Research of Low SNR Crack Denoising Based on Kernel Anisotropic Diffusion

We propose a new kernel anisotropic diffusion filter for denoising low signal-to-noise ratio (SNR) crack images of straddle-type monorail beam surface. This method incorporates an enhancement operator into the diffusion steps, and then the low dimensional data is promoted to high dimensional space according to the characteristic of noise uniformly distributing in the multidimensional space, also average absolute difference value of automatic diffusion termination criterion is introduced to enhance the efficiency of diffusion. In addition, a different edge enhancement weight is explored to achieve the appropriate result. This method achieved much better noise suppression with minimum edge blurring compared with median filter and the original anisotropic diffusion filter when it was tested on the low SNR crack image of straddle-type monorail beam surface of 0.4mm width.