Zerotree Wavelet Coding Using Fractal Prediction

Fractal image coding can be seen as spectral prediction of coefficients in the wavelet decomposition of an image. Fine scale coefficients are described by shifted and scaled coefficients from the next coarser frequency scale. We analyzed the ability of fractal coders to predict wavelet coefficients and found that the prediction of entire wavelet subtrees-as done by conventional spatial domain fractal coders-leads to poor coding results. This paper discusses how to design an improved fractal coder in the wavelet domain and proposes a new hybrid fractal-wavelet zerotree coding scheme. Although the coding results of the new coder outperform conventional fractal coders, they are unfortunately inferior compared with a pure wavelet version of this zerotree coder.