Enhancing Compatibility Accuracy in Jigsaw Puzzle Assembly through Multi-Color Space
Him Kafle, Amit Banerjee · 2024
Ensuring precise pairwise compatibility is pivotal in jigsaw puzzle solving. This study mainly focuses on determining the neighbors of the jigsaw pieces by calculating their compatibility. The objective is to enhance the true positive neighbors of the jigsaws to reduce the computational overheads. For this, the paper studies the compatibility measures proposed by the researchers utilizing various color spaces, such as RGB, LAB, and HSV. Finally, it proposes a compatibility measure using a weighted average on the multi-color space. The proposed methodology examines the jigsaw pieces in RGB, LAB, and HSV formats and computes the weighted average for all four sides of each pair of square jigsaw pieces. The key idea of our work is to explore a measure that can be applicable to diverse images, such as underwater and fire images. The paper evaluates the performance of the proposed metric on various standard datasets, such as MIT-432, McGill-540, Pomeranz-805, the EUVP dataset for underwater images, and the D-fire dataset for fire images. The experimental evaluation shows that the proposed measure enhances the true positive neighbor selection by 10% for both colored and underwater pictures in comparison to the approaches that use single-color space.