Algorithm Beats Jigsaw-Solving Record
We have met our match at the genteel pastime of jigsaw puzzles. It seems an algorithm can now whiz through 10,000 pieces in 24 hours. The speedy solver could also help piece together shredded documents or archaeological artefacts.
Andrew Gallagher at Cornell University in Ithaca, New York, wrote the algorithm while working at photography firm Kodak. By mimicking the way a human solves jigsaws, it beat last year’s record of 3300 pieces. The algorithm can even solve multiple puzzles at the same time, where the pieces have been mixed up together.
Unlike other software that only analyses the edges of the pieces, Gallagher’s looks at how colour patterns spread across many pieces. For example, if one piece becomes progressively lighter from left to right, it is likely that the piece nestles between a lighter piece on the left and a darker one on the right.
The algorithm only works on jigsaws with square pieces, which are harder to solve because the shape offers no clues. The algorithm calculates a score for each pair, stores the best matches, and uses these to assemble the whole puzzle.