A new method for visualisation and segmentation of vessel structures in 3D magnetic resonance angiography (MRA) images is presented. This method uses a simple statistical model of the information stored along parallel rays within the data set to derive a 2D projection image. Although similar to the maximum image projection (MIP) method, the new method uses a single parameter to achieve a higher contrast-to-noise ratio at a modest computational cost. The same idea is employed to provide a means of segmenting a 3D data set in order to derive a region of support for the purpose of reconstructing image sequences with high temporal resolution.