Mobile camera parameter recovery in an unknown environment without point features

We show that unknown camera and vehicle parameters may be recovered from images of an unknown environment with no point features. Our method complements established geometric methods which require point features. Given an image sequence, we construct a series of light field models over a range of parameter values. The 'sharpest' model is seen to correspond to the correct value of the parameter, and our method may be viewed as an abstract form of focusing. We describe experiments with a one-dimensional camera mounted on a robot moving about in a two-dimensional environment that does not contain point features. The results demonstrate that accurate parameter values can be recovered from real images using methods that do not require point features.