How Are Representations Affected by Scene Statistics in an Adaptive Active Vision System?