Analysis of methods for construction of efficient linear local features for digital signals and images description

Analysis of several methods for the construction of efficient linear local features for digital signals and images description is given in the work. This analysis is carried out for the best method selection. As a linear local feature (LLF) of a digital signal a pare is considered which consists of a finite impulse response (FIR) and an algorithm for linear convolution computation of the signal with this FIR. Efficient LLF demonstrates optimal behavior, such as the algorithm has minimal computational complexity in some class and FIR of LLF is the best matched to a quality criterion of an applied task. For particular applied task the method for efficient LLF construction is specified by the formulation of LLF construction task and by the algorithm for its solution. Two formulations of efficient LLF construction task are considered in the current work: the particular and extended particular task formulations. Algorithms for exact solution of particular task and approximate solution of extended particular task were suggested in authors' previous works. The comparison of constructed LLF is carried out with the use of different criterions that characterize features properties.