An optimal performance investigation for bilateral filter under four different image types

Abstract

The merit of bilateral filter is that it preserves the edges of the image and smooth the image at the same time. This can be done by controlling two important parameters — spatial domain parameter and range domain parameter. The contribution of this paper is two-fold. First, we investigate the optimal parameters of the bilateral filter under four different types of images — aerial, face, object and texture. The simulation results show that the parameters — spatial domain parameter and range parameter — should be changed depending on the noise power and the details of image. If the noise power is high, the parameters should be set at high levels. These values should be decreased on the increasing in an SNR. Second, we determine the appropriate types of image that the bilateral filter can be applied to perform filtering. Then, we compare bilateral filter’s performance to Gaussian filter and switching bilateral filter. The simulation results show that the bilateral filter gives the best performance when perform de-nosing of aerial, face and object image where the detail of these images are less than texture image.