Solving computational square jigsaw puzzles with a novel pairwise compatibility measure

Abstract

The most challenging aspect of rebuilding the puzzle is finding out the right pair of pieces of the image. To do this, we need, an accurate estimation of the pairwise compatibility measure between local patches and the assembly strategy. In this paper, we propose a novel pairwise compatibility measure for the assembly of computational square jigsaw puzzles using Gist and color distance. This possible combination of gradient and color features improves the assembling pieces and can deal with the majority of the problems encountered in state-of-the-art. We also propose the rotation-based strategy to enable working on multiple parts and rebuild completed puzzles from local matching candidates. The experimental results on the frequently used data-sets sufficiently show that the suggested compatibility metric surpasses the latest approaches.

Keywords

• Compatibility metrics
• Jigsaw puzzle solver
• Mosaicing image
• Gist scene descriptor
• Global image feature
• Image reconstruction