Camera interpolation

If controlling a single camera is difficult, constructing a camera path is even more so. The traditional approach is to specify a sequence of key frames, then interpolate from one key frame to the next. This works well for some camera motions, but can produce unexpected results for certain motions. A simultaneous rotation and translation can result in the object of interest flying off of the screen. A center of projection change results in substantial jitter as well.

We present two solutions to the camera interpolation problem. The first uses linear matrix interpolation to interpolate the camera matrices. The interpolation can be a linear interpolation, a spline-weighted blend, or an interpolating blend. The second approach performs the interpolation in image space.

Image-space interpolation also provides additional control over the trajectories and sizes of selected objects in screen space. This lets the user use far fewer key frames. To support image-space interpolation we developed the Four-point camera representation, which is more suited to solving image-space constraint problems.

See also camera control and non-linear projection.

Image-space interpolation movies

The first three movies demonstrate the different behaviour of three different image-space solvers and traditional key framing. The next two movies demonstrate directly editing the camera using image-space constraints.

Linear matrix interpolation movies

Papers

Students