The volume visualization techniques present two classical computer problems: long execution time and large memory requirements. The memory requirements becomes more investigated with the popularization of the distributed applications. This problem is important also, to make possible the acces to volume visualization techniques from a personal computer or low endworkstation with limited memory. In this thesis we propose a new lossy compression scheme for volumetric data based on the local cosine transform. This method is appropriated for further volume visualization application because it provides good compression rates, minimizes reconstruction errors, and allows local decompression of the volume. We analyse the compression results and estimate some scheme parameter variations, investigating the adaptivity properties and different space decomposition arrangements.