Interactive viewpoint modification for multiple perspective views

From two real time camera video outputs a limited range of new perspective views can be created consistent with the individual camera positions and resolutions. This approach is readily extended to multiple cameras which improves range and or resolution. This has application in video teleconferencing where user control of the viewpoint would provide new capability for remote presentation of true three dimensional objects. This would have value for either a single perspective view or for simultaneous presentation of a stereo pair of separate left and right eye views. The user could remotely change the perspective view in a manner equivalent to typical physical observer motion in the examination of a real object. This could also have application in passive navigation for autonomous vehicles where information on trajectories predicted from video input would supplement input from other instrumentation. For both applications, new views can be created either by explicitly creating a three dimensional scene model, from the multiple views of by using interpolation and masking operations to create new views from an implicit model. Algorithms for performing these operations are presented and performance limitations are determined.