No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Quality Assessment and Restoration of Typewritten Document Images
Abstract
. We present a useful method for assessing the quality of a typewritten document image and automatically selecting an optimal restoration method based on that assessment. We use five quality measures that assess the severity of background speckle, touching characters, and broken characters. A linear classifier uses these measures to select a restoration method. On a 139-document corpus, our methodology reduced the corpus OCR character error rate from 20.27% to 12.60%. Key words: Optical character recognition - Document quality assessment - Document image restoration. 1. Introduction Not all of today's OCR is performed on clean laser-written documents. Many organizations have huge archives of typewritten material, much of it of marginal quality. For example, the U.S. Department of Energy has an archive of over 300 million classified documents consisting of fixed-width, fixed-pitch typewritten documents, teletypewriter output, and carbon copies on aging fibrous paper. As part of the dec...
... For example, [5], [12]- [14], [18]- [23], [33], [35], [37], [47], [55]- [81], [83]- [88] For example, [5], [20], [28], [31], [37]- [39], [41]- [48], [50], [54], [88], [90]- [106] For example, [26], [29], [51], [90], [52] For example, [10], [11], [12], [38], [113], [ ...
... For modeling-based approaches, the consumers of document images are typically a downstream process, for example, OCR. As a result, document image quality is defined based on, for example, the OCR accuracy and DIQA metrics are factors that can be used to reliably predict OCR accuracy [5], [13], [41]- [47]. Meanwhile, in the human perception-based approaches, objective document image quality measures are computed according to human perception, as the end-users/clients of document images may be individuals (humans) [5], [10]- [12], [26], [38], [39], [48]- [55]. ...
... Filtering methods alone and along with quality assessment have also been proposed in the literature to perform preprocessing on historical and degraded document images to enhance their quality and readability [5], [47], [78]- [83]. Eye-tracking technology has further been used to aggregate the reading behavior of a group of individuals for providing objective quality feedback [55]. ...
The rapid emergence of new portable capturing technologies has significantly increased the number and diversity of document images acquired for business and personal applications. The performance of document image processing systems and applications depends directly on the quality of the document images captured. Therefore, estimating the document's image quality is an essential step in the early stages of the document analysis pipeline. This paper surveys research on Document Image Quality Assessment (DIQA). We first provide a detailed analysis of both subjective and objective DIQA methods. Subjective methods, including ratings and pair-wise comparison-based approaches, are based on human opinions. Objective methods are based on quantitative measurements, including document modeling and human perception-based methods. Second, we summarize the types and sources of document degradations and techniques used to model degradations. In addition, we thoroughly review two standard measures to characterize document image quality: Optical Character Recognition (OCR)-based and objective human perception-based. Finally, we outline open challenges regarding developing DIQA methods and provide insightful discussion and future research directions for this problem. This survey will become an essential resource for the document analysis research community and serve as a basis for future research.
... In relation to the human readability, document images are presented to individuals and asked them to look at the documents, read and assess their readabilities, whereas as in relation to the machine readability, document images are generally subjected to an optical character recognition (OCR) process and the accuracy obtained from OCR is considered as degree of machine readability of the documents [22]. In both human and machine readability, poor quality document images might result in low document image readability (documents are unreadable) [17][18][19][20][21][22][23]. ...
... In addition to the methods in the literature that provide only a metric as the quality of a document image, there are also many studies which propose employing different preprocessing and filtering techniques in order to enhance the quality of document images (especially historical and degraded ones), results in improving the readability of those documents [17][18][19][20][21][22][23][24]. Different pre-processing and filtering operations [17,18,19], binarization methods [20], and quality assessment followed by filtering operations [21,22,23] have been used to improve document image quality and readability. ...
... In addition to the methods in the literature that provide only a metric as the quality of a document image, there are also many studies which propose employing different preprocessing and filtering techniques in order to enhance the quality of document images (especially historical and degraded ones), results in improving the readability of those documents [17][18][19][20][21][22][23][24]. Different pre-processing and filtering operations [17,18,19], binarization methods [20], and quality assessment followed by filtering operations [21,22,23] have been used to improve document image quality and readability. Eye tracking technology has also been used to aggregate reading behaviour of many readers in order to provide objective quality feedback [24]. ...
Based on the hypothesis that a good / poor quality document image is most probably a readable / unreadable document, document image quality and readability have interchangeably been used in the literature. These two terms, however, have different meanings implying two different perspectives of looking at document images by human being. In document images, the level of quality and the degree of readability may have a relation / correlation considering human perception. However, to the best of our knowledge there is no specific study to characterise this relation and also validate the abovementioned hypothesis. In this work, at first, we created a dataset composed of mostly camera-based document images with various distortion levels. Each document image has then been assessed with regard to two different measures, the level of quality and the degree of readability, by different individuals. A detailed Normalised Cross Correlation analysis along with different statistical analysis based on Shapiro-Wilks and Wilcoxon tests has further been provided to demonstrate how document image quality and readability are linked. Our findings indicate that the quality and readability were somewhat different in terms of the population distributions. However, the correlation between quality and readability was 0.99, which implies document quality and readability are highly correlated based on human perception.
... Néanmoins l'écart constaté avec la sélection manuelle est important (environ 5%), d'autant plus que les auteurs ne comparent pas leur méthode avec la sélection optimale qui peut être obtenue à partir de la vérité-terrain de l'OCR en utilisant une méthode de type "brute-force" (test de toutes les méthodes possibles). Les auteurs de [CHK99] utilisent une approche similaire dans l'objectif de sélectionner une méthode de restauration optimale pour chaque image de document et par conséquent d'améliorer les performances de l'OCR. Les auteurs disposent d'un corpus (de 139 documents) et de leurs vérités-terrains OCR associées ainsi que de 14 méthodes de restauration. ...
... D'autres descripteurs comme les moments statistiques calculés sur les profils verticaux et horizontaux peuvent caractériser d'une certaine façon la mise en page d'un document (nombre de ligne ou de colonnes). -Les descripteurs moyens-niveaux utilisent des connaissances sur la composition physique du document par exemple, la taille moyenne d'un caractère, l'intensité moyenne (local ou global) de l'encre, etc. Plusieurs descripteurs de ce type ont déjà été utilisés pour caractériser des images de documents binaires [BKN95,CHK99,CHKW97]. Ces descripteurs doivent être adapté aux images en niveaux de gris, et peuvent être complétés par des descripteurs résultants d'analyse plus complexes comme l'étude de la similarité structurelle de deux documents. ...
