HeNLM-LA3D: A Three-Dimensional Locally Adaptive Hermite Functions Expansion Based Non-Local Means Algorithm for CT Applications

摘要

A three-dimensional filtering algorithm for CT images (HeNLM-LA3D) has been proposed that is based on expanding the pixel neighborhood into Hermite functions, which form an orthonormal basis. Accounting for Hermite functions properties, pixel neighborhoods are oriented according to principal com- ponents of the structure tensor. The filtering parameter is adaptively adjusted to local estimates of the noise level. A noise estimation algorithm is proposed.