With the rapid spread of image capture devices, the size of family album grows quickly up. It is difficult for family members to find out an expected photo just based on their fragmentary memory. In this paper, we attempt to develop a system named Photo Linker, to help people quickly get their expected photos by fully exploring the clues remembered by themselves. The key idea is to build the relationship among all photos by shared faces and scenes. The system includes three key components: (1) face detection and recognition; (2) scene classification; (3) association searching. In particular, an algorithm called "Association Correction" based on FP-Growth is discovered to improve the face recognition rate of family members. Moreover, a friendly interface is designed and implemented to facilitate the photo searching process. The experimental results show that the system can effectively and efficiently build the relationship among photos and make the photo finding process more convenient.