Complexidade computacional de um algoritmo competitivo aplicado ao projeto de quantizadores vetoriais

Abstract

In the present paper, the computational complexity of a competitive learning algorithm applied to vector quantization (VQ) codebook design is investigated. Analytical expressions (as a function of the codebook size, the dimension of the codevectors, the number of training vectors and the number of iterations performed) are derived for the number of operations (multiplications, divisions, additions, subtractions and comparisons) performed by the competitive algorithm. Analytical expressions are also derived for the tradional LBG (Linde-Buzo-Gray) algorithm. Regarding VQ codebook design for image coding, results show that the computational complexity of the competitive algorithm is lower than that of LBG.