Adaptive robust filters in MRI

Abstract

An adaptive noise filter that can be used in MRI for noise reduction is presented. The algorithm is mainly based on the robust estimator, mode. Using the mode as the gray scale value estimator it is possible to differentiate the structures of interest from the background noise. Noise reduction is one of the most common image correction procedures, used in the enhancement of digital images. Wildly used noise reduction filters for digital imaging are based on median estimation, median filters. Eery time noise reduction filters are applied to an image there is a general softening or blurring of it, in particular mean filters are characterized by a strong softening effect for the case of high amplitude noise levels, practically destroying all the fine features in the filtered image. This problem is significantly reduced when median filters are used. The adaptive mode filter proposed in this work have very good noise reduction effect without a strong softening effect and is comparable in CPU time to the median filters. This fine resolution is achieved because the filter changes the mode estimator according to the difference of the deviation of the mode calculated for each the neighborhood of pixels with the global deviation of the mode for each class in the image. We consider it robust, because it uses the mode of gray-scale intensity distribution of the pixels neighborhood and its mode deviation.