The non-linear filter such as a median filter plays an important role among various methods of enhancing an image and removing random noise. This type of filter is one of an extended family of rank-order filters. On the other hand, a directional filte...
The non-linear filter such as a median filter plays an important role among various methods of enhancing an image and removing random noise. This type of filter is one of an extended family of rank-order filters. On the other hand, a directional filter is a linear filter that minimizes the stair-case artifacts commonly produced from bi-linear interpolated images and simply zoomed images.
In this paper, the directional rank-order filter is proposed, in which the two kinds of linear and non-linear filters are integrated effectively to remove stair-case artifacts for zoomed images, using three types of ranks; maximum, median, minimum. The optimum direction for a given pattern is first estimated from surrounding pixels in a window, then the optimum rank is also estimated for a given pattern of an image. The simulation result shows that the proposed new type of filter can enhance and remove stair-case effects effectively for bi-linear interpolated images or noisy images.