Watershed Segmentation Based on Multi-Scale Morphological Filtering

Image segmentation based on watershed method always results in over-segmentation, makes the contour of object blurred with irrelevant watershed lines. To alleviate this problem, a robust multi-scale morphological filter is introduced to eliminate the noise and simplify the input image and morphological gradient image while preserving the main contour of object. A modified fast region merging algorithm is also proposed to improve segmentation result. Experiments show that the proposed algorithm is efficient.