We previously proposed a method of seismic endoscopy and a related prototype tool to acquire directional information and to produce three-dimensional (3-D) seismic images in a cylindrical volume surrounding a borehole, with an investigation radius of several metres. Basic imaging algorithms were developed where the azimuthal move out (AMO) process is combined to a timescale method to refocus directional information and analyse the surrounding medium. Further processing tools, however, were necessary for separating the various types of waves recorded. The present paper describes a multiscale and dynamic azimuthal filtering to separate the far-field waves from borehole or tube waves, generated by the pipe. We call this filter Stoneley move out (SMO), because it characterizes Stoneley waves associated with the tube. It is then possible to reconstruct details of the far field by an inverse wavelet transform. The comparison with classical methods such as the covariance method is discussed. Applications on noisy synthetic and experimental data are presented.