Nonsubsampled Contourlet Transform based expectation maximization method for segmentation of images

In this paper, a new algorithm for Gaussian Mixture Model based image segmentation is presented. The proposed method is an extension to traditional Gaussian Mixtures Expectation-Maximization Segmentation. The standard Gaussian Mixture Model(GMM) is the most widely used method for image segmentation where the model parameters can be estimated by the Expectation Maximization algorithm. The Gaussian Mixture Model considers each pixel as independent and does not take into account the spatial correlation between the neighbouring pixels. Hence the segmentation result obtained using standard GMM is highly sensitive to noise. In this paper, the Nonsubsampled Contourlet Transform is employed to overcome this drawback. The Nonsumsampled contourlet transform employs a trous filters for pyramidal decomposition which are isotropic and shift invariant. The a trous algorithm uses a low pass filter to obtain successive approximations of the original image. The low pass output of the nonsubsampled contourlet transform incorporates the spatial correlation between the neighboring pixels and hence gives better segmentation accuracy when subjected to Expectation Maximization. The presented algorithm is used to segment synthetic images and real images with regions of varying pixel intensities. The performance of the algorithm is compared with the standard GMM and the recently proposed method which addresses the pixel labeling problem by replacing each pixel value in an image with the average value of its neighbors including itself. The results suggested that the proposed algorithm is superior in segmentation accuracy when compared with the existing methods.