... des épaisseurs des caractères composantes connexes de l'image Ce descripteur semble peu robuste aux changements de police au sein d'une même page.Évaluation de la qualité par des descripteursEn 1999 M.Cannon[CHK99] introduit une dernière mesure Font Size Factor (FSF) qui norme la taille de la fonte (afin d'être invariant en échelle). FSF est définie par: CHK99]. ...
This PhD. thesis deals with quality evaluation of digitized document images. In order to measure the quality of a document image, we propose to create new features dedicated to the characterization of most commons degradations. We also propose to use these features to create prediction models able to predict the performances of different types of document analysis algorithms. The features are defined by analyzing the impact of a specific degradation on the results of an algorithm and then used to create statistical regressors.The relevance of the proposed features and predictions models, is analyzed in several experimentations. The first one aims to predict the performance of different binarization methods. The second experiment aims to create an automatic procedure able to select the best binarization method for each image. At last, the third experiment aims to create a prediction model for two commonly used OCRs. This work on performance prediction algorithms is also an opportunity to discuss the scientific problems of creating ground-truth for performance evaluation.
... Blando et al. developed two metrics namely white speckle factor (WSF) and broken character factor (BCF) for OCR evaluation. Cannon et al. [11] include small speckle factor (SSF), touching character factor (TCF) and font size factor (FSF) for predicting quality and choose appropriate filter to restore the degraded document. Souza et al. [12] propose a method for automatically selecting the best among the restoration filters. ...
... Souza et al. [12] propose a method for automatically selecting the best among the restoration filters. The above works [10], [11], [12] observe only the noise present in the document image to propose metrics and OCR engines are required to evaluate the quality of document image. In our case, we avoid the use of OCR to a great extent and ensure quality metrics are script independent. ...
... This count gives an idea about the amount of salt and pepper noise, which indirectly points to the quality of a document. Cannon et al. [11] and Souza et al. [12] use median filtering to remove speckles and improve degraded document image. Figures 6 (a) and (b) show normalized counts for number of pixels converted by median filter of size 3x3 and 5x5, respectively. ...
We propose a set of metrics that evaluate the uniformity, sharpness, continuity, noise, stroke width variance, pulse width ratio, transient pixels density, entropy and variance of components to quantify the quality of a document image. The measures are intended to be used in any optical character recognition (OCR) engine to a priori estimate the expected performance of the OCR. The suggested measures have been evaluated on many document images, which have different scripts. The quality of a document image is manually annotated by users to create a ground truth. The idea is to correlate the values of the measures with the user annotated data. If the measure calculated matches the annotated description, then the metric is accepted; else it is rejected. In the set of metrics proposed, some of them are accepted and the rest are rejected. We have defined metrics that are easily estimatable. The metrics proposed in this paper are based on the feedback of homely grown OCR engines for Indic (Tamil and Kannada) languages. The metrics are independent of the scripts, and depend only on the quality and age of the paper and the printing. Experiments and results for each proposed metric are discussed. Actual recognition of the printed text is not performed to evaluate the proposed metrics. Sometimes, a document image containing broken characters results in good document image as per the evaluated metrics, which is part of the unsolved challenges. The proposed measures work on gray scale document images and fail to provide reliable information on binarized document image.
... These methods utilized quality metrics calibrated from the image scanned for OCR. More recently an approach has surfaced where authors are designing systems to automatically choose a filter that when applied to the degraded document image improves the image quality and therefore the OCR performance [9,13,14]. Here the choice of filter is related to the level and type of degradation present in the document image. ...
... Most are defined to quantitatively capture the different types of degradations seen in large document collections. Eight different quality measures defined in [7,9,10,13] were implemented for these experiments. Some quality metrics aim to measure the same effect, but the method used to calculate the degradation effect has been defined differently by different authors. ...
... Small Speckle Factor (SSF) [9]: This measures the amount of black background speckle in the image. It identifies all the black connected components in an image that contain between 6 and FS pixels. ...
OCR often performs poorly on degraded documents. One approach to improving performance is to determine a good filter to improve the appearance of the document image before sending it to the OCR engine. Quality metrics have been measured in document images to determine what type of filtering would most likely improve the OCR response for that document image. In this paper those same quality metrics are measured for several word images degraded by known parameters in a document degradation model. The correlation between the degradation model parameters and the quality metrics is measured. High correlations do appear in many places that were expected. They are also absent in some expected places and offer a comparison of quality metric definitions proposed by different authors.
... Proper assessment of the quality of document images is therefore beneficial for the OCR process. A similar note could also be drawn in other document processing applications such as document restoration [4] and document image enhancement [5], where document image quality assessment (DIQA) proves to be relevant. ...
... Meanwhile, as pristine images are often unavailable in practical scenarios, much effort has been devoted to the development of no-reference (NR) DIQA models where the extraction of descriptive image features is a critical step. Comparing to hand-crafted features [4], [17], learning based feature extraction has attracted more attention. ...
... If severe enough, either of them can reduce the performance of a document analysis system significantly. Several document degradation models [72,64,73], methods for document quality assessment [74,75], and document enhancement algorithms [76,77,78] have been presented in previous work. One common enhancement approach is window-based morphological filtering [76,77,78]. ...
... These algorithms can be further categorized as manually designed, semi-manually designed, or automatically trained approaches. The kFill algorithm, proposed by O'Gorman [78], is a manually designed approach and has been used by several other researchers [74,79]. Experiments show it is effective for removing salt-and-pepper noise. ...
... • recognition of hand-printed text in forms [16,72,77] • handwriting recognition in personal checks [48] • postal envelope and parcel address readers [46] • OCR in portable and handheld devices [33,34,60] Document image enhancement [12,18,55,88] involves (automatically) choosing and applying appropriate image filters to the source document image to help the given OCR engine better recognize characters and words. ...
... Let us discuss some of the noteworthy papers on OCR-related image enhancement. Cannon et al. [12] introduced QUARC, a system for enhancing and restoring images of typewriter-prepared documents. They devised a set of quality assessment measures and image restoration techniques to automatically enhance typewritten document imagery for OCR processing. ...
This report explores the latest advances in the field of digital document recognition. With the focus on printed document imagery, we discuss the major developments in optical character recognition (OCR) and document image enhancement/restoration in application to Latin and non-Latin scripts. In addition, we review and discuss the available technologies for hand-written document recognition. In this report, we also provide some benchmark results on available OCR engines.
