Anisotropic Gaussian filtering for infrared image

To achieve much better noise suppression with minimum edge blurring, an adaptive anisotropic Gaussian filtering algorithm was proposed. In this algorithm, the scale of filter's long axes was determined by Intensity-Dependent-Spread model in each pixel, and the ratio between filter's long axes and short axes was determined by the degree of local smoothness, then the orientation of the filter's long axes was also calculated according to the gray gradient orientation in the position. The experimental result shows that the algorithm has good capability in noise suppression and edge-preserving.