Neural mappings and space-variant image processing

The authors consider cortical organization (i.e. layering, excitatory feedback, and mapping) an adaptation to special information processing needs that might be of value in technical applications as well. They present three examples of computationally advantageous mappings that are found in neurobiology. These consist of a space-variant filter formed by a coordinate transform and a uniform local operation two dimensional vector field and its representation by scalar excitation values in a compound (nested) map and integration of image segmentation data compound map. The authors conclude that the coding of stimulus information into cortical position is a general tool for neural information processing. The strategy of spatial coding is not limited to early vision problems, especially if compound mappings are considered.