... Au cours du traitement, elle peut également permettre de sélectionner, parmi un ensemble de systèmes, le meilleur OCR pour un document donné (Ablavsky et al., 2003) ou encore de paramétrer au mieux l'OCR en fonction du document. (par exemple par la sélection automatique du système de restauration d'images le plus approprié au document (Cannon et al., 1999)) Enfin, les techniques mises en oeuvre par un système de prédiction peuvent être utilisées en aval du traitement OCR pour permettre de contrôler la qualité de ce dernier grâce au calcul automatique d'une métrique de qualité de l'OCRisation, à l'échelle de la page (afin de palier aux défauts des mesures classiques, souvent calculées à l'échelle de l'ouvrage) afin de remplacer les contrôles visuels, basés sur un échantillonnage des documents, qui sont particulièrement longs, coûteux et bien souvent lacunaires. ...
... Dans l'article (Cannon et al., 1999), les mesures définies dans (Blando et al., 1995) sont cette fois utilisées afin de caractériser la qualité d'images de documents dans le but de sélectionner la méthode de restauration corrective la plus appropriée à chaque image. Les auteurs ajoutent aussi deux nouvelles mesures : la taille de police estimée et le Small Speckle Factor (SSF) qui mesure la quantité de bruit (ou de tâches) présente sur le fond du document. ...
... Because noise-whether biological, mechanical, electrical, or digital-is a fundamental issue in communication, substantial research has aimed at improving document legibility by suppressing a great variety of what are usually considered manifestations of noise, including: ink bleed-through [46], see-through [48], foxing [163], shadows [95], termite bites [151], cross-outs [30], photocopied low-contrast carbon copies [36], low resolution raster images [141], and background texture interference [133] (for a history of image Content courtesy of Springer Nature, terms of use apply. Rights reserved. ...
This article develops theoretical, algorithmic, perceptual, and interaction aspects of script legibility enhancement in the visible light spectrum for the purpose of scholarly editing of papyri texts. Novel legibility enhancement algorithms based on color processing and visual illusions are compared to classic methods in a user experience experiment. (1) The proposed methods outperformed the comparison methods. (2) Users exhibited a broad behavioral spectrum, under the influence of factors such as personality and social conditioning, tasks and application domains, expertise level and image quality, and affordances of software, hardware, and interfaces. No single enhancement method satisfied all factor configurations. Therefore, it is suggested to offer users a broad choice of methods to facilitate personalization, contextualization, and complementarity. (3) A distinction is made between casual and critical vision on the basis of signal ambiguity and error consequences. The criteria of a paradigm for enhancing images for critical applications comprise: interpreting images skeptically; approaching enhancement as a system problem; considering all image structures as potential information; and making uncertainty and alternative interpretations explicit, both visually and numerically.
Supplementary information:
The online version contains supplementary material available at 10.1007/s10032-021-00386-0.
... Because noise -whether biological, mechanical, electrical, or digital -is a fundamental issue in communication, substantial research has aimed at improving document legibility by suppressing a great variety of what are usually considered manifestations of noise, including: bleed-through [32], foxing [102], shadows [59], termite bites [94], crossouts [20], photocopied low-contrast carbon copies [26], low resolution raster images [87], and background texture interference [82] (for a history of image denoising, see [70]). The type of visual media can also dictate the typology of enhancement methods (e.g., methods for pictures and for movies differ in whether the time dimension is available as a source of contextual information for optimizing image processing) [16]. ...
Purpose: This article develops theoretical, algorithmic, perceptual, and interaction aspects of script legibility enhancement in the visible light spectrum for the purpose of scholarly editing of papyri texts. - Methods: Novel legibility enhancement algorithms based on color processing and visual illusions are proposed and compared to classic methods. A user experience experiment was carried out to evaluate the solutions and better understand the problem on an empirical basis. - Results: (1) The proposed methods outperformed the comparison methods. (2) The methods that most successfully enhanced script legibility were those that leverage human perception. (3) Users exhibited a broad behavioral spectrum of text-deciphering strategies, under the influence of factors such as personality and social conditioning, tasks and application domains, expertise level and image quality, and affordances of software, hardware, and interfaces. No single method satisfied all factor configurations. Therefore, using synergetically a range of enhancement methods and interaction modalities is suggested for optimal results and user satisfaction. (4) A paradigm of legibility enhancement for critical applications is outlined, comprising the following criteria: interpreting images skeptically; approaching enhancement as a system problem; considering all image structures as potential information; deriving interpretations from connections across distinct spatial locations; and making uncertainty and alternative interpretations explicit, both visually and numerically.
... In [10] the assessment of the image quality of the document is the weighted sum of the image clarity assessment and the font parameters assessment. The latter, in turn, is the sum of three estimates: the number of dark specks around the text, imitating the speckle structure, the estimate of the inter-letter space and the estimate of the size of the inter-letter space to the total size of the letter, which were proposed in [11] and adapted by the authors of the article for their own document format. The second group of methods includes the work [12], which proposes a method for calculating the image quality assessment based on calculating the maps of the mean square deviations of the brightness gradient calculated on the text areas of the image. ...
During the process of document recognition in a video stream using a mobile device camera, the image quality of the document varies greatly from frame to frame. Sometimes recognition system is required not only to recognize all the specified attributes of the document, but also to select final document image of the best quality. This is necessary, for example, for archiving or providing various services; in some countries it can be required by law. In this case, recognition system needs to assess the quality of frames in the video stream and choose the “best” frame. In this paper we considered the solution to such a problem where the “best” frame means the presence of all specified attributes in a readable form in the document image. The method was set up on a private dataset, and then tested on documents from the open MIDV-2019 dataset. A practically applicable result was obtained for use in recognition systems.
... Similarly, Souza et al. proposed to employ the statistics of connected components and the font size to measure the document quality, which were utilized to guide the filter selection to improve the document image's visual quality [26]. In order to assess the quality of typewritten document images and restore them, Cannon et al. used similar features, such as background speckle, touching characters, and broken characters to measure the quality of the documents [6]. By using stroke's gradient and average height-width ratio, Peng et al. estimated the document image's quality based on the Support Vector Regression (SVR), which was trained depended on normalized OCR scores [22]. ...
