Upper left: A very "bad" 2-D velocity model which is designed to contain plenty features both in the depth direction and the horizontal direction. To the geophysists, the model is not realistic, but it is a really good mathematical model for MRA;

Upper middle: the same model with 34:1 lossy compression (I am not working on data compression, otherwise I can go further and get much higher compression ratio by applying quatitization and coding techniques);

Upper right: the residue of the compression.

The above plots are the MRA displays of the models. The MRA display is a new way to look at a velocity model. I strongly encourage to look at your model in this domain. It tells both spatial and spectral information of your model, and also tells your confidence level in your inversion.

Again, the upper left is the lossless MRA display of the model, upper middle is the lossy MRA display of the model with a 34:1 compression ratio; the upper right is the residue. It is not supprising why the model in MRA model space can be well compressed, when you find there are so many zeros in the MRA displays, in comparison with their spatial distributions.