A New Method for Edge Detection Based on the Criterion of Separability

This paper presents a new edge detection algorithm based on calculating the difference value of two clusters. An edge is defined as a boundary that separates two adjacent regions that are relatively homogenous. For each image pixel, a window is first defined by placing the pixel at the center, and this window is partitioned into two sub-regions respectively in four different directions. An appropriate function is then selected to estimate the difference between each pair of two adjacent regions and to calculate edge information in terms of edge strength and direction by maximizing the difference value. Finally, the non-maxima suppression is adopted to derive the output edge map. Experiments on a variety of noise contaminated images show that the new algorithm is more robust under noisy conditions. In addition, the proposed algorithm can also be applied to color or multi-spectral images.