Efficient Mutually-Calculated Features for Linear Local Description of Signals and Images

V. Myasnikov (Russia)


Signal processing, linear local features.


Problem of linear local features construction for digital signals and images description which are qualitatively and computationally efficient is considered in the work. As linear local feature a pare is considered which consist of finite impulse response (FIR) and an algorithm for linear convolution computation of a signal with this FIR. Mutually-calculated linear local features have united algorithm which performs simultaneous (mutual) computation of multiple linear convolutions of input signal with several FIR set at once. Efficient linear local features has to demonstrate optimal behavior, that is algorithm has to have minimal computational complexity and FIR set has to be the best matched to quality criterion of given applied task. The approach of mutually-calculated efficient linear local features construction, based on the sequences set of special type is proposed. Examples of such sequence sets are given.

Important Links:

Go Back