Constructing efficient linear local features in image processing and analysis problems

Abstract

We give formalize the problem of creating single and sets of linear local features for digital signals and images. A linear local feature is a pair consisting of a finite impulse characteristics (FIC) and an algorithm for computing the linear convolution of the signal with that FIC. Efficient linear local features should show optimal behavior: the algorithm should have minimal computational complexity, and the FIC should best accommodate the quality criterion for a specific applied problem. We offer formulations for the problems of creating single features and sets of efficient linear local features, where the process of computing features is connected with sequences of a special kind. We give examples of such sequences.