The important information in a high-resolution two-dimensional NMR spectrum or 2D NMR image is localized in a small fraction of the overall data block. Hence data files should be subject to compression by suitable manipulation. We propose two methods for data compression: elimination of background noise and logarithmic scaling of the data. By combining the two methods, one can obtain a compression factor of 30 or more without significant loss of spectral information content. The method is particularly suitable for the storage of massive data sets. One or more two-dimensional spectra can be reduced and stored on one floppy disk. These methods and their versatility are demonstrated via compression of two-dimensional NMR spectra of a small protein (turkey ovomucoid third domain) and via compression of the NMR image of a phantom.