Skip to Main Content
In this paper, we introduce two novel edge detection algorithms based on a negative alpha weighted quadratic filter. The goal of this work is to utilize the characteristics of the nonlinear filter to preserve and enhance edges for the purpose of edge detection. Unlike traditional edge detection algorithms, which detect edges by using derivatives, the proposed algorithms operate on local regions and modify the color tones of uniform regions while preserving the original edges. We also incorporate the luminance masking feature of the Human Visual System by masking the gradient image before edge labeling. Experimental simulations show that the proposed algorithms can extract fine edge information from images contaminated by noise and affected by non-uniform illumination; the obtained edge maps are more consistent to the edges perceived by the human eye. Comparison with existing algorithms will be also presented.