Image Segmentation Based on Global Extraction and Local Repair of Boundaries

As the edge lines obtained from the Canny's edge detection operator are not closed, it difficult to use the edge map for the image segmentation. Moreover, the detection result contains some edge lines have nothing to do with the target partition, also resulting in poor segmentation results. This paper presents a method combining global edge extraction with local edge repair, in conjunction with the morphological processing to obtain closed edge lines and remove irrelevant edge. GELR is short for this method.