Salnikov D. Hardware and software means of adaptive nonlinear digital filtering based on noise detectors

The dissertation is devoted to the development of software and hardware means of adaptive nonlinear digital image filters based on noise detectors. The scientific results are: 1) the algorithm of the adaptive weighted median digital image filter is improved, the software model with calculations with a fixed point in Python language is implemented; 2) the algorithm of the adaptive weighted median digital image filter was further developed by implementing the units for estimating the weights and the mechanism of bypassing the calculations, which allows reducing the hardware costs for its implementation on the basis of FPGA; 3) for the first time a digital filtering algorithm with post-filtering decision-making is proposed, which does not have the disadvantages inherent in the most productive adaptive filters, in particular, the adaptive weighted median switching filter; 4) improved the method of implementing a filter with post-filtering decision-making based on FPGA or processor architectures with a vector set of instructions, in particular ARM NEON and Intel SSE/AVX. The proposed methods and tools allow reducing the hardware cost of removing pulsed noise from images by adaptive digital filtering in systems using as hardware platforms programmable logic integrated circuits and/or processor architectures with vector instruction extension and increase the ratio "performance/hardware costs».