An efficient segmentation algorithm based on mathematical morphology and improved watershed

Image separation is a critical issue toward the recognition and analysis phase in many image processing tasks. This paper describes an efficient segmentation algorithm based on mathematical morphology and improved watershed which uses the immersion-based watershed transform applied to the fusion image of multi-scale morphological gradient and distance image to decrease the known drawback of watershed, oversegmentation, notably. Furthermore, oversegmentation is posteriorly reduced by a region merging strategy to obtain meaningful results. The presented segmentation technique is tested on a series of images and numerical validation of the results is provided, demonstrating the strength of the algorithm for image segmentation.