Morphological paradigm for loss-function-based design of digital filters
Conference Paper
Overview
Identity
Additional Document Info
Other
View All
Overview
abstract
Statistically based automatic design of nonlinear image processing algorithms has been used successfully for binary image enhancement, specifically in restoration and resolution conversion of documents. The present paper introduces an extension of the methodology in two directions. First, it proposes to use the representation- optimization paradigm for general algorithm development in the context of system transformations. Second, it employs a statistical loss function to generalize the mean-absolute- error approach taken in previous work.