The first step of creating an automatic jigsaw puzzle solver is figuring out how to align the puzzle pieces. Using an image recognition algorithm, the machine can recognize the inner and outer locks on the pieces and multiple points of convexity to determine which way they should be oriented in the puzzle. Then, it creates several versions of the puzzle piece image, taking into account the different orientations and rotations of the pieces. Finally, the process joins the pieces together to create the final image.
Automatic Jigsaw puzzle solvers may take advantage of parallel processing. The goal of the program is to extract puzzle pieces from the image, figure out where they go together, and assemble the scene. The final output is an image of the assembled scene. The process may require non-imagelike intermediate representations to complete the puzzle. If the algorithm is able to solve a jigsaw puzzle without a reference image, it will be much faster.
The second method of an automatic jigsaw puzzle solver is machine learning. Machine learning algorithms are trained to identify patterns of jigsaw puzzles. By applying machine learning algorithms to the puzzle images, these programs are capable of solving jigsaw puzzles in a variety of sizes, shapes, and types. In addition to solving puzzles automatically, they also improve hand-eye coordination, visual acuity, and color and shape identification.