LAUSR

This paper aims to propose a novel contextual adaptive fourth-order smoothing method for noise removal. The classical Laplacian operator is first modified into a multi-scale and robust form, which combines two distinct discontinuity measures simultaneously, i.e., inhomogeneity and local spatial gradient. Then the contextual adaptive fourth-order smoothing method is proposed based on the new Laplacian operator. Moreover, a gain control function is employed for the gain control of smoothing in terms of contextual discontinuities to better preserve the important features. Experimental results support that the proposed method achieves the best performance among the comparative methods with respect to some objective evaluation metrics and visual effects.