Bio-inspired Motion-Based Object Segmentation

Although motion extraction requires high computational resources and normally produces very noisy patterns in real sequences, it provides useful cues to achieve an efficient segmentation of independent moving objects. Our goal is to employ basic knowledge about biological vision systems to address this problem. We use the Reichardt motion detectors as first extraction primitive to characterize the motion in scene. The saliency map is noisy, therefore we use a neural structure that takes full advantage of the neural population coding, and extracts the structure of motion by means of local competition. This scheme is used to efficiently segment independent moving objects. In order to evaluate the model, we apply it to a real-life case of an automatic watch-up system for car-overtaking situations seen from the rear-view mirror. We describe how a simple, competitive, neural processing scheme can take full advantage of this motion structure for segmenting overtaking-cars.