Robust Trilateral Filtering with Reachability-Based Impulse Detection for Mixed Noise Removal

In this paper, a new color image denoising filter is proposed with the goal of efficiently eliminating mixed noise consisting of Gaussian and impulsive noise. The new approach builds upon the trilateral filtering concept by introducing a sophisticated impulse detection strategy founded on a generalized reachability notion and an enhanced Rank-Ordered Absolute Differences similarity metric. In order to enhance the accuracy of impulse detection even more, in a preprocessing stage, a bilateral filter is employed to suppress Gaussian noise without hindering structural details. Experimental evaluation on benchmark test images reveals that the presented technique performs better than state-of-The-Art techniques as far as denoising quality is concerned.