Document image quality assessment (DIQA), which predicts the visual quality of the document images, can not only be applied to estimate document’s optical character recognition (OCR) performance prior to any actual recognition, but also provides immediate feedback on whether the documents meet the quality requirements for other high level document processing and analysis tasks. In this work, we present a deep neural network (DNN) to accomplish the DIQA task, where a Saimese based deep convolutional neural network (DCNN) is employed with customized losses to improve system’s capability of linearity and monotonicity to predict the quality of document images. Based on the proposed network along with the new losses, the obtained DCNN achieves the state-of-the-art quality assessment performance on the public datasets. The source codes and pre-trained models are available at https://gitlab.com/xujun.peng/DIQA-linearity-monotonicity.
... Similarly, Souza et al. proposed to employ the statistics of connected components and the font size to measure the document quality, which were utilized to guide the filter selection to improve the document image's visual quality [26]. In order to assess the quality of typewritten document images and restore them, Cannon et al. used similar features, such as background speckle, touching characters, and broken characters to measure the quality of the documents [6]. By using stroke's gradient and average height-width ratio, Peng et al. estimated the document image's quality based on the Support Vector Regression (SVR), which was trained depended on normalized OCR scores [22]. ...
... The majority of the datasets that have been introduced are either not available or not available to the public [18][19][20][21]. The very few quality metrics for DIQA are either not available or not available to the public [18,[20][21][22][23][24]. The authors of [25] surveyed DIQA/VDIQA metrics and datasets. ...
The huge amount of degraded documents stored in libraries and archives around the world needs automatic procedures of enhancement, classification, transliteration, etc. While high-quality images of these documents are in general easy to be captured, the amount of damage these documents contain before imaging is unknown. It is highly desirable to measure the severity of degradation that each document image contains. The degradation assessment can be used in tuning parameters of processing algorithms, selecting the proper algorithm, finding damaged or exceptional documents, among other applications. In this paper, the first dataset of degraded document images along with the human opinion scores for each document image is introduced in order to evaluate the image quality assessment metrics on historical document images. In this research, human judgments on the overall quality of the document image are used instead of the previously used OCR performance. Also, we propose an objective no reference quality metric based on the statistics of the mean subtracted contrast normalized (MSCN) coefficients computed from segmented layers of each document image. The segmentation into four layers of foreground and background is done on the basis of an analysis of the log-Gabor filters. This segmentation is based on the assumption that the sensitivity of the human visual system (HVS) is different at the locations of text and non-text. Experimental results show that the proposed metric has comparable or better performance than the state-of-the-art metrics, while it has a moderate complexity. The developed dataset as well as the Matlab source code of the proposed metric is available at http://www.synchromedia.ca/system/files/VDIQA.zip.
... Earlier DOC-IQA methods used image based features such as font size, stroke thickness, gradient of edges, connected components statistics, touching/broken character factor, small speckle factor and other hand-crafted features [10], [11], [12], [13]. However, computation of these factors relies on heuristics determined for an image or a set of similar images. ...
Performance of most of the recognition engines for document images is effected by quality of the image being processed and the selection of parameter values for the pre-processing algorithm. Usually the choice of such parameters is done empirically. In this paper, we propose a novel framework for automatic selection of optimal parameters for pre-processing algorithm by estimating the quality of the document image. Recognition accuracy can be used as a metric for document quality assessment. We learn filters that capture the script
properties and degradation to predict recognition accuracy. An EM base d framework has been formulated to iteratively learn optimal parameters for document image pre-processing. In the E-step, we estimate the expected accuracy using the current set of parameters and filters. In the M-step we compute parameters to
maximize the expected recognition accuracy found in E-step. The experiments validate the efficacy of the proposed methodology for document image pre-processing applications.
... In this paper, we have developed such a monitoring system to predict the OCR results. Many research have been done in similar direction to asses the quality of the document regarding OCR [2], [3], [4], [5], but not to predict the OCR results. We assume that the performance of OCR are varying mainly due to the variability of font, noise, image quality and typographic problem that have a direct impact on text recognition. ...
In this paper, we describe a novel and simple technique for prediction of OCR results without using any OCR. The technique uses a bag of allographs to characterize textual components. Then a support vector regression (SVR) technique is used to build a predictor based on the bag of allographs. The performance of the system is evaluated on a corpus of historical documents. The proposed technique produces correct prediction of OCR results on training and test documents within the range of standard deviation of 4.18% and 6.54% respectively. The proposed system has been designed as a tool to assist selection of corpora in libraries and specify the typical performance that can be expected on the selection.
... These values are then used as inputs for different types of semisupervised classification algorithms. The authors of [5] use the features of [3] with three new ones from [6] to select a restoration algorithm using a linear classifier. ...
This article proposes an approach to predict the result of binarization algorithms on a given document image according to its state of degradation. Indeed, historical documents suffer from different types of degradation which result in binarization errors. We intend to characterize the degradation of a document image by using different features based on the intensity, quantity and location of the degradation. These features allow us to build prediction models of binarization algorithms that are very accurate according to R2 values and p-values. The prediction models are used to select the best binarization algorithm for a given document image. Obviously, this image-by-image strategy improves the binarization of the entire dataset.
... From the literature survey, few algorithms have been investigated on recognize the broken character. Cannon et al [14,15] proposed a method for automatically improving the quality of degraded image in a typewritten archive. Oguro et al [16] describe a method of three steps solutions for restoring faxed document by producing gray level image. ...
In this paper we focus on the different kinds of degradation in printed Bangla script. The working module of any Optical character Recognition system almost depends upon printing and paper of the input document image. A number of OCR techniques are available and claim correctly identified accuracy in printed document image in Indian and foreign script. A few report have been found on the recognition of the degraded Indian language document. The degradation in any scanned printed document can be of many types. In this paper, we have proposed different kind of degradation problem available in scanned printed Bangla script. We are identifying the different kind of degradation in printed Bangla language document image. Accordingly we have discussed problem associated with each kind of degradation in printed Bangla script document. The some possible solutions have also been discusses.
... This approach has been implemented in a system called ImageRefiner . Our work was originally inspired by the QUARC system [7] developed at Los Alamos National Labs and related work [24]. As any system based on machine learning, ImageRefiner operates in two modes: training and application (image refining , in our case). ...
We approach the analysis of electronic documents as a multi-stage process, which we implement via a multi-filter document processing framework that provides (a) flexibility for research prototyping, (b) efficiency for development, and (c) reliabil-ity for deployment. In the context of this framework, we present our multi-stage so-lutions to multi-engine Arabic OCR (MEMOE) and Arabic handwriting recognition (AHWR). We also describe our adaptive pre-OCR document image clean-up system called ImageRefiner. Experimental results are reported for all mentioned systems.
