A non-local bilateral filter for image denoising

In this paper, under the non-local means framework, we propose a non-local bilateral filter algorithm for image denoising based on the neighborhoods' gray value and the corresponding neighborhoods' Gaussian curvature. We also adopt a new method to provide the optimum denoising parameter h based on the discrete wavelet transform and the smoothing spline estimation. Meanwhile the generalized cross-validation criterion is adopted to give the parameter in smoothing spline estimation. Experiment results demonstrate that the proposed method is robust and efficient for both additive and multiplicative noise.