This paper describes a multi-view based robust face recognition method which overcomes variation in facial pose. Many recent researches of face recognition deal with pose variation problem, but most of those are not practical and focus on laboratory environment which is controlled for experiments. In this paper, we propose high dimensional feature for multi-view based face recognition and real-view based matching method for real world scenario. Experimental results shows that this approach is robust and can produce high accuracy in real world face recognition scenario.