Reconstructed images in digital tomosynthesis (DTS) are affected by artifacts due to blur from planes other than the fulcrum plane. A wavelet-based method has been developed for the discrimination and subsequent removal of unrelated structures from the reconstructed plane. The approach exploits both the specific pattern of noise in DTS and the spatial locality of the wavelet transformation. The technique was implemented on a DTS clinical protoype system. Experimental evaluation on angiographic types of images demonstrated excellent noise differentiation and elimination. The method is therefore particularly useful for certain medical imaging applications such as vascular DTS imaging.