Efficient structures for two-dimensional quadratic filters

Abstract

Matrix decomposition techniques are used in order to obtain efficient structures for 2-D quadratic digital filters. The proposed structures consist of parallel branches which operate simultaneously on the input array. Each branch is formed by a 2-D FIR filter in cascade with a squaring operator, and all the partial results are summed up to obtain the output value. In the paper it is shown that, in particular, the lower-upper (LU) decomposition permits computational and hardware advantages.