Edge Segmentation based on Illumination Invariant Feature Detector Phase Congruency

Image Segmentation is necessary stage in image investigation. In computer vision, image segmentation is the process of dividing digital image into multiple segments. Dividing the image into various parts simplifies the analysis of image. In numerous image analysis and computer visualization uses, including remote sensing, robot vision, medical image analysis, and pattern recognition, the subdivision of image into segments is an important preprocessing stage. Finding the outlines of objects in the images is the fundamental step in the segmentation of pictures. Zero crossover, Canny, Gradient dependent, zero cross, Laplacian are examples of traditional edge detection techniques. This study presents a novel method based on phase congruency for edge segmentation. As compared to traditional edge detection methods, advantage of phase congruency is its invariance to uneven illumination. To obtain edge segmentation outcomes, edge lines were obtained using phase congruency. Long edges give an approximate sense of the object's arrangement in the picture. Color likeness between edges is determined while short edges are removed. Color similar edges assigned same label. Performance analysis shows that proposed method produces good results for synthetic images and natural real world images in which objects are well separated by boundaries whereas it results in broken edges for images having complex structure.