This talk addresses three significant problems in ultrasound image processing: radio-frequency (RF) recovery, image reconstruction and segmentation. With the purpose of statistical data processing, more important than having good looking images is to have standardized and realistic models to describe the data. Here, a nonlinear compression parametric function is used to model the log-compressed ultrasound images displayed by the ecographs, providing an estimate of the RF signal, which is known to be Rayleigh distributed. Image reconstruction from noisy and incomplete observations is usually an ill-posed problem. A Bayesian framework may be adopted do deal with this task by minimizing a two-term energy function. The first term pushes the solution toward the data and the second regularizes the solution. A Bayesian algorithm for ultrasound image reconstruction and de-noising is proposed where an edge preserving prior is used to reduce the smoothing effect at the transitions. The prior distribution is based on log-Euclidean potential functions which are particular suitable in reconstruction problems under the constraint of positivity, which is the case, where the observations are modeled by a Rayleigh distribution. Under the motivation of identifying (labeling) potential foci of atherosclerotic plaque rupture according to its echo-morphological characteristics, a ‘new’ segmentation method is proposed. The local characterization of plaques is performed by using a 3D Graph-Cuts robust method. This methodology allows to binary segment the data by minimizing an energy function that uses spatial correlation among neighboring pixels/voxels in order to remove small misclassified regions. Results show that this labeling procedure is less noisy and favors clustering, being more meaningful from a clinical point of view than the one obtained with simple thresholding.