Liveside have just posted about a cool new technology MS is building called DeepPhoto. It looks like a mashup of tools like Virtual Earth and PhotoSynth and allows photo’s to be mapped against the geometric data being built up by the virtual earth team.
With all of this information combined you can do some pretty interesting things. The video demo the LiveSide guys recorded shows how knowing the real world 3D references for objects in a photo allows the technology to de-haze images that are hazy for example. The real interesting stuff for me though comes at the end of the demo.
A photo of the Yosemite valley is mapped against the geometric model from Virtual Earth. This allows the software to increase the accuracy of the 3D view by extrapolating colours and textures from the real image. You can then move around this composite view in 3D – in effect filling in the gaps between the photos in a PhotoSynth scene.
Working for a company that does loads of GIS work, and produces visualisations of new projects, I can see this sort of technology as being genuinely useful to business. Assuming there’s some sort of API for adding your own data into the views (hyperlinks and additional models for example) it would be a very powerful way of visualising the sort of change we plan and design every day.
The LiveSide Video:
