Wavelet compression schemes have become powerful image compression [1, 13, 11, 6, 14, 5, 2, 10]. The EZW method [13] utilizen prediction of wavelet coefficient significance to effectively and efficiently compress images. A combination of EZW with vector quantization was reported [11]. An even more powerful method SPIHT has been recently preposed [10]. In this paper we propose another compression scheme, ModLVQ, which is based on SPIHT and includes the following modification: a modified lattice vector codebook, a more efficient coder, effective timing of the vector coding, and elimination of a minor inefficiency in the original algorithm. The experimental results are encouraging, especially on busy images ( which are more difficult to compress) and complex portions of other images, numerically and subjectively.