Abstract: We present an improved wavelet image compression algorithm based on a wavelet-tree structure. It keeps and inherits all advantages of EZW and SPIHT, but improves their sensitiveness to channel noise. In this paper, we first focus on the rules of sorting, lattice vector quantizers and hybrid source-channel coding which are correlative with this new algorithm, then compare PSNR performance with SPIHT in the noisy channel environments.