Spectral Density Driven Wavelet Representation of 2-D Images

The representation of images is a fundamental issue. Wavelet transform is the foundation on which the state of the art still-image compression algorithms are based. These algorithms exploit good energy compaction property of wavelet representation. In this work we propose a new paradigm in the wavelet based image representation. The proposed work uses the spectral property of a given image for the representation and can be applied to images having arbitrary spectra. The method is valid for both orthogonal as well as biorthogonal wavelets. The existing wavelet based schemes give good compression performance for images having spectra, which are dominantly low pass and where the signal energy decreases with increasing frequency. The simulation results clearly demonstrate the superior energy compaction property of the proposed scheme when compared with the standard wavelet representation.