Pavement image denoising based on shearlet treansform

A novel pavement image denoising method based on shearlet transform is proposed in this paper. Because that the pavement crack has continuous liner geometrical feature which can be captured by shearlets very efficiently with more directions than wavelets, the pavement image denoising method based on the shearlet transform can obtain a great improvement than traditional method. Background fitting is used to deal with the low frequency component of the image, which can balance the energy distribution of the pavement image. Then coarse scale coefficients of shearlet are selected under multiple thresholds. The coefficients obtained by low threshold is used for reconstruction of the main parts of cracks, and the coefficients obtained by high threshold is employed to extract crack position and direction information, which is fused with the threshold at fine scale to distinguish the noise and fine parts of cracks. The experimental results show that this method can smooth the most of noisy spots but keep the cracks details well and have less pseudo-Gibbs artifacts.