One of the most important tasks in Cardiac Magnetic resonance Cine (CMRC) consists in identifying and describing normal and abnormal dynamic heart patterns, a task usually performed by physicians. Segmentation and tracking may support decisions during a particular treatment, but their performance is dependent on the quality of the video. The acquired signal, on the other hand, is contaminated with noise coming from physiological movements and devices, resulting in cardiac blurred boundaries. This paper presents a novel method that automatically identifies flow heart patterns by establishing similarities between two consecutive frames to which a local jet feature analysis has been applied. Once a vector motion field is calculated, spatially connected regions with minimal variance are found as the sources of movement and different statistics objectively estimate movement patterns of these regions. The utility of this method is illustrated by comparing the temporal series of these regions between normal and abnormal patients.