Efficient segmentation of cellular images using gradient-based methods and simple morphological filters

This paper presents an algorithm for segmenting two-dimensional (2-D) microscope images of overlapping regions of cell nuclei. Algorithms for segmenting such images must be fast as well as capable of handling subtle variability between objects if they are to be of practical clinical use. The authors attempt to address these issues by the application of techniques that exploit local features, extract object boundaries, perform skeletonization using the distance transform and gradient operators, and evaluate junction segment candidates based on boundary criteria.