Multibandwidth wavelet-transform image coding using zerotrees
E.S. Demirciler, George H. Freeman · 2002
A wavelet transform with multibandwidth channels is investigated for image coding. It offers improved frequency resolution and good compression efficiency. For effective coding, the multibandwidth transform is accompanied by (1) the use of zerotrees (proposed by Shapiro (see IEEE Trans. Signal Proc., vol.41, p.3445-3462, 1993)) to predict insignificant information across scales; and (2) entropy coding (done by an adaptive arithmetic code). Optimised bit allocations for the orthonormal and biorthogonal multibandwidth decompositions are derived. The use of different wavelets to perform different stages of the decomposition is shown to be advantageous. Empirical compression results are presented for individual test images.