An improved trilateral filter for Gaussian and impulse noise removal
Liu Ying-hui, Gao Kun, NI Guo-qiang · 2010
In this paper, we improve the trilateral filter for removing the mix of Gaussian and impulse noise. The new algorithm incorporates Rank-Ordered Absolute Differences (ROAD) Statistic for detecting outliers in gradient domain and intensity domain of image with impulse noise. A switching mechanism is adopted in gradient bilateral filter and intensity bilateral filter for smoothing the gradients and intensities of impulse noise samples and impulse noise-free samples with different parameters. By introduce the impulse detector in both gradient domain and intensity domain, the proposed algorithm has demonstrated superior performance in suppressing noise, which include Gaussian, impulse, and mixed noise. Compared to most other nonlinear filters, the proposed algorithm consistently yields good results in a sharply-bounded, gradient piecewise-linear approximation which provides stronger noise reduction and better edge-limited smoothing behavior.