This paper introduces a new framework for two-phase image segmentation, namely the Fuzzy Region Competition. A generic formulation is developed that extends in a convex way several existing supervised or unsupervised models. The method provides globally minimizing solutions and can be solved with efficient convex optimization tools. We also show by some classical examples that the results are in practice almost insensitive to initial conditions. Motivated by medical applications, in particular angiography, we finally derive a fast algorithm for segmenting images into two non-overlapping smooth regions. Compared to existing methods, this last model has the unique advantage of featuring closedform solutions for the approximation functions in each region based on normalized convolutions. Results are shown in 2D and 3D.