... Such measurements could be incorporated into summarization algorithms for the purpose of avoiding problematic regions, thereby improving the overall readability of the summary. Past work on attempting to quantify document image quality for predicting OCR accuracy [2,3,9] addresses a related problem, but one which exhibits some significant differences. ...
We investigate the problem of summarizing text documents that contain errors as a result of optical character recognition. Each stage in the process is tested, the error effects analyzed, and possible solu-tions suggested. Our experimental results show that current approaches, which are developed to deal with clean text, suffer significant degradation even with slight increases in the noise level of a document. We conclude by proposing possible ways of improving the performance of noisy document summarization.
... There are multiple sources of noise in the images we analysed, the two principal ones being the speckle created by the printing process and those inherent to microfilming. We dealt with speckle in much the same way as Cannon, et al. described [1]. ...
Despite the current practice of re-keying most documents placed in digital libraries, we continue to try to improve accuracy of automated recognition techniques for obtaining document image content. This task is made more dif-ficult when the document in question has been rendered in letterpress, subjected to hundreds of years of the ageing process and been microfilmed before scanning. In supporting the accurate capture of textual information, we endeavoured to leave intact a previously described document reconstruction technique, where we combine information from a language model and character image pattern matching to iteratively reduce ambiguity in document images. Combining word shape information and lists of similar bitmap patterns in a document at least partially resolves the character content without optical char-acter recognition.
... However, due to the unconstrained nature of handwriting, the above-mentioned degradation models are not directly applicable to handwriting . To evaluate the quality of typewritten document images and automatically select an optimal restoration method, [31] used five quality measures to assess the extent of background speckle noise, and touching and broken characters. For unconstrained handwriting, the 'font size' information is not available. ...
The automation of business form processing is attracting intensive research interests due to its wide application and its
reduction of the heavy workload due to manual processing. Preparing clean and clear images for the recognition engines is
often taken for granted as a trivial task that requires little attention. In reality, handwritten data usually touch or cross
the preprinted form frames and texts, creating tremendous problems for the recognition engines. In this paper, we contribute
answers to two questions: “Why do we need cleaning and enhancement procedures in form processing systems?” and “How can we
clean and enhance the hand-filled items with easy implementation and high processing speed?” Here, we propose a generic system
including only cleaning and enhancing phases. In the cleaning phase, the system registers a template to the input form by
aligning corresponding landmarks. A unified morphological scheme is proposed to remove the form frames and restore the broken
handwriting from gray or binary images. When the handwriting is found touching or crossing preprinted texts, morphological
operations based on statistical features are used to clean it. In applications where a black-and-white scanning mode is adopted,
handwriting may contain broken or hollow strokes due to improper thresholding parameters. Therefore, we have designed a module
to enhance the image quality based on morphological operations. Subjective and objective evaluations have been studied to
show the effectiveness of the proposed procedures.
... [ Beyondgeneral-purpose image classification algorithms no scientific investigation of the evaluation of overall scan quality analysis is known to the authors. For typewritten documents Cannon et al. [10] describe how to measure the small speckle factor (amount of black background speckle), white speckle factor (fattened character strokes), touching character factor (degree to which neighboring characters touch), broken character factor (degree to which individual characters are broken) and the font size factor (degradations that accompany an increase or decrease in the size of the font). Holley [11] gives some insights into analyzing and improving OCR accuracy in a large-scale historic newspaper digitalization project. ...
This paper describes how to make use of e-books that look like printed books in a knowledge network. After an overview of digitalization efforts and current digital library initiatives we introduce quality measures for the digitalization process. After digitalization an Interactive Internet Book (IIB) has to offer a kind of digital binding, annotation efforts and sophisticated ways for user interaction. We claim that the quality and the enhancements of an Interactive Internet Book go far beyond what is traditionally assumed: it is not enough to scan books. The scans have to be of high quality, allow good OCR to permit full text searches; books need not only be "packaged" but also need meta-data and functionalities that one can expect from a computer supported medium that go far beyond what is possible with traditional printed books. Those factors are critical for the use of e-books in social media environments, yet this is often still overlooked. Finally, we describe a working prototype and demonstrate the advantages obtained with a use case.
... Some progress has been made in recent years in the segmentation and restoration of degraded historical documents. Cannon et al. 7 introduced a measure for quantifying document image quality to predict OCR accuracy for a typewritten document archive. Based on the information about the quality of the image, they train a linear classifier that will predict the best restoration method. ...
In previous work we showed that Look Up Table (LUT) classifiers can be trained to learn patterns of degradation and correction in historical document images. The effectiveness of the classifiers is directly proportional to the size of the pixel neighborhood it considers. However, the computational cost increases almost exponentially with the neighborhood size. In this paper, we propose a novel algorithm that encodes the neighborhood information efficiently using a shape descriptor. Using shape descriptor features, we are able to characterize the pixel neighborhood of document images with much fewer bits and so obtain an efficient system with significantly reduced computational cost. Experimental results demonstrate the effectiveness and efficiency of the proposed approach.
... Such measurements could be incorporated into text processing algorithms for the purpose of avoiding problematic regions, thereby improving the overall readability . Past work on attempting to quantify document image quality for predicting OCR accuracy [1, 2, 6] addresses a related problem, but one which exhibits some significant differences . One possibility would be to establish a robust index that measures whether a given section of text is processable. ...
We investigate the problem of evaluating the performance of text processing algorithms on inputs that contain errors as a result of optical character recognition. A new hierarchical paradigm is proposed based on approximate string matching, allowing each stage in the processing pipeline to be tested, the error effects analyzed, and possible solutions suggested.
