A robust modified Gaussian mixture model with rough set for image segmentation

Accurate image segmentation is an essential step in image processing, where Gaussian mixture models with spatial constraint play an important role and have been proven effective for image segmentation. Nevertheless, most methods suffer from one or more challenges such as limited robustness to outliers, over-smoothness for segmentations, sensitive to initializations and manually setting parameters. To address these issues and further improve the accuracy for image segmentation, in this paper, a robust modified Gaussian mixture model combining with rough set theory is proposed for image segmentation. Firstly, to make the Gaussian mixture models more robust to noise, a new spatial weight factor is constructed to replace the conditional probability of an image pixel with the calculation of the probabilities of pixels in its immediate neighborhood. Secondly, to further reduce the over-smoothness for segmentations, a novel prior factor is proposed by incorporating the spatial information amongst neighborhood pixels. Finally, each Gaussian component is characterized by three automatically determined rough regions, and accordingly the posterior probability of each pixel is estimated with respect to the region it locates. We compare our algorithm to state-of-the-art segmentation approaches in both synthetic and real images to demonstrate the superior performance of the proposed algorithm.