Plato's Cave: Information Theory

Information Theory

According to Information Theory, [information compression] is an essential component to visual abstraction. For example image surfaces can be compressed to image edges, the edge information can be compressed to corners between straight line segments, and patterns of corners can be encoded by the symmetry of their relative positions (e.g. a triangle has 3-way symmetry, a square has 4-way symmetry, etc.). Information theory therefore suggests that higher levels in the visual hierarchy encode patterns of regularity in lower levels, which in turn encode regularity at still lower levels and so on.

The compressed representation by itself however is meaningless without a decompression algorithm to unpack that information on demand. This suggests that visual perception involves the two separate and complimentary functions of abstraction and completion. Abstraction involves the elimination of redundant information, while completion involves the restoration or decompression of that redundant information on the basis of either the compressed representation, or based on the visual input. Visual illusions such as the Kanizsa figure clearly illustrate that the visual system is capable of completing the "redundant" edge and surface information suggested by the information at the corners of the figure.

Return to argument

Return to Steve Lehar