The quality of the image is critical to Optical Character Recognition (OCR), poor quality images will lead OCR to generate unreliable results. There are relative high ratio of low quality images in practical OCR-based application scenarios, how to evaluate quality of image and filter out unqualified images by document image quality assessment (DIQA) algorithms effectively is a big challenge for these scenarios. Current DIQA algorithms mainly focus on the overall image features rather than the text region, while the quality of the text region is dominant factor for OCR. In this paper, we propose a document image quality assessment algorithm based on information entropy in text region of image. Our algorithmic framework mainly consists of three networks to detect, extract and evaluate text region in image respectively. We build a quality prediction network based on HyperNet, and use the information entropy of the text region as the score weight, so that the final score can reflect the quality of the text region better. Finally, testing results on benchmark dataset SmartDoc-QA and our constructed dataset DocImage1k demonstrate that the proposed algorithm achieves excellent performance.KeywordsDIQAOCRHyperNetInformation entropy
We introduce a novel two-stage system for document image quality assessment (DIQA). The first-stage model of our system was trained on synthetic data to explicitly extract blur and text size features. The second-stage model was trained to assess the quality of optical character recognition (OCR) based on the extracted features. The proposed system was tested on two publicly available datasets: SmartDoc-QA and SOC. The discrepancies in the results between our system and current state-of-the art methods are within statistical error. At the same time, our results are balanced for both datasets in terms of Pearson and Spearman Correlation Coefficients. In the proposed approach, features are extracted from image patches taken at different scales, thus making the system more stable and tolerant of variations in text size. Additionally, our approach results in a flexible and scalable solution that allows a trade-off between accuracy and speed. The source code is publicly available on github: https://github.com/RodinDmitry/QA-Two-Step.
In this paper, we propose a method to segment seals and evaluate their quality. Seals with inferior qualities are not suitable for verification. To enhance the robustness of seal system, we put forward a strategy to assess the quality of extracted seal images. First we propose a method to segment seals and get the characters. Then by human assessment, we assign different characters with proper scores as ground-truth. We utilize a series of features and the SVR regression to predict the quality. Finally we use Optical Character Recognition rates to test the effectiveness. Experimental results prove that our proposed method is very effective.
With the proliferation of cameras on mobile devices there is an increased desire to image document pages as an alternative to scanning. However, the quality of captured document images is often lower than its scanned equivalent due to hardware limitations and stability issues. In this context, automatic assessment of the quality of captured images is useful for many applications. Although there has been a lot of work on developing computational methods and creating standard datasets for natural scene image quality assessment, until recently quality estimation of camera captured document images has not been given much attention. One traditional quality indicator for document images is the Optical Character Recognition (OCR) accuracy. In this work, we present a dataset of camera captured document images containing varying levels of focal-blur introduced manually during capture. For each image we obtained the character level OCR accuracy. Our dataset can be used to evaluate methods for predicting OCR quality of captured documents as well as enhancements. In order to make the dataset publicly and freely available, originals from two existing datasets - University of Washington dataset and Tobacco Database were selected. We present a case study with three recent methods for predicting the OCR quality of images on our dataset.
This paper proposes a deep learning approach for document image quality assessment. Given a noise corrupted document image, we estimate its quality score as a prediction of OCR accuracy. First the document image is divided into patches and non-informative patches are sifted out using Otsu's binarization technique. Second, quality scores are obtained for all selected patches using a Convolutional Neural Network (CNN), and the patch scores are averaged over the image to obtain the document score. The proposed CNN contains two layers of convolution, location blind max-min pooling, and Rectified Linear Units in the fully connected layers. Experiments on two document quality datasets show our method achieved the state of the art performance.
No-reference image quality assessment (NR-IQA) aims at computing an image quality score that best correlates with either human perceived image quality or an objective quality measure, without any prior knowledge of reference images. Although learning-based NR-IQA methods have achieved the best state-of-the-art results so far, those methods perform well only on the datasets on which they were trained. The datasets usually contain homogeneous documents, whereas in reality, document images come from different sources. It is unrealistic to collect training samples of images from every possible capturing device and every document type. Hence, we argue that a metric-based IQA method is more suitable for heterogeneous documents. We propose a NR-IQA method with the objective quality measure of OCR accuracy. The method combines distortion-specific quality metrics. The final quality score is calculated taking into account the proportions of, and the dependency among different distortions. Experimental results show that the method achieves competitive results with learning-based NR-IQA methods on standard datasets, and performs better on heterogeneous documents.
A summary of materials used in creating documents, methods of creating the printed document, and methods to acquire a digital version of that document are presented. Current as well as historical methods, materials, and processes are presented. Along with this, a discussion of places where image degradations can enter the process is included. All this is related to how these aspects could affect document recognition ability.
This article proposes an approach to predict the result of binarization algorithms on a given document image according to its state of degradation. Indeed, historical documents suffer from different types of degradation which result in binarization errors. We intend to characterize the degradation of a document image by using different features based on the intensity, quantity and location of the degradation. These features allow us to build prediction models of binarization algorithms that are very accurate according to values and p values. The prediction models are used to select the best binarization algorithm for a given document image. Obviously, this image-by-image strategy improves the binarization of the entire dataset.
This paper describes new capabilities of ImageRefiner, an automatic image enhancement system based on ma-chine learning (ML). ImageRefiner was initially designed as a pre-OCR cleanup filter for bitonal (black-and-white) document images. Using a single neural network, ImageRefiner learned which image enhancement transforma-tions (filters) were best suited for a given document image and a given OCR engine, based on various image measurements (characteristics). The new release improves ImageRefiner in three major ways. First, to process grayscale document images, we have included three grayscale filters based on smart thresholding and noise fil-tering, as well as five image characteristics that are all byproducts of various thresholding techniques. Second, we have implemented additional ML algorithms, including a neural network ensemble and several "all-pairs" classifiers. Third, we have introduced a measure that evaluates overall performance of the system in terms of cumulative improvement of OCR accuracy. Our experiments indicate that OCR accuracy on enhanced grayscale images is higher than that of both the original grayscale images and the corresponding bitonal images obtained by scanning the same documents. We have noticed that the system's performance may suffer when document characteristics are correlated.
Purpose
The purpose of this paper is to present a novel image sensor technology for raised characters based on line structured‐light. It can convert raised character's three‐dimensional (3D) features into image's grayscale levels.
Design/methodology/approach
The measurement principle and mathematical model are described. An experimental device is established and system parameters are calibrated. A grayscale conversion algorithm is proposed to convert the distortion of laser stripe to the grayscale intensity of image. The article also introduces a four‐factor method to assess the image quality of characters.
Findings
Experimental results show that the method can get high‐contrast images of raised characters that are conventionally low‐contrast with the background. Besides, the method does not need complicated calibration and mass computation, which makes the system structure simple and increases the speed of image acquisition.
Originality/value
The paper presents a novel image acquisition method for raised characters.
