This paper investigates variational region-level criterion for supervised and unsupervised texturebased image segmentation. The focus is given to the demonstration of the effectiveness and robustness of this region-based formulation compared to most common variational approaches. The main contributions of this global criterion are twofold. First, the proposed methods circumvent a major problem related to classical texture based segmentation approaches. Existing methods, even if they use different and various texture features, are mainly stated as the optimization of a criterion evaluating punctual pixel likelihoods or similarity measure computed within local neighborhood. The former approaches require sufficient dissimilarity between used feature statistics. The latter involve an additional limitation which is the choice of the neighborhood size and shape. These two parameters and especially the neighborhood size significantly influence the classification performances: the neighborhood must be large enough to capture texture structures and small enough to warrant segmentation accuracy. These parameters are often set experimentally. To address these limitations, the proposed methods are stated at the region-level, both for stating the overall variational criterion and the observation-driven texture criterion. It resorts to an energy criterion on image regions: image regions are characterized by non-parametric distributions of their responses to a set of filters. In supervised case the segmentation algorithm consists in the minimization of a similarity measure between regions features and texture prototypes and a boundary based functional that imposes smoothness and regularity on region boundaries. In unsupervised case, the segmentation consists in the maximization of the dissimilarity between regions. The proposed similarity-based criteria are generic and permit optimally fusing various types of texture features. It is defined as a weighted sum of Kullback-Leibler divergences between feature distributions. The optimization of the proposed variational criteria is carried out using a level-set formulation. The effectiveness and the robustness of this formulation at region-level, compared to classical active contour methods, are evaluated for various Brodatz and natural images.