Starting from a variational formulation, we present a model for image segmentation that employs both region statistics and edge information. This combination allows for improved flexibility, making the proposed model suitable to process a wider class of images than purely region-based and edge-based models. We perform several simulations with real images that attest to the versatility of the model. We also show another set of experiments on images with certain pathologies that suggest opportunities for improvement.