Document Image Analysis

This research proposes the neural network (NN)-fuzzy logic control (FLC)-based methodology, which is designed with two stages of execution. Stage-1 is composed with the NN approach; it takes the input from the scanned image, the input from the classified clustered group, processes it with hidden layers, and generates the final output, which is input to the FLC in stage 2. The input of stage 2 converts the input into fuzzified data (0s and 1s) and trains the FLC system with fuzzified set values. The trained data has been simulated for further stages in the FLC-based decision-making approach and integrated to identify the character patterns based on the trained dataset. FLC-based planting and normalization techniques are integrated for retraining the data. The dataset has been framed with more than 6200 handwritten image characters, which were written by 235 writers. It aligned with the pre-processing, classification, and filtering stages and achieved a 99.8% recognition rate.

The carbonized Herculaneum scrolls represent a unique challenge for text recovery due to their fragile state and the visual similarity between ink and papyrus substrate. This study presents an iterative, human in-the-loop approach for ink detection in Scroll 5 (PHerc. 172) using a TimeSformer-based deep learning model. Building upon preprocessed X-ray Phase-Contrast Tomography (XPCT) data from the Vesuvius Challenge, we employ progressive model refinement and expert-verified labeling to improve the identification of ancient Greek letterforms. Our approach demonstrates that precise, trace-based annotation combined with repeated training cycles leads to a significant increase in legible character recognition from 169 to 368 letters across 15 scroll segments. The results underscore the value of interdisciplinary collaboration and iterative feedback in advancing the digital decipherment of ancient texts.

Given the existence of digital scanners, printers and fax machines, documents can undergo a history of sequential reproductions. One of the most important determiners of the quality of the resulting image is the set of underlying resolutions at which the images were scanned and binarized. In particular, a low resolution scan produces a noticeable degradation of image quality, and produces a set of printed fonts that cause omnifont OCR systems to operate with a relatively high error rate. This error rate can be reduced if the OCR system is trained on text having fax scanner degradations, but this also requires that the OCR system can determine in advance if such degraded fonts are present. Methods are found for determining the lowest resolution that a binary document image has been scanned (or printed) in the past. Observing that the signature for a low resolution scan is contained in the boundary contours of the black components, the connected components that constitute the contours are measured, and both the size histogram and its power spectrum are used to determine the underlying low resolution, if any. The primary application is to fax, both standard and fine. Degradation of both font appearance and OCR performance is far more severe for standard than fine fax, so the most important practical problem is to identify documents of binary scanned text that have a standard fax signature. Only a small part of the page image is required to determine the existence of such a signature; consequently, the image is pre-processed to locate a subregion that is well suited for analysis. Results for discriminating between originals, standard fax and fine fax on a small but diverse test suite are given.

A class of shift-variant reduction operations is introduced, that is useful for performing efficient and controllable shape and texture transformations between resolution levels. In their most general form, the operations proceed in three steps: (a) convolve a binary image with a kernel of arbitrary size; (b) threshold the result; (c) subsample to produce the reduced image. Taking a binary structuring element for the kernel, the threshold convolution on a binary image is equivalent to a rank order filter, and the full reduction operation is a threshold reduction. Threshold reductions that use convolution filters and subsample tiles of equal size are optimized by combining the three operations, using only logical raster operations and producing threshold convolution values only at the sampling points. For 2x reduction, the four possible threshold values (1, 2, 3, and 4) refer to the minimum number of ON pixels within each 2x2 tile for which a pixel in the reduced image will be ON. Algorithms for boolean raster operations are given for 2x, 3x, and 4x threshold reduction, and lookup tables that efficiently implement column raster operations are provided. Threshold reduction rates of 2.5x½¼ pixel/second can be achieved with a Sun SparcStation2 Ì Å . A mask-forming image analysis cycle of threshold reduction, augmented by morphology and followed by replicative expansion to full resolution, is described, and some general properties of the cycle are derived. A simple application of threshold reduction to document image analysis, the extraction of halftone regions from scanned images that also contain text and line graphics, is illustrated. A sequence of 2x reductions with first low and then high thresholds is used to create a reduced image consisting of a mask over the halftone regions. In this way, the extraction occurs as a natural consequence of the reductions.

A particularly effective method for analyzing document images, that consist of large numbers of binary pixels, is to generate reduced images whose pixels represent enhancements of textural densities in the full-resolution image. These reduced images are generated using an integrated combination of filtering and subsampling. Previously reported methods used thresholding over a square grid, and cascaded these threshold reduction operations. Here, the approach is generalized to a sequence of arbitrary filtering/subsample operations, with emphasis on several particular filtering operations that respond to salient textural qualities of document images, such as halftones, lines or blocks of text, and horizontal or vertical rules. As with threshold reductions, these generalized "textured reductions" are performed with no regard for connected components. Consequently, the results are typically robust to noise processes that can vitiate analysis based on connected components. Examples of image analysis and segmentation operations using textured reductions are given. Some properties can be determined very quickly; for example, the existence or absence of halftone regions in a full page image can be established in about 10 milliseconds.

