Abstract: To help users play the jigsaw puzzle more efficiently while enjoying the game, a navigational jigsaw puzzle method is proposed, which is driven by local image region matching. Firstly, all of the jigsaw puzzle pieces are automatically detected and segmented from an input original jigsaw puzzle image based on Canny algorithm. Secondly, the quantity descriptor for each jigsaw puzzle piece is automatically constructed by using D2 distribution function and statistical histogram graph. Meanwhile, the adjacent topological relationships among the above-mentioned pieces are described by developing an attributed adjacency graph. Now, when a user needs some help during the assembling process, the above two steps will be carried out on the current unfinished jigsaw puzzle image. After that, between the current jigsaw puzzle image and its original jigsaw puzzle image, the piece correspondence relationship will be established according to descriptor similarity, KM(Kuhn-Munkres) algorithm and adjacent topological relationship consistency. Finally, the next feasible candidate piece in the current jigsaw puzzle image will be automatically identified and prompted for the user based on the above correspondence relationship. To validate the proposed method, a corresponding prototype system is developed. Navigated jigsaw puzzle experiments also have been performed on the system with some typical jigsaw puzzle images. The results show that the proposed method,without decreasing the game enjoyments, is intuitive and efficient. Compared with the mainstream jigsaw puzzle methods, the proposed method does not need any real-time tracking or database, and can be worked on smart devices with cameras. Thus, it has the potential to be used more widely.