Image Compression by Gabor Expansion

Abstract

Transform-based coding methods are popular in data compression. In the paper, an easily implemented method is proposed for the weighting factors of the Gabor decomposition. The method is based on the least-mean- squares error (LMSE) approach. The solution of the LMSE problem shows that the weighting factors can be extracted by simple multiplication between a matrix and the vector of data. If the set Gabor functions are chosen to be independent of the test images, this matrix is constant. Images are reconstructed by multiplying the matrix of Gabor functions and the vector of weighting factors. The choise of Gabor functions in the decomposition allows that the resulting decomposition has a pyramidal structure. In the paper is proposed simple codec system for pyramidal Gabor expansion for image compression.