DocEnTr: An End-to-End Document Image Enhancement Transformer

DocEnTr: An End-to-End Document Image Enhancement Transformer. Souibgui, M. A., Biswas, S., Jemni, S. K., Kessentini, Y., Fornés, A., Lladós, J., & Pal, U. In 2022 26th International Conference on Pattern Recognition (ICPR), pages 1699–1705, August, 2022. ISSN: 2831-7475

Paper doi abstract bibtex

Document images can be affected by many degradation scenarios, which cause recognition and processing difficulties. In this age of digitization, it is important to denoise them for proper usage. To address this challenge, we present a new encoder-decoder architecture based on vision transformers to enhance both machine-printed and handwritten document images, in an end-to-end fashion. The encoder operates directly on the pixel patches with their positional information without the use of any convolutional layers, while the decoder reconstructs a clean image from the encoded patches. Conducted experiments show a superiority of the proposed model compared to the state-of-the-art methods on several DIBCO benchmarks. Code and models will be publicly available at: https://github.com/dali92002/DocEnTR.

@inproceedings{souibgui_docentr_2022,
	title = {{DocEnTr}: {An} {End}-to-{End} {Document} {Image} {Enhancement} {Transformer}},
	shorttitle = {{DocEnTr}},
	url = {https://ieeexplore.ieee.org/document/9956101},
	doi = {10.1109/ICPR56361.2022.9956101},
	abstract = {Document images can be affected by many degradation scenarios, which cause recognition and processing difficulties. In this age of digitization, it is important to denoise them for proper usage. To address this challenge, we present a new encoder-decoder architecture based on vision transformers to enhance both machine-printed and handwritten document images, in an end-to-end fashion. The encoder operates directly on the pixel patches with their positional information without the use of any convolutional layers, while the decoder reconstructs a clean image from the encoded patches. Conducted experiments show a superiority of the proposed model compared to the state-of-the-art methods on several DIBCO benchmarks. Code and models will be publicly available at: https://github.com/dali92002/DocEnTR.},
	urldate = {2024-01-04},
	booktitle = {2022 26th {International} {Conference} on {Pattern} {Recognition} ({ICPR})},
	author = {Souibgui, Mohamed Ali and Biswas, Sanket and Jemni, Sana Khamekhem and Kessentini, Yousri and Fornés, Alicia and Lladós, Josep and Pal, Umapada},
	month = aug,
	year = {2022},
	note = {ISSN: 2831-7475},
	pages = {1699--1705},
}

Downloads: 0

{"_id":"GGRWrDdGo6safLTMt","bibbaseid":"souibgui-biswas-jemni-kessentini-forns-llads-pal-docentranendtoenddocumentimageenhancementtransformer-2022","author_short":["Souibgui, M. A.","Biswas, S.","Jemni, S. K.","Kessentini, Y.","Fornés, A.","Lladós, J.","Pal, U."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","title":"DocEnTr: An End-to-End Document Image Enhancement Transformer","shorttitle":"DocEnTr","url":"https://ieeexplore.ieee.org/document/9956101","doi":"10.1109/ICPR56361.2022.9956101","abstract":"Document images can be affected by many degradation scenarios, which cause recognition and processing difficulties. In this age of digitization, it is important to denoise them for proper usage. To address this challenge, we present a new encoder-decoder architecture based on vision transformers to enhance both machine-printed and handwritten document images, in an end-to-end fashion. The encoder operates directly on the pixel patches with their positional information without the use of any convolutional layers, while the decoder reconstructs a clean image from the encoded patches. Conducted experiments show a superiority of the proposed model compared to the state-of-the-art methods on several DIBCO benchmarks. Code and models will be publicly available at: https://github.com/dali92002/DocEnTR.","urldate":"2024-01-04","booktitle":"2022 26th International Conference on Pattern Recognition (ICPR)","author":[{"propositions":[],"lastnames":["Souibgui"],"firstnames":["Mohamed","Ali"],"suffixes":[]},{"propositions":[],"lastnames":["Biswas"],"firstnames":["Sanket"],"suffixes":[]},{"propositions":[],"lastnames":["Jemni"],"firstnames":["Sana","Khamekhem"],"suffixes":[]},{"propositions":[],"lastnames":["Kessentini"],"firstnames":["Yousri"],"suffixes":[]},{"propositions":[],"lastnames":["Fornés"],"firstnames":["Alicia"],"suffixes":[]},{"propositions":[],"lastnames":["Lladós"],"firstnames":["Josep"],"suffixes":[]},{"propositions":[],"lastnames":["Pal"],"firstnames":["Umapada"],"suffixes":[]}],"month":"August","year":"2022","note":"ISSN: 2831-7475","pages":"1699–1705","bibtex":"@inproceedings{souibgui_docentr_2022,\n\ttitle = {{DocEnTr}: {An} {End}-to-{End} {Document} {Image} {Enhancement} {Transformer}},\n\tshorttitle = {{DocEnTr}},\n\turl = {https://ieeexplore.ieee.org/document/9956101},\n\tdoi = {10.1109/ICPR56361.2022.9956101},\n\tabstract = {Document images can be affected by many degradation scenarios, which cause recognition and processing difficulties. In this age of digitization, it is important to denoise them for proper usage. To address this challenge, we present a new encoder-decoder architecture based on vision transformers to enhance both machine-printed and handwritten document images, in an end-to-end fashion. The encoder operates directly on the pixel patches with their positional information without the use of any convolutional layers, while the decoder reconstructs a clean image from the encoded patches. Conducted experiments show a superiority of the proposed model compared to the state-of-the-art methods on several DIBCO benchmarks. Code and models will be publicly available at: https://github.com/dali92002/DocEnTR.},\n\turldate = {2024-01-04},\n\tbooktitle = {2022 26th {International} {Conference} on {Pattern} {Recognition} ({ICPR})},\n\tauthor = {Souibgui, Mohamed Ali and Biswas, Sanket and Jemni, Sana Khamekhem and Kessentini, Yousri and Fornés, Alicia and Lladós, Josep and Pal, Umapada},\n\tmonth = aug,\n\tyear = {2022},\n\tnote = {ISSN: 2831-7475},\n\tpages = {1699--1705},\n}\n\n\n\n","author_short":["Souibgui, M. A.","Biswas, S.","Jemni, S. K.","Kessentini, Y.","Fornés, A.","Lladós, J.","Pal, U."],"key":"souibgui_docentr_2022","id":"souibgui_docentr_2022","bibbaseid":"souibgui-biswas-jemni-kessentini-forns-llads-pal-docentranendtoenddocumentimageenhancementtransformer-2022","role":"author","urls":{"Paper":"https://ieeexplore.ieee.org/document/9956101"},"metadata":{"authorlinks":{}},"html":""},"bibtype":"inproceedings","biburl":"https://bibbase.org/zotero/fsimonetta","dataSources":["pzyFFGWvxG2bs63zP","MSmFwabid8wvXA9PQ"],"keywords":[],"search_terms":["docentr","end","end","document","image","enhancement","transformer","souibgui","biswas","jemni","kessentini","fornés","lladós","pal"],"title":"DocEnTr: An End-to-End Document Image Enhancement Transformer","year":2022}