#34: 3-D models from exhaustive, 2-D data
I’ve been thinking a lot lately about the relationship between photography and sculpture. In particular, I’ve been waiting for a software tool that would allow me to take, say, 20 digital photographs (what other kind is there?) and then stitch these together to create a seamless, quasi-continuous CAD-like model capable of showing the object in question from any angle.
I know that there are prototypes available in University labs all over the place -some of them may even work under the right conditions, but life’s too short.
My alternative, quick-and-dirty solution, given that I’m not that good at patience, is to take a picture of an object from almost every possible angle. No interpolation, no frame-to-frame registration difficulties, no advanced algorithms. If you need to know what that junction looks like between those four suspension members near the cylindrical brake cylinder? Simply dial up view number 2,002,003,040,105. Rather than rely on processing all those edges and splines, concentrate instead on memory (especially now that Moore’s law is running out of steam and storage costs are falling).
All these images would be stored by the location, distance and orientation of the camera as it is automatically moved around the object. It might even make sense to do what they do in astronomy and build a camera with multiple CCD sensors, arrayed perhaps in a hemispherical configuration.
What about zooming-in for extra detail? Well, I imagine that all the images would be recorded at some convenient minimum distance (determined by the acuity characteristics of the visual system, perhaps). The only zooming that could occur would then be zooming out. Alternatively, the whole process could be undertaken using handheld movie cameras, creating a convenient, continuous steam of images.
Ultimately, I’d like to see this kind of model displayed in a digital photoframe. Simply turning the frame relative to your viewpoint (monitored by wireless sensors as used in the latest Nintendo devices) would allow a different angle of the object in question to be displayed…in other words, a low-cost, digital hologram.
So is this really a model, in the sense of predicting things previously unknown? Possibly not, but I’m still thinking about that one.