Preparing clean and clear images for the recognition engines is often taken for granted as a trivial task that requires little attention. Most of the existing OCRs have been designed in such a way that which correctly identify fine printed documents in all scripts. The performance of standard machine printed OCR system works fails, if it is tested on documents with distorted characters. This paper presents an approach to overcome the difficulties presented in such distorted type written documents especially with broken characters. As a first step, isolation of character is forwarded using character position location and character localization and enclosing it in a matrix which will be analyzing and repairing in the later part of our study. An attempt is incorporated using shape and line tracing method for recognition of distorted broken characters and then it is fine tuned by lexical knowledge.
To maintain, control and enhance the quality of document images and minimize the negative impact of degradations on various analysis and processing systems, it is critical to understand the types and sources of degradations and develop reliable methods for estimating the levels of degradations. This paper provides a brief survey of research on the topic of document image quality assessment. We first present a detailed analysis of the types and sources of document degradations. We then review techniques for document image degradation modeling. Finally, we discuss objective measures and subjective experiments that are used to characterize document image quality.
Large degradations in document images impede their readability and deteriorate the performance of automated document processing systems. Document image quality (IQ) metrics have been defined through optical character recognition (OCR) accuracy. Such metrics, however, do not always correlate with human perception of IQ. When enhancing document images with the goal of improving readability, e.g., in historical documents where OCR performance is low and/or where it is necessary to preserve the original context, it is important to understand human perception of quality. The goal of this paper is to design a system that enables the learning and estimation of human perception of document IQ. Such a metric can be used to compare existing document enhancement methods and guide automated document enhancement. Moreover, the proposed methodology is designed as a general framework that can be applied in a wide range of applications.
This paper describes document processing techniques used in ImageRefiner, the automatic image enhancement system developed by CACI International Inc. Though other methods are used in the system, we discuss two techniques that are novel and well tested or particularly important in the system. The first is a novel segmentation method that segments the text image file into "homogeneous" segments. The second is the use of a neural network to select the best transformation for each segment. Our experiments show that after applying the transformation selected by the neural network method to each specific segment, the fully processed images usually have more accurate OCR output. On average, the OCR accuracy for processed images is 35% better than the original images for a test set of Arabic files.
This paper describes a new automatic color thresholding based on wavelet denoising and color clustering with K-means in order to segment text information in a camera-based image. Several parameters bring different information and this paper tries to explain how to use this complementarity. It is mainly based on the discrimination between two kinds of backgrounds: clean or complex. On one hand, this separation is useful to apply a particular algorithm on each of these cases and on the other hand to decrease the computation time for clean cases for which a faster method could be considered. Finally, several experiments were done to discuss results and to conclude that the use of a discrimination between kinds of backgrounds gives better results in terms of Precision and Recall.
The massive digitization of heritage documents has raised new prospects for research like degraded document image restoration.
Degradations harm the legibility of the digitized documents and limit their processing. As a solution, we propose to tackle
the problem of degraded text characters with PDE (partial differential equation)-based approaches. Existing PDE approaches
do not preserve singularities and edge continuities while smoothing. Hence, we propose a new anisotropic diffusion by adding
new constraints to the Weickert coherence-enhancing diffusion filter in order to control the diffusion process and to eliminate
the inherent corner rounding. A qualitative improvement in the singularity preservation is thus achieved. Experiments conducted
on degraded document images illustrate the effectiveness of the proposed method compared with other anisotropic diffusion
approaches. We illustrate the performance with the study of the optical recognition accuracy rates.
KeywordsDegraded text characters–Document images–PDE-based approaches–Optical character recognition–Restoration–Reconstruction–Enhancement
Large degradations in document images impede their readability as well as substantially deteriorating the performance of automated document processing systems. Image quality metrics have been defined to correlate with OCR accuracy. However, this does not always correlate with human perception of image quality. When enhancing document images with the goal of improving readability, it is important to understand human perception of quality. The goal of this work is to evaluate human perception of degradation and correlate it to known degradation parameters and existing image quality metrics. The information captured enables the learning and estimation of human perception of document image quality.
In previous work we showed that shape descriptor features can be used in Look Up Table (LUT) classifiers to learn patterns of degradation and correction in historical document images. The algorithm encodes the pixel neighborhood information effectively using a variant of shape descriptor. However, the generation of the shape descriptor features was approached in a heuristic manner. In this work, we propose a system of learning the shape features from the training data set by using neural networks: Multilayer Perceptrons (MLP) for feature extraction. Given that the MLP maybe restricted by a limited dataset, we apply a feature selection algorithm to generalize, and thus improve, the feature set obtained from the MLP. We validate the effectiveness and efficiency of the proposed approach via experimental results.
Noise presents a serious challenge in optical character recognition, as well as in the downstream applications that make use of its outputs as inputs. In this paper, we describe a paradigm for measuring the impact of recognition errors on the stages of a standard text analysis pipeline: sentence boundary detection, tokenization, and part-of-speech tagging. Employing a hierarchical methodology based on approximate string matching for classifying errors, their cascading effects as they travel through the pipeline are isolated and analyzed. We present experimental results based on injecting single errors into a large corpus of test documents to study their varying impacts depending on the nature of the error and the character(s) involved. While most such errors are found to be localized, in the worst case some can have an amplifying effect that extends well beyond the site of the original error, thereby degrading the performance of the end-to-end system.
A key task for connectionist research is the development and analysis of learning algorithms. An examination is made of several supervised learning algorithms for single-cell and network models. The heart of these algorithms is the pocket algorithm, a modification of perceptron learning that makes perceptron learning well-behaved with nonseparable training data, even if the data are noisy and contradictory. Features of these algorithms include speed algorithms fast enough to handle large sets of training data; network scaling properties, i.e. network methods scale up almost as well as single-cell models when the number of inputs is increased; analytic tractability, i.e. upper bounds on classification error are derivable; online learning, i.e. some variants can learn continually, without referring to previous data; and winner-take-all groups or choice groups, i.e. algorithms can be adapted to select one out of a number of possible classifications. These learning algorithms are suitable for applications in machine learning, pattern recognition, and connectionist expert systems.
To improve the performance of an optical character recognition (OCR) system, an adaptive technique that restores touching or broken character images is proposed. By using the output from an OCR system and a distorted text image, this technique trains an adaptive restoration filter and then applies the filter to the distorted text image that the OCR system could not recognize. To demonstrate the performance of this technique, two synthesized images containing only touching characters and two synthesized images containing only broken characters were processed. The results show that this technique can improve both pixel and character accuracy of text images containing touching or broken characters
A generalized form of the cross‐validation criterion is applied to the choice and assessment of prediction using the data‐analytic concept of a prescription. The examples used to illustrate the application are drawn from the problem areas of univariate estimation, linear regression and analysis of variance.
