Аннотация:We propose a new algorithm for noise suppression – HeNLM. It is a modification of an existing local jets based method LJNLM-LR. In our method, the neighborhood of a pixel is expanded in a series of Hermite functions, which, unlike Gaussian derivatives, form an orthonormal basis. Testing of our method against Local Jets and NLM has shown better preservation of detail-rich areas, sharp edges in CT scans, without loss of medically-important information.