An image edge detection and segmentation algorithm based on small-world phenomenon
Naijian Chen, Sun’an Wang · 2008
This paper presents a novel approach, depending on the threshold and clustering probability, to generalize small-world phenomenon to image edge detection and segmentation. It begins with computing the optimal threshold, based on global image features. Then, the algorithm iterates two steps. 1) The small-world effect. Searching the pixel with sudden changes of an image attribute such as luminance from its neighboring pixels, the algorithm forms a candidate set of edges’ pixels based on the optimal threshold. 2) Graph edge clustering. It clusters candidate pixels with assigned probability into edges to segment image in HS color-space. Through pre-setting the probability of clustering, the algorithm could change the threshold in certain range and apply it from overall features to partial attributes, and segment the image from rough to detail. The example images are included to illustrate the stability and effectiveness of the proposed approach.