Maximum Likelihood Estimation of Gaussian Mixture Models using PSO for Image Segmentation

Gaussian mixture model-based clustering algorithm is one of the advanced techniques applied to enhance the image segmentation performance. However, segmentation process is still encountering some critical difficulties: the model is quite sensitive to initialization, and easily gets trapped in local maxima. To address these problems in image segmentation, we proposed a novel clustering algorithm employing the arbitrary covariance matrices that uses particle swarm optimization for the estimation of Gaussian Mixture Models. Such model can be able to prevent the effective use of population-base algorithms during clustering, and the arbitrary covariance matrices allow independently updating individual parameters, while retaining the validity of the matrix. Then we present the solution that involves an optimization formulation to identify the correspondence between different parameter orderings of candidate solutions. The experimental results show that our method provides a simple segmentation process and the better quality of segmented images comparing to other methods. Furthermore, our method would provide an advanced technique for multi-dimensional image analysis and computer vision systems that can apply for various science and technology sector.