A self-governing hybrid model for noise removal

Abstract

Denoising methods based on using fourth-order partial deferential equations (PDEs) are providing a good combination of the noise smoothing and the edge preservation without creating blocky effects on the smooth regions of the image. However, finding an optimal choice of model parameters for numerical solver of these techniques is a challenging problem and generally, these model parameters are image-content dependent. In this paper, a hybrid fourth-order PDE-based filter is proposed so that it does not need a manual adjustment of the model parameters. It is shown that by setting the numerical solver of proposed filter for operation at a minor time step-size derived under a data-independent stability condition, the filter can still provide a significantly fast convergence rate. Therefore, the model parameters are reduced to one parameter estimated by using a well-studied mechanism applying in the second-order nonlinear diffusion denoising techniques. Simulation results show that the proposed method can provide a denoised image with higher quality in comparison with that of the existing methods. © 2009 Springer Berlin Heidelberg.