Doroshenko Y. Hardware and software means for tree-structure based nonlinear digital filtering

Research object: processes digital signals and images filtering, methods and means of ensuring the efficiency of processes and filtering options of digital signals and images for computer systems, their devices and components. Research purpose: the increase the ratio of "efficiency / hardware - software" costs, hardware for signal processing and image through the development and analysis of a new class of nonlinear digital filters based on tree structures and order statistics. Research methods and apparatus: theory of digital signal processing, statistical methods and spatial processing of random processes and fields, combinatorial analysis and the theory of partitions, methods of discrete optimization and graph theory, theory of digital automata and Boolean algebra. Theoretical and practical results: the models, methods and algorithms, hardware and software implementation of tree- structured median filter, which can be used in the construction of high-speed devices and systems. Novelty: developed a new promising class of nonlinear digital filters - tree- structured median filters, which have the best ratio performance / hardware costs compared to existing and offered their semigraphical model - M-ary tree with an odd number of leaves; algorithms for the synthesis of possible structures of filters to select the optimal structure among possible criteria for the degree of noise suppression and on the tree, which is built on a mask weights the prototype filter; further developed nonlinear digital filters with a stack architecture, which differ from existing through the implementation of positive Boolean functions based on graph model M-ary tree. Degree of introduction: Kharkiv Morozov Machine Building Design Bureau; the firm "Diagnostic systems"; National Technical University "Kharkiv Polytechnical Institute". Sphere of the use: images and signals processing.