An improved watershed image segmentation algorithm based on particle swarm and region growing was proposed to solve the problems of noisesensitivity and over-segmentation. The improved algorithm, combining region growing with the classical watershed algorithm, was established by constructing an objective function based on Shannon entropy to determine the parameter of the region growing. The regional disparity degree was calculated by the gray mean, and the smaller region was merged with the neighbor region with a minimal disparity degree. The particle swarm optimization algorithm was employed to search the global optimization of the objective function. Ex-perimental results show that this improved algorithm is better than other image segmentation methods, and can solve effectively the problem of over-segmentation that existed with the watershed algorithm. The segmentation results con-form to the visual characteristics of the human eye, so this algorithm is therefore an effective, accurate, and practi-cal image segmentation method.