One-Dimensional Processing for Edge Detection using Hilbert Transform

This paper presents a new method for edge detection using one-dimensional processing. The Discrete Hilbert Transform of a Gaussian function is used as an edge detection lter. The image is smoothed using 1-D Gaussian along the horizontal (or vertical) scan lines to reduce noise. Detection lter is then used in the orthogonal direction, i.e., along vertical (or horizontal) scan lines to detect the edges. The proposed method di ers from the traditional approaches based on 2-D operators in the sense that smoothing is done along one direction and the detection lter is applied along the orthogonal direction. The traditional 2-D operators smooth the image in all directions, thus resulting in some loss of edge information. Performance of the proposed method is compared with Canny's method for a set of real-world images. We also compare the performance of the proposed lter with the rst order derivative of Gaussian (1-D Canny operator) for di erent 1-D edge pro les.