Structure adaptive anisotropic image filtering

Noise filtering of images is essentially a smoothing process, and it is an issue that has been addressed for many years. The most commonly used low-pass filtering methods blur important image structures such as edges and lines, and thus reduce image contrast and damage image fidelity. This paper presents a structure adaptive anisotropic filtering technique with its application to processing magnetic resonance images. It differs from other techniques in that, instead of using local gradients as a means of controlling the anisotropism of filters, it uses both a local intensity orientation and an anisotropic measure of level contours to control the shape and extent of the filter kernel. This ensures that corners and junctions are well preserved throughout the filtering process. The following two aspects of the proposed technique demonstrate the advantage of using this filtering method. Firstly, the use of local orientation detection provides a robust and convenient way for shaping the filter kernel. Secondly, the structural adaptiveness of the filtering process ensures that salient image features are non-symmetrically enhanced.