NM images filtering using NPR QMF filters dependent on the images spectrum

In this paper we present our approach on pre-processing of the heart region dynamic NM images. The presented method combines a discrete wavelet transform (DWT) realized via linear phase near perfect reconstruction (NPR) quadrature mirror filters (QMF) with a specific strategy for selecting the filters cut-off frequency and the filtering into the time direction of the consecutive images. The aim of the method is to develop sophisticated diagnostic software that would automatically offer the optimal positions and the shapes of the regions of interest needed for heart studies.