Bottom-up and top-down influences on attentional allocation in natural dynamic scenes