Abstract:
Spiral Architecture is a new and powerful approach to a general-purpose machine
vision system. Earlier research on the Spiral Architecture focused on the properties of
undistorted translation, rotation and scaling. The research reported here uses the properties
of the pixel address labelling scheme. The property of interest is the physical proximity of the
hexagonal pixels with neighbouring addresses. Rectangular systems may for instance, have
vertical physically adjacent pixels but the address distance is the length of a scan line. It is
demonstrated that in the Spiral Architecture, unlike the rectangular system. many images
have the property that neighbouring pixels have similar intensities thus giving opportunities
for better compression.
