COMPUTATIONAL PERFORMANCE OF HOLE FILLING MORPHOLOGICAL ALGORITHMS FOR BINARY IMAGES

The presence of holes, cracks or scratches in one or more object regions in a binary image usually results from quantizing or thresholding a gray scale image. However, for further processing or quantitative binary image analysis, those artifacts must be removed by filling the corresponding object regions. In this paper, a computational performance analysis is realized for the class of hole filling algorithms based on mathematical morphology. Two fundamental techniques, supervised and unsupervised, are described in detail based on marker images that may be composed of pixel subsets chosen within an object region artifact, formed by external near by points to object regions, or from selected background pixel subsets. A mathematical description spanning the different variants is given on how this kind of algorithms converge to the desired result. In addition, illustrative examples using representative binary images are provided to test and compare the computational performance in terms of the number of iterations corresponding to each morphological hole filling algorithm for binary images.