Processing of Visual Imagery by an Adaptive Model of the Visual System. Its Performance and Its Significance.

摘要

From a review of biological evidence and assumptions in the human visual system, a model is derived for the recognition of visual patterns. The abstraction of information under a smearing transformation by adaptive spatial filtering suggests a useful redefinition of the classical pattern recognition approach involving the extraction of distinct features or properties on a feature space. The digital simulation of the model for pattern recognition is described,and the performance of the model on a variety of tasks is reported. The scope of this simulation is through to exceed that for any single pattern recognition model reported elsewhere. The significance of the performance of the model to several Air Force problems is discussed,and recommendations for continued investigations are suggested. (Author)