Shredded banknotes reconstruction using AKAZE points

Area of Science:

• Computer Vision
• Pattern Recognition
• Digital Forensics

Background:

• Shredded banknote reconstruction is a complex problem akin to large-scale jigsaw puzzles.
• Existing methods struggle with real-world banknote fragments due to unique challenges.
• High computational complexity necessitates advanced reconstruction techniques.

Purpose of the Study:

• To develop an effective method for reconstructing shredded banknotes.
• To improve fragment matching and alignment accuracy in real-world scenarios.
• To establish a reliable criterion for evaluating reconstruction success rates.

Main Methods:

• Utilized the AKAZE (Accelerated-KAZE) keypoint-based method for effective fragment matching.
• Proposed a novel fragment alignment technique using convex hulls of matched keypoints to estimate orientation.
• Developed a new criterion to assess the success rates of reconstructed banknotes.

Main Results:

• The AKAZE method demonstrated effectiveness in matching banknote fragments.
• The proposed convex hull-based alignment method accurately estimated fragment orientations.
• New datasets of real and synthetic banknote fragments were created for comprehensive testing.

Conclusions:

• The study presents a significant advancement in shredded banknote reconstruction.
• The integrated approach of AKAZE matching and novel alignment shows promise for real-world applications.
• The developed success rate criterion provides a quantitative measure for reconstruction quality.