Face detection is an important technique and has been employed in many applications. Many literatures have been proposed to extract the features of face. However, in many cases, the human face may be obscured by other people or objects which cause the face cannot be detected. This paper focuses on how to do partial face detection and discusses what the significant feature parts on face are. We employed SURF algorithm to compute the reliable feature parts on face so that the partial face can be detected. In order to evaluate the degree of importance for each feature part, this paper introduces CV value to appraise each feature part of face. Finally, the experimental result demonstrates the effectiveness of the proposed algorithm.