Abstract: The SUSAN filtering operator can not self-adaptively modulate filter coefficients due to the use of an axisymmetric Gaussian filter. In order to solve this problem, this article presents an anisotropic SUSAN filter using an anisotropic Gaussian filter instead of a Gaussian filter in the SUSAN filtering operator. The variances of long axis and short axis are determined by the variance and the pixel neighborhood smoothness of the partial image. The long axis direction is determined by the gradient direction of the point. The SUSAN filter threshold is determined by the first order norm of the difference of the gray value and the mean value. For the infrared dim target detection, experimental results show that the anisotropic SUSAN filter could preserve the image edge information, the signal-to-noise ratio gain and signal-to-clutter ratio gain greatly improved in the residual image, the size of the target got a better retention, and the false alarm decreased.