Image Segmentation Using Glowworm Swarm Optimization for Finding Initial Seed

The segmentation is a challenging task in digital image processing. There are various methods available for performing segmentation. Clustering based image segmentation is an important technique in image segmentation scenarios. Under this Expectation-Maximization algorithm (GMM- EM) for image segmentation is taken here for analysis. This algorithm is a popular tool for estimating model parameters, especially mixture models and it is a hill-climbing approach. But Image segmentation using GMM-EM algorithm has several drawbacks, such as local maxima, plateau and ridges. The important drawbacks are local maxima which mean that it is sensitive to initialization. To overcome the problem of local maxima this paper proposes a new Glowworm swarm optimization (GSO) algorithm along with EM algorithm. In the initial stage, the GSO is executed for a short period for automatic clustering. The result from the GSO is used as the initial optimal seed of the EM algorithm. Compared to the existing method of segmentation using only with GMM, this paper GSO based EM gives better results with minimum time and errors.