Pyramidal Seeded Region Growing Algorithm and Its Use in Image Segmentation

Abstract

Improvement of “Seeded Region Growing” (SRG) segmentation algorithm based on the pyramidal representation of image is described. Segmentation starts from the proper coarse level of pyramid using seed points chosen by the operator. Segmented contours are projected to the level below. On each subsequent level, SRG algorithm is applied only to pixels inside the window of variable size near the projected contour which leads to the linear dependence of execution time on the image size. Implementation exploiting the graphic user interface allows various forms of the interactive control of image segmentation.