Visual processing for action resists similarity of relevant and irrelevant object features