Digital images are very often corrupted by noise, hence the development of effective algorithms for data denoising is a very important issue. Image denoising, however, is not a trivial task because the noise should be cancelled while preserving image details and textures. Since scalar metrics cannot evaluate these key features, vector metrics were introduced. This paper presents a new vector method that can yield very accurate measurements. The method is based on a novel estimation procedure that exploits both spatial and amplitude information about a filtered pixel in order to decide if that pixel is actually affected by residual noise or by unwanted distortion. Results of many computer simulations dealing with images corrupted by Gaussian noise show that the proposed method performs better than other vector and non vector metrics in the literature.