This paper presents a methodology for model based restoration of degraded document imagery. The methodology has the advantages of being able to adapt to nonuniform page degradations and of being based on a model of image defects that is estimated directly from a set of calibrating degraded document images. Further, unlike other global filtering schemes, our methodology filters only words that have been misspelled by the OCR with a high probability. In the first stage of the process, we extract a training sample of candidate misspelled word subimages from the set of calibration images before and after the degradation that we wish to undo. These word subimages are registered to extract defect pixels. The second stage of our methodology uses a vector quantization based algorithm to construct a summary model of the defect pixels. The final stage of the algorithm uses the summary model to restore degraded document images. We evaluate the performance of the methodology for a variety of parameter settings on a real world sample of degraded FAX transmitted documents. The methodology eliminates up to 56.4% of the OCR character errors introduced as a result of FAX transmission for our sample experiment.
Determining the readability of documents is an important task. Human readability pertains to the scenario when a document image is ultimately presented to a human to read. Machine readability pertains to the scenario when the document is subjected to an OCR process. In either case, poor image quality might render a document un-readable. A document image which is human readable is often not machine readable. It is often advisable to filter out documents of poor image quality before sending them to either machine or human for reading. This paper is about the design of such a filter. We describe various factors which affect document image quality and the accuracy of predicting the extent of human and machine readability possible using metrics based on document image quality. We illustrate the interdependence of image quality measurement and enhancement by means of two applications that have been implemented: (1) reading handwritten addresses on mailpieces and (2) reading handwritten U.S. Census forms. We also illustrate the degradation of OCR performance as a function of image quality. On an experimental test set of 517 document images, the image quality metric (measuring fragmentation due to poor binarization) correctly predicted 90% of the time that certain documents were of poor quality (fragmented characters) and hence not machine readable.
Neural Network Learning and Expert Systems is the first book to present a unified and in-depth development of neural network learning algorithms and neural network expert systems. Especially suitable for students and researchers in computer science, engineering, and psychology, this text and reference provides a systematic development of neural network learning algorithms from a computational perspective, coupled with an extensive exploration of neural network expert systems which shows how the power of neural network learning can be harnessed to generate expert systems automatically.
Features include a comprehensive treatment of the standard learning algorithms (with many proofs), along with much original research on algorithms and expert systems. Additional chapters explore constructive algorithms, introduce computational learning theory, and focus on expert system applications to noisy and redundant problems.
For students there is a large collection of exercises, as well as a series of programming projects that lead to an extensive neural network software package. All of the neural network models examined can be implemented using standard programming languages on a microcomputer.
Bradford Books imprint
In many OCR systems, character segmentation is a necessary
preprocessing step for character recognition. It is a critical step
because incorrectly segmented characters are not likely to be correctly
recognized. The most difficult cases in character segmentation are
broken characters and touching characters. The problem of segmenting
touching characters in various fonts and size in machine-printed text is
addressed. The author classifies the touching characters into five
categories: touching characters in fixed-pitch fonts, proportional and
serif fonts, ambiguous touching characters, and strings with broken and
touching characters. Different methods for detecting multiple character
segments and for segmenting touching characters in these categories are
developed. The methods use features of characters and fonts and profile
models
An approach to supervised training of document-specific character templates from sample page images and unaligned transcriptions is presented. The template estimation problem is formulated as one of constrained maximum likelihoodparameter estimation within the document image decoding (DID) framework. This leads to a two-phase iterative training algorithm consisting of transcription alignment and aligned template estimation (ATE) steps. The ATE step is the heart of the algorithm and involves assigning template pixel colors to maximize likelihoodwhile satisifying a template disjointness constraint. The training algorithm is demonstrated on a variety of English documents, including newspaper columns, 15 th century books, degraded images of 19 th century newspapers and connected text in a script-like font. Three applications enabled by the training procedure are described--- high-accuracy document-specific decoding, transcription error visualization and printer font generation. 1. INTRODU...
Method for Repairing Optical Character Recognition Performing Different Repair Operations Based on Measured Image Characteristics
- Lougheed
Lougheed et. al, "Method for Repairing Optical Character Recognition Performing Different Repair
Operations Based on Measured Image Characteristics," United States Patent 5,142,589, 1992.
On the segmentation of touching characters Quality assessment and restoration of typewritten document images 89
- Yi Lu
- M Cannon
Yi Lu: On the segmentation of touching characters. Proc. ICDAR'93 2nd Int. Conf. on Document Analysis and Recognition, Tsukuba, Japan, 1993, p. 440 M. Cannon et al.: Quality assessment and restoration of typewritten document images 89
Enhancement and Restoration of Digital Documents
- P Robert
- Edward R Loce
- Dougherty
Robert P. Loce and Edward R. Dougherty, Enhancement and Restoration of Digital Documents, SPIE
Optical Engineering Press, 1997, p192.
10. ibid., p198.
ScanFix Software 206 West 6 th Avenue
- Tmssequoia
TMSSequoia, " ScanFix Software, " 206 West 6 th Avenue, Stillwater, OK, 74074 ©1997.
An Automated System for Numerically Rating Document Image Quality
- Michael Cannon
Michael Cannon et. al, "An Automated System for Numerically Rating Document Image Quality,"
Proceedings 1997 Symposium on Document Image Understanding Technology, Annapolis, MD,
p162.
Prediction of OCR Accuracy Using Simple Image Features
- R Luis
- Blando
Luis R. Blando, et. al, "Prediction of OCR Accuracy Using Simple Image Features," Proceedings,
ICDAR'95 Third International Conference on Document Analysis and Recognition, Montreal, 1995,
p319.
Document Image Restoration Using Binary Morphological Filters
- Jisheng Liang
Jisheng Liang et. al, "Document Image Restoration Using Binary Morphological Filters,"
Proceedings, Document Image Recognition III, SPIE '96, San Jose, CA, January 1996, p274.
Model Based Restoration of Document Images for OCR
- M Y Jaisimha
M. Y. Jaisimha et. al, "Model Based Restoration of Document Images for OCR," Proceedings,
Document Image Recognition III, SPIE '96, San Jose, CA, January 1996, p297.