This paper presents a velocity estimator based on a digital version of the so called Elementary Motion Detector (EMD). Inspired by insect vision, this model benefits from a low complexity motion detection algorithm and is able to estimate velocities in four directions. It can handle noisy images with a pre-filtering step which highlights the important features to be detected. Using a specific velocity tuned detector called Elementary Velocity Detector (EVD) applied to different resolutions of the same input, it gains time efficiency by estimating different speeds in parallel. The responses of the different EVDs are then combined together at the input resolution size.