Abstract: Document unwarping attempts to undo physical deformations of the paper and recover a ’flatbed’ scanned document-image for downstream tasks such as OCR. Current state-of-the-art relies on global unwarping of the document which is not robust to local deformation changes. Moreover, a global unwarping often produces spurious warping artifacts in less warped regions to compensate for severe warps present in other parts of the document. In this paper, we propose the first end-to-end trainable piece-wise unwarping method that predicts local deformation fields and stitches them together with global information to obtain an improved unwarping. The proposed piece-wise formulation results in 4% improvement in terms of multi-scale structural similarity (MS-SSIM) and shows better performance in terms of OCR metrics, character error rate (CER) and word error rate (WER) compared to the state-of-the-art.

*** Code Release Update: *** We still need some time to release the code and models for this paper. In case you're interested to compare with this method and have an upcoming deadline, please feel free to contact me at: sadas@cs.stonybrook.edu. You can also send me your evaluation set (If it's not DocUNet Benchmark), I'll send you the unwarped results.

Unwarping Demo of Different Views