On a New Nonlinear Image Filtering Technique

A generalization of the varying weight trimmed mean (VWTM) filter is proposed for the removal of both additive and impulse noise from corrupted images by introducing the center pixel to the original VWTM filter. Unlike many traditional nonlinear filters, one key feature of the VWTM filter is that the weights of the pixels to be averaged are not constant. They vary based on their differences to the median value. By adding the weight varying center pixel to the VWTM filtering formalism, the center pixel is therefore involved in adjusting its filtering output. Studies of the proposed filtering technique with several other methods show that the present method, in comparison, is extremely robust and efficient for removing noise from highly corrupted images.