Compact Pyramidal Architectures

In the last decade, several massively parallel architectures have been proposed and sometimes built to enhance the processing of large quantities of sensory information. Computer vision is the main field of application of such machines; low level tasks deal with regular data structures which match naturally with arrays of simple processing elements. Recently, multiresolution techniques have gained much attention as a bridge between preliminary and intermediate stages of information processing. Accordingly, pyramidal architectures have emerged as a natural extension of arrays. The hierarchical nature of multiresolution algorithms can be easily exploited on such machines, especially on the class of homogeneous, compact pyramids. In this contribution, the main features of these systems will be analyzed as case studies for the alternative solutions to the problem of designing and building a pyramidal system.