
In the real world, the control points could be subtly marked, subtle enough that only those in the know would know to query the database with a photograph of that scene. The inspiration was to allow unobtrusive graffiti through 4 small dots in a public place: the associated value could have arbitrary ostentatious data.
Next, allow the use of artificial images (with control points) as keys. One interesting way it could work, vaguely similar to geocaching, is the control points are geographically separated points in the real world. Perhaps something special has been placed at the control points that the querier will notice. The querier then marks the control points on a virtual map (e.g., Google maps) and the image of that specific view of the map with marked points serves as a key into the image database.
No comments :
Post a Comment