A conjointly well localized quadrature mirror filterbank

This paper proposes the use of conjointly time-frequency well localized filters that constitute a perfect reconstruction quadrature mirror filterbank. The time-frequency measure to determine optimality is the product of a filter´s time and frequency variances, which has already been proposed. An in depth analysis of parameter selection and novel use of polynomial fitting to ensure perfection reconstruction of the optimal quadrature mirror filterbank are provided in this paper. The output of the separably extended 2-D analysis filters of the optimized 1-D perfect reconstruction quadrature mirror filter-bank illustrates the best space-frequency localized decomposition of the original image.