An image denoising algorithm is proposed with an overcomplete set of linear transforms for removing white Gaussian noises of the image.The denoising result is gotten by weight averaging.The method computes the weight with the sparse concentrate index.The weight depends on sparse decomposition of localized transform coefficients.Sparser the image region is

the larger its weight is.The method has no need to devise elaborate statistic models on the transform coefficients and more sophisticated transforms for image singularities.The method is simple and obtains very high denoising performance

especially for the image with singularities.Experiments results show the effectivity of the method.