This paper describes an approach to detect vessel bifurcations and crossovers in retinal images. This approach is based on an novel method to estimate multiple orientations at each pixel of a gray image. The main orientations are provided by directional openings whose outputs are regularized in order to extend the orientation information to the whole image. The detection of vessel bifurcations is based on the coexistence of two or more than two different main orientations in the same pixel. Results on retinal images show the robustness and accuracy of the proposed method to detect vessel bifurcations.