VARIABLES-RESOLUTION DIGITAL VISION.

A few techniques for analyzing digital images on nonuniform or variable resolution artificial retinas are described. Analyzing images by variable-resolution algorithms can be computationally more efficient than by uniform-resolution algorithms. In fact, some of the variable-resolution techniques, when specialized to uniform mosaics, seem to be more efficient than some methods developed specifically for uniform mosaics. In particular, the following techniques are described. 1. The representation of images by minimum-perimeter polygons on variable-resolution mosaics. 2. The use of the theory of linear separability for detecting cellular straightness in variable-resolution mosaics. 3. A procedure for finding optimum triangulations on nonuniform mosaics of points, leading to a new procedure for constructing contour maps from nonuniformly distributed data.