Fractal image compression by the classification in the wavelet transform domain

In the fractal image compression the domain-range comparison step of the encoding is very computationally intensive. Therefore in order to minimize the number of domains compared with a range a classification scheme is used. We propose a new theory for classification of domain-range blocks. The classification uses non-decimated separable discrete wavelet transform. The encoding using the proposed classification is compared with that by Y. Fisher (1995), which uses the average and the variance as features of images and classifies domain-range blocks into 72 classes. The Y. Fisher's classification uses the variance which represents only a messy degree of image intensities. The new classification proposed in this paper represents more effective features of images and classifies domain-range blocks into 432 classes by using the average and the power. With this classification we are able to encode faster and realize a high image quality in fractal image compression.