Edge and color contexts based object representation and tracking

In this paper, we present a novel framework of visual tracking based on the edge-color and edge-edge contexts. By employing the non-uniform quantization of the color space to highlight the object, and canny operator and motion detection to gain the object edges, we derive an edge-color co-occurrence centroid (ECCC) based object appearance representation. We subsequently select the stable centroids to estimate the one step shifting vector for localizing the object, and adopt the edge-edge distance histograms (DHs) matching to acquire the zooming factor for updating the object scale in the next frame. Experiments on both standard and real world video sequences demonstrate that the proposed method is robust enough for tracking the object in scenarios of partial occlusion and background disturbance.