Segmentation of Images Using Watershed and MSER: A State-of-the-Art Review

Abstract

The success of watershed transform in image processing and image analysis domain is due to its ability to induce closed object boundaries and needs less computation time with reference to other segmentation methods. This review delivers a thorough study of conventional image segmentation methods like watershed algorithm, marker-controlled watershed and the Maximally Stable Extremal Region (MSER) algorithm using Magnetic Resonance Images (MRI). This study determines the major accomplishments in the performance metrics of the relevant algorithms in the three areas such as tracking of fibres, licence plate, and faces. In addition, this review examines the key outcomes and focuses on the lessons learned and thus forms the foundation for future research. The results of the literature indicate that the MSER-based procedures outperforms well as compared with the other methods. The experimental results of the applications shows that MSER provides improved speed and stability.