关键词: 图像分割, 分水岭, 形态学混合开闭, 区域合并

Abstract: Aiming at the problem of over-segmentation of traditional watershed algorithm in image segmentation process, a watershed improvement algorithm based on morphological markers and region merging is proposed. Firstly, the color image is converted into gray image, the image reconstruction is realized by morphological hybrid opening and closing operation, the image boundary is strengthened by Laplace sharpening, the target object and background are marked, the amplitude image is corrected, and the image is coarsely segmented by watershed transformation. Then the MRSM region merging method is used to achieve effective segmentation of the target and background. Compared with clustering segmentation and H-minima-based watershed algorithm and Otsu class maximum spacing algorithm, the improved algorithm has obvious advantages in segmentation effect, algorithm running time, segmentation accuracy and recall rate, and average with other algorithms mentioned above. The split time is increased by 3.86 ms, 2.88 ms and 5.64 ms. The split intersection ratio is 96.42% higher than the IOU average, which is 12.65%, 2.77% and 3.07% higher than the average of other algorithms.

Key words: image segmentation, watershed, morphological hybrid opening and closing, regional merger