In this paper, cosine-modulated filter banks are introduced into fract
al image coding to remove the block effects which exist in conventiona
l block-based fractal image compression both in spatial and frequency
domains. The proposed method offers not only an efficient solution to
such block effects but also better objective performances. Compared wi
th the fractal image compression methods based on discrete cosine tran
sform, about 1dB SNR gains can be achieved.