A fast algorithm for wavelet packet decomposition using statistical properties of images

Abstract

The recent attention of image coding community is paid to wavelet transform, which is of interest for the nonstationary signal analysis because it provides good localization in time and frequency. There are two directions of study in decomposing an image by wavelet transform, that is octave form and wavelet packet. Of this, wavelet packet works adequately for all images with various properties, but it has a difficulty of requiring relatively massive processing time. In this paper, using a statistical properties, such as correlation and variance, in spatial domain and octave band, a new algorithm for generating wavelet packet is presented. The described scheme has an ability of decomposing a image fastly. Simulation results have shown us a good performance in a view point of processing time and band decomposition ability comparing to the Vetterli's method.