Mosaicing of Torn paper documents
Recovery of ripped-up documents is a problem that springs up in Archival study and Investigation science. Efficient and successful joining of torn pieces of papers to reconstruct the original documents is an important and challenging issue which if automated would alleviate the manual effort, which is difficult and time-consuming. This paper focuses on the reconstruction of documents ripped up by hand.
Document reconstruction problems can be solved in two steps, finding an initial set of matching fragment pairs as candidate matches followed by resolving the ambiguity among these candidate matches to reconstruct the original document. Document fragments are represented by their contours and candidate matches are obtained by local curve matching of individual fragments. Local curve matching may result in many ambiguous matches and hence a search technique is required to resolve the ambiguities for global reconstruction.The best-first strategy is one such search technique which selects and merges the best matching pair to form a new fragment. This is repeated until there is only one fragment left which gives the recovered document.
Since document fragments often have similar contour segments in real applications, invalid matching occurs frequently using the local search strategies. Intensive backtracking in such cases makes these local search techniques inefficient. To overcome this, we resort to a global approach to disambiguate the candidate matches. The global consistency approach is based on the fact although some incorrect matches may score higher than the correct one locally, the global score of the correct configuration will be much higher than that of an invalid one. A globally consistent solution is found by exploiting the relationships among all candidate matches.
Approach
This approach uses Dynamic programming methodology to solve the problem of matching contours. To know about the approach in detail read here. For details about the implementation of GUI and dependencies read here.
Successful Merge results
Authors and Contributors
Automosaic was created by Srikar Polipalli, Srinivasan Sivanandan (@srinivasans) ,Subham Ghosh (@raltgz), Sourya Kakarla (@ma08) and Srinath Mandava (@srinath9795)
This project was placed 3rd (Bronze winning) in Inter hall OpenSoft contest 2014 conducted by Technology Students Gymkhana, Indian Institute of Technology, Kharagpur.
Support or Contact
Having trouble with AutoMosaic ? Check out our user manual or contact us and we’ll help you sort it out.
References
- Biswas, P. Bhowmick, and B. B. Bhattacharya, Reconstruction of Torn Documents Using Contour Maps, In Proc. ICIP 2005, III:517-520.
- L Zhu, Z Zhou and D Hu, Globally Consistent Reconstruction of Ripped-Up Documents, IEEE Transactions on Pattern Analysis and Machine Intelligence, Volume 30 , Issue 1 (January 2008)
- P. Faber, A Theoretical Framework for Relaxation Processes in Pattern Recognition: Application to Robust Nonparametric Contour Generalization, IEEE Trans. Pattern Analysis and Machine
- C. Solana, E. Justino, L.S. Oliveira, and F. Bortolozzi, Document Reconstruction Based on Feature Matching, Proc. 18th Brazilian Symp. Computer Graphics and Image Processing, pp. 163-170, 2005
- J. McBride and B.B. Kimia, Archaeological Fragment Reassembly Using Curve-Matching, Proc. IEEE Conf. Computer Vision and Pattern Recognition, pp. 1-8, 2003.
- O. Faugeras and M. Berth, Improving Consistency and Reducing Ambiguity in Stochastic Labeling: An Optimization Approach, IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 3, no. 4, pp. 412-424, 1981.Intelligence, vol. 25, no. 8, pp. 1021-1027, Aug. 2003.