Entropy-based trilateral filtering for noise removal in digital images

Bilateral filtering has been a popular denoising technique that smooths images while preserving edges by means of a nonlinear combination of adjacent pixel values. We propose an entropy-based trilateral (EnTri) filter that extends the classical bilateral filter for noise removal in digital images. A new median-metric weighting function is incorporated into the geometric and radiometric components, followed by an entropy function to balance the contribution between the weights. The entropy function is used to adaptively detect the local intensity variations on each pixel. Our EnTri framework replaces the intensity value on each pixel with an average value weighted by the three components between neighboring pixels. A variety of images contaminated with various levels of noise were used to assess the performance of this new filtering method. Experimental results indicate that the EnTri filter outperformed several existing methods in both visual image quality and restored signal quantity.