Efficient Noise Removing based Optimized Smart Dynamic Gaussian Filter

Abstract

filter has been used extensively in signal image processing for many years. Gaussian or Gaussian derivative filtering is in several ways optimal for applications requiring low-pass filters or running averages. In this paper, a highly efficient noise removing technique based on a modified dynamic Gaussian filter is introduced. Called smooth filter, the Gaussian filter is known to be more efficient for conserving details and slight borders then other filters. In the proposed approach, we developed a variable shape low pass filter that can be used for efficient noise removal even with impulsive noise. In this study, the filter selects automatically the processed windows based on an automatic noise targeting in such a way that the image does not lose its characteristics. An optimal magnitude and support extent of the Gaussian filters is continually computed in an iterative method for each selected windows of the image. This approach is approved experimentally using salt and pepper noise. In fact Gaussian filter is not appropriate for removal of impulsive (salt and pepper) noise that needs filters based on statistical approach. Nevertheless high efficiency in removing high densities of noise difficult to remove even using median filter is shown. In addition the image quality is preserved. This proposed method combines the behavior of an intelligent dynamic low-pass filter that eliminates only high frequencies corresponding to noise and a filter based statistical approach such as median filter that removes efficiently impulsive noise and conserves details.