We apply noise suppression based on a feed-forward neural network (FFNN) onto hand vein images. A neural network is employed to classify corrupted and noncorrupted pixels. Filtering is only carried out on corrupted pixels, keeping the noncorrupted ones unchanged. The emphasis is on the selection of relevant inputs and training patterns. With appropriate choice of patterns, the assiduous task of FFNN training becomes effortless and noise detection in pixels becomes efficient. Comparative analysis on 20 sets of hand vein images using different filtering algorithms shows that filtering based on a FFNN scheme outperforms its counterparts.