Requirements

2. Requirements for the Analysis of Visual Stimulus

According to the author, the visual system and any other artificial imitating the human visual system we may devise, must meet the following three criteria (greater details regarding these criteria is provided in the following sections):
  1. Abstractness: ability to establish abstract properties and relations.
  2. Open-endedness: ability to establish a large variety of properties and relations including new ones.
  3. Complexity: Ability to efficiently handle the complexity involved in determining spatial relations  e.g. scale independence.

2.1 Abstractness

Abstractness can be defined easily using an example (as done in the paper).  Consider a set of shapes S, satisfying some property P (for example, the set of all shapes which are closed).  Set S may be very large (as in this example) and it may be impossible to determine whether some shape B belongs to S by template matching (e.g. comparing each shape in S to shape B).  However, determining whether B belongs to S can be easily tested by taking advantage of any regularities the set S may contain.  Regularities allow for membership testing in the set S to be "broken down" into a set of operations which allow one to test whether B is included in the set S quicker and more efficiently in comparison to template matching.  For example, given some polygon B, the human system doesn't template match with every possible closed polygon.  Rather, other means are used.  For example, if the shape is closed, then placing some mark on the border of the shape, and tracing the boundary of the shape, will eventually bring us back to the starting point only if the shape is closed (this operation consists of marking and tracing, which are described in greater detail in future sections).

Abstract Shape Properties refers to the establishment of properties and relations of some large set by using any regularities it may contain.  According to the author, the human visual system has this ability and it is performed by using a combination of processes which in turn are composed of several basic "fundamental" or  "elemental" operations.  The processes which are comprised of these elemental operations are referred to as visual Rroutines.

2.2 Open Endedness

The processes used to determine some property share the same elemental operations e.g. the elemental operations are not created by some process when it is need but rather, are available for any process to use.  The open-endedness requirement requires that these operations be combined to perform new computations as required by the visual system.

2.3 Complexity

The elemental operations used by some process will be the same operation and as mentioned above, are not created by some process when needed.  Furthermore, these elemental operations which may be operating on different locations share some of their "resources".  For example, there is no independent module for each possible location in the visual field.