This work focus on the problem of image compression, with exploring the techniques of subband coding. The basic structure, used in the sirst part of this tesis, encompass the uniform decomposition of the image into 16 subbands. This procedure aims at reproducing the reults of Woods [1]. The component of the 16 subbands are quatized and coded and bits are optimally allocated among the subbands to minimize the mean-squared error. The quantizers desingned match the Generelized Gaussian Distribuition, which model the subband components. In the coding process, the lowest subband is DPCM coded while the higher subbands are coded with PCM. As an innovation, it is proposed the use of the algorithm LZW for coding without error (compaction) the quantized subbands. In the compactation process, the Huffamn and LZW algorithms are used. The simulation results are presented in terms of rate (bits/pel) versus peak signal-to-noise and subjective quality. The performance of the subband decomposition tecnique obtained with the Huffamn´s algorithm is about 2dB better than that obtained with the LZW. The universality of the Lempel-Ziv algorithm is, however, an advantage that leads us to think that further investigation should still be pursued.