Multiscale Image Segmentation Using Energy Minimization

We consider the role of multiscale prior information of the object in the form of a Bayesian framework to address the posterior inference problem. The multiscale prior is implicitly estimated from the given image. We show how the multiscale prior effectively exploits the available image data for hierarchical modeling and exploiting posterior inference scheme to determine the posterior likelihood at each iteration with definite number of iteration steps. Extensive experiments show that this method achieves robust multiscale image segmentation results in the presence of dynamic Gaussian noises.