Model-based characterization of statistically optimal design for morphological shape recognition algorithms via the hit-or-miss transform
Academic Article
Overview
Research
Identity
Additional Document Info
Other
View All
Overview
abstract
Shape and character recognition via the morphological hit-ormiss transform is studied in a nondeterministic setting and the basic paradigm is thereby extended. A shape is viewed as a random process satisfying various model constraints and the recognition process is analyzed relative to the process. In particular, expectations of various types of recognition errors are analyzed. Rather than demand sufficient conditions for perfect recognition, the classical statistical approach is taken and the paradigm is one of measuring recognition efficiency and drawing from the probability model a criterion of optimality. In the framework of hit-ormiss shape recognition, optimization results from finding a class of hit-or-miss structuring element pairs that results in minimal error, as measured by expectation relative to the shape-process model. 1992.