Jigsaw puzzle solving techniques and applications: a survey

Markaki, Smaragda; Panagiotakis, Costas

doi:10.1007/s00371-022-02598-9

Cited by 11 publications

(5 citation statements)

References 65 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This problem is simplified by limiting the registration to a classification problem of nine classes, e. g. a center image and its 8-connected neighbors. The problem shows similarities with the jigsaw puzzle problem [5]. However, no inference can be made based on the image's shape.…”

Section: Motivation and Objectivementioning

confidence: 99%

P01 - Non-Overlap Image Registration

Siemens¹,

Kästner²,

Reithmeier³

2023

Poster

View full text Add to dashboard Cite

This work aims to predict the relative position of non-overlapping image pairs consisting of a moving and a fixed image. For this purpose, a modified VGG16 convolutional neural network is proposed. The network is trained on a large dataset with microtopographic measurement data of different materials and processing methods. The proposed method shows a high prediction accuracy on the test data and the potential for developing non-overlap registration algorithms.

show abstract

Section: Motivation and Objectivementioning

confidence: 99%

P01 - Non-Overlap Image Registration

Siemens¹,

Kästner²,

Reithmeier³

2023

Poster

View full text Add to dashboard Cite

show abstract

“…In a recent survey addressing the solving of Jigsaw puzzles [13], emphasis is placed on the utilization of deep learning techniques exclusively for methodologies employing square or rectangular patches extracted from images. This particular perspective proves valuable for discerning the relationship between two fragments belonging to the same object [18] or predicting the permutation of image patches to compose a comprehensive image [11].…”

Section: Related Workmentioning

confidence: 99%

Matching Apictorial Puzzle Pieces Using Deep Learning

Chiș

2024

Studia UBB Informatica

View full text Add to dashboard Cite

Finding matches between puzzle pieces is a difficult problem relevant to applications that involve restoring broken objects. The main difficulty comes from the similarity of the puzzle pieces and the very small difference between a pair of pieces that almost match and one that does. The proposed solution is based on deep learning models and has two steps: firstly, the pieces are segmented from images with a U-Net model; then, matches are found with a Siamese Neural Network. To reach our goal, we created our own dataset, containing 462 images and just as many masks. With these masks, we built 3318 pairs of images, half of them representing pieces that fit together and half that do not. Our most relevant result is estimating correctly for 290 out of 332 pairs whether they match. Keywords: U-Net, Siamese architecture, Edge-matching, Puzzle Pieces.

show abstract

“…1) and real data. The real datasets of coastlines correspond to various areas around the world and come from the European Environment Agency 3 , and the National Oceanic and Atmospheric Administration (NOAA), National Centers for Environmental Information 4 . Specifically, we present the following pairs of coastlines, most of them share common geological origin:…”

Section: Datasetsmentioning

confidence: 99%

“…This problem can be viewed as an extension of the jigsaw puzzle problem [1][2][3][4], where only rotational and translational transformations are supported. The jigsaw puzzle problem requires assembling a given number of puzzle pieces of arbitrary or specific shape (e.g., square) from a broken single object.…”

Section: Introductionmentioning

confidence: 99%

Coastline matching via a graph-based approach

et al. 2022

Self Cite

View full text Add to dashboard Cite

This paper studies the problem of unsupervised detection of geometrically similar fragments (segments) in curves, in the context of boundary matching. The goal is to determine all pairs of sub-curves that are geometrically similar, under local scale invariance. In particular, we aim to locate the existence of a similar section (independent of length and/or orientation) in the second curve, to a section of the first curve, as indicated by the user. The proposed approach is based on a suitable distance matrix of the two given curves. Additionally, a suitable objective function is proposed to capture the trade-off between the similarity of the common sub-sequences and their lengths. The goal of the algorithm is to minimize this objective function via an efficient graph-based approach that capitalizes on Dynamic Time Warping to compare the two subcurves. We apply the proposed technique in the context of geometric matching of coastline pairs. This application is crucial for investigating the forcing factors related to the coastline evolution. The proposed method was successfully applied to global coastline data, yielding a bipartite graph with analytical point-to-point correspondences.

show abstract

Jigsaw puzzle solving techniques and applications: a survey

Cited by 11 publications

References 65 publications

P01 - Non-Overlap Image Registration

P01 - Non-Overlap Image Registration

Matching Apictorial Puzzle Pieces Using Deep Learning

Coastline matching via a graph-based approach

Contact Info

Product

Resources

About