International Journal of Computer and Applications

Optical Character Recognition (OCR) can open up understudied historical documents to computational analysis, but the accuracy of OCR software varies. This article reports a benchmarking experiment comparing the performance of Tesseract, Amazon Textract, and Google Document AI on images of English and Arabic text. English-language book scans (n = 322) and Arabic-language article scans (n = 100) were replicated 43 times with different types of artificial noise for a corpus of 18,568 documents, generating 51,304 process requests. Document AI delivered the best results, and the server-based processors (Textract and Document AI) performed substantially better than Tesseract, especially on noisy documents. Accuracy for English was considerably higher than for Arabic. Specifying the relative performance of three leading OCR products and the differential effects of commonly found noise types can help scholars identify better OCR solutions for their research needs. The test materials have been preserved in the openly available “Noisy OCR Dataset” (NOD) for reuse in future benchmarking studies.

/pdf/ocr-with-tesseract-amazon-textract-and-google-document-ai-a-30q7a6rx98.pdf

OCR with Tesseract, Amazon Textract, and Google Document AI: a benchmarking experiment

Mass digitization of historical documents is a challenging problem for optical character recognition (OCR) tools. Issues include noisy backgrounds and faded text due to aging, border/marginal noise, bleed-through, skewing, warping, as well as irregular fonts and page layouts. As a result, OCR tools often produce a large number of spurious bounding boxes (BBs) in addition to those that correspond to words in the document. This paper presents an iterative classification algorithm to automatically label BBs (i.e., as text or noise) based on their spatial distribution and geometry. The approach uses a rule-base classifier to generate initial text/noise labels for each BB, followed by an iterative classifier that refines the initial labels by incorporating local information to each BB, its spatial location, shape and size. When evaluated on a dataset containing over 72,000 manually-labeled BBs from 159 historical documents, the algorithm can classify BBs with 0.95 precision and 0.96 recall. Further evaluation on a collection of 6,775 documents with ground-truth transcriptions shows that the algorithm can also be used to predict document quality (0.7 correlation) and improve OCR transcriptions in 85% of the cases.

https://www.aaai.org/ocs/index.php/AAAI/AAAI15/paper/download/9965/9458

Automatic assessment of OCR quality in historical documents

Digitization and Recognition of handwritten documents have become popular with the growth of advances in Computing. Even after the introduction of new technologies, handwritten documents continue to be the means of communication and recording information in various fields. Handwritten Character Recognition (HCR) is a subject of research for decades. Many HCR techniques have been developed for applications where documents are written by hand like bank cheques, mail addresses, legal documents, medical prescriptions etc. and have achieved satisfactory performance. But active research is still going on in character recognition from historical documents. HCR becomes more challenging for Ancient Scripts due to the presence of various kinds of noise, low intensity, discoloration etc. Vast amount of knowledge in the field of literature, architecture, history and civilization, economic factors, medicine etc. have been inscribed and preserved in historical documents in various forms. Even though several techniques have been developed for HCR for Ancient Scripts, promising results have not been achieved. The paper reviews some of the major works carried out in HCR for Ancient handwritten documents.

Character Recognition in Historical Handwritten Documents – A Survey

Optical character recognition (OCR) engines work poorly on texts published with premodern printing technologies. Engaging the key technological contributors from the IMPACT project, an earlier project attempting to solve the OCR problem for early modern and modern texts, the Early Modern OCR Project (eMOP) of Texas A8M received funding from the Andrew W. Mellon Foundation to improve OCR outputs for early modern texts from the Eighteenth Century Collections Online (ECCO) and Early English Books Online (EEBO) proprietary database products—or some 45 million pages. Added to print problems are the poor quality of the page images in these collections, which would be too time consuming and expensive to reimage. This article describes eMOP's attempts to OCR 307,000 documents digitized from microfilm to make our cultural heritage available for current and future researchers. We describe the reasoning behind our choices as we undertook the project based on other relevant studies; discoveries we made; the data and the system we developed for processing it; the software, algorithms, training procedures, and tools that we developed; and future directions that should be taken for further work in developing OCR engines for cultural heritage materials.

Mass Digitization of Early Modern Texts With Optical Character Recognition

of historical documents helps preserve these documents. As these documents have existed for a long time, various types of noise creep in. In our paper we have analyzed the different types of noise that occur in printed and handwritten historical documents mainly based on Kannada (Kannada is a language used in Karnataka, a southern state in India) documents and created a taxonomy for the same. We have also characterized each noise type based on factors such as their source, their effect on characters and the associated challenges in character recognition. We have also catalogued the different noise detection, removal and restoration techniques that are reported in the literature for each of the prominent noise types, and identified areas relating to noise detection, removal for further research focus.

/pdf/a-language-independent-characterization-of-document-image-b3wxqbng92.pdf

A language independent Characterization of Document Image Noise in Historical Scripts

Degraded character recognition is one of the most challenging topic in the field of Kannada character recognition. The degraded characters which are broken and deformed will have missing features and will be difficult for any recognition method. Rebuilding the degraded character is very important for better recognition. This paper proposes a novel method to rebuild the broken characters. These characters are thinned and the endpoints of the lines are obtained. The line segments are effectively rebuilt so as to preserve the degraded character. Experimental results on this method are presented to establish its efficiency.

A novel local enhancement technique for rebuilding Broken characters in a degraded Kannada script

Recognition of historical printed degraded Kannada characters is not solved completely and remains as a challenge to the researchers still. In this paper, a scale for measuring degradation of a character is proposed. Further, the degradation is characterised to high, medium and low based on this scale, and use it to study the efficiency of the character restoration technique designed. A new approach, fit discriminant analysis (FDA) for recognition is proposed and compares its recognition accuracy with the existing techniques support vector machines (SVM) and Fisher linear discriminant (FLD) analysis. Through extensive experimentation it is established that rebuilding of characters improves the recognition accuracy of learning-based approaches SVM, FDA, and FLD significantly. Further, it is established that the proposed approach FDA gives the best recognition accuracy for historical printed degraded documents. It is also proved that training-testing set applying the proposed degradation measure is required for better recognition accuracy.

N. Sandhya

Papers

A language independent Characterization of Document Image Noise in Historical Scripts

A novel local enhancement technique for rebuilding Broken characters in a degraded Kannada script

An efficient approach for handling degradation in character recognition