Inversion of seismic amplitude into seismic impedance can be a difficult and time-consuming process. In this paper, a simple yet effective alternative approach is presented. The method relies on clustering similar seismic amplitude features, thereby exposing distinct geological features (geobodies) directly from the amplitude data. The method consists of first scanning the amplitude with a pre-defined window to obtain realizations of windowed seismic amplitude features. Next a minimum message length (MML) clustering algorithm is applied to search for similarities within the set of windowed amplitude features. Each group of similar amplitude features is given a unique cluster label. It is shown that the 3D cube of cluster labels often reveals existence of distinct geobodies. This approach is illustrated on a simple 2D synthetic case and on a real 3D seismic data set of a submarine channel deposit.