The diagnosis of cardiovascular disease is usually assisted by resonance angiography (MRA) or computed tomography angiography (CTA) imaging. The identification of abnormal vascular architecture from angiographic three-dimensional images is therefore crucial to the diagnosis of cardiovascular disease. Automated detection and quantification of vascular structure and architecture thus holds significant clinical value. In this work, we employ a Lindenmayer system to represent vascular trees from angiographic images and describe a quantitative measure based on the Tokunaga taxonomy to differentiate vascular architectures. Synthetic vessel architectures with varying bifurcation patterns were compared and results showed that this architectural measure is proportional to the level of branching. In real MRA images, this measure was able to differentiate between normal and abnormal intracerebral vasculature containing an aneurysm. Hence, this methodology not only allows for compact representation of vascular architectures but also provides a quantitative metric of bifurcation complexity, which has the potential to characterize different types of vascular abnormalities.