Abstract: An image compression method based on the features of image wavelet decomposition coefficient and zerotree structure is proposed in this paper. The matrix of wavelet image and the multiresolution matrix elements are depicted too. The encoding structure and the expressions which calculate the maximal absolute value of each subband wavelet coefficient and the mean square error ignoring this subband are given. With a Lena image of 512×512×8, the simulation is carried out. The results are compared with those of EZW. With the same peak signal to noise ratio(PSNR), this method gains obviously higher compression ratio and has better performance.