A vital part of the publication process of ancient cuneiform tablets is creating hand-copies, which are 2D line art representations of the 3D cuneiform clay tablets, created manually by scholars. This research provides an innovative method using Convolutional Neural Networks (CNNs) to identify strokes, the constituent parts of cuneiform characters, and display them as vectors — semi-automatically creating cuneiform hand-copies. This is a major step in optical character recognition (OCR) for cuneiform texts, which would contribute significantly to their digitization and create efficient tools for dealing with the unique challenges of Mesopotamian cultural heritage. Our research has resulted in the successful identification of horizontal strokes in 2D images of cuneiform tablets, some of them from very different periods, separated by hundreds of years from each other. With the Detecto algorithm, we achieved an F-measure of 81.7% and an accuracy of 90.5%. The data and code of the project are available on GitHub. An innovative approach towards cuneiform OCR of identifying strokes instead of signs offers insight into the challenges and methodologies of quantitative epigraphy.

Today’s digital era, the attention towards camera based text processing has increased many folds. This has led to the development of multiple text processing methods. Most of the procedures follow a scene text detection manner and further improvement is needed in camera captured images. In this research paper we present a survey and analysis of the existing methods in text line detection for camera captured images. In this work we have identified text-line detection methods such as MSER with Ada boost, State Estimation, Texture Based SVM, Chain Level Classification, CMSER Algorithm, Graph Model, and Fuzzy Method. MSER with Ada boost is the better one, it is robust and efficient.

In this paper, we propose a new approach based on sparse coding for single textual image Super-Resolution (SR). The proposed approach is able to build more representative dictionaries learned from a large training Low-Resolution/High-Resolution (LR/HR) patch pair database. In fact, an intelligent clustering is employed to partition such database into several clusters from which multiple coupled LR/HR dictionaries are constructed. Based on the assumption that patches of the same cluster live in the same subspace, we exploit for each local LR patch its similarity to clusters in order to adaptively select the appropriate learned dictionary over that such patch can be well sparsely represented. The obtained sparse representation is hence applied to generate a local HR patch from the corresponding HR dictionary. Experiments on textual images show that the proposed approach outperforms its counterparts in visual fidelity as well as in numerical measures.

Demographic handwriting-based classification problems, such as gender and handedness categorizations, present interesting applications in disciplines like Forensic Biometrics. This work describes an experimental study on the suitability of deep neural networks to three automatic demographic problems: gender, handedness, and combined gender-and-handedness classifications, respectively. Our research was carried out on two public handwriting databases: the IAM dataset containing English texts and the KHATT one with Arabic texts. The considered problems present a high intrinsic difficulty when extracting specific relevant features for discriminating the involved subclasses. Our solution is based on convolutional neural networks since these models had proven better capabilities to extract good features when compared to hand-crafted ones. Our work also describes the first approach to the combined gender-and-handedness prediction, which has not been addressed before by other researchers. Moreover, the proposed solutions have been designed using a unique network configuration for the three considered demographic problems, which has the advantage of simplifying the design complexity and debugging of these deep architectures when handling related handwriting problems. Finally, the comparison of achieved results to those presented in related works revealed the best average accuracy in the gender classification problem for the considered datasets.

Demographic handwriting-based classification problems, such as gender and handedness categorizations, present interesting applications in disciplines like Forensic Biometrics. This work describes an experimental study on the suitability of deep neural networks to three automatic demographic problems: gender, handedness, and combined gender-and-handedness classifications, respectively. Our research was carried out on two public handwriting databases: the IAM dataset containing English texts and the KHATT one with Arabic texts. The considered problems present a high intrinsic difficulty when extracting specific relevant features for discriminating the involved subclasses. Our solution is based on convolutional neural networks since these models had proven better capabilities to extract good features when compared to hand-crafted ones. Our work also describes the first approach to the combined gender-and-handedness prediction, which has not been addressed before by other researchers. Moreover, the proposed solutions have been designed using a unique network configuration for the three considered demographic problems, which has the advantage of simplifying the design complexity and debugging of these deep architectures when handling related handwriting problems. Finally, the comparison of achieved results to those presented in related works revealed the best average accuracy in the gender classification problem for the considered datasets.

Demographic handwriting-based classification problems, such as gender and handedness categorizations, present interesting applications in disciplines like Forensic Biometrics. This work describes an experimental study on the suitability of deep neural networks to three automatic demographic problems: gender, handedness, and combined gender-and-handedness classifications, respectively. Our research was carried out on two public handwriting databases: the IAM dataset containing English texts and the KHATT one with Arabic texts. The considered problems present a high intrinsic difficulty when extracting specific relevant features for discriminating the involved subclasses. Our solution is based on convolutional neural networks since these models had proven better capabilities to extract good features when compared to hand-crafted ones. Our work also describes the first approach to the combined gender-and-handedness prediction, which has not been addressed before by other researchers. M...

Gérer les périodiques sous la direction de Géraldine Barron BAO #17 (2009) Favoriser la réussite des étudiants sous la direction de Carine El Bekri-Dinoird BAO #18 (2009) Mettre en oeuvre un plan de classement sous la direction de Bertrand Calenge) BAO #20 (2010) Mettre en oeuvre un service de questions-réponses en ligne sous la direction de Claire Nguyen BAO #21 (2010) Communiquer ! Les bibliothécaires, les décideurs et les journalistes sous la direction de Jean-Philippe Accart BAO #22 (2011) Mener l'enquête ! Guide des études de publics en bibliothèque sous la direction de Christophe Evans BAO #23 (2011) Créer des services innovants. Stratégies et répertoire d'actions pour les bibliothèques sous la direction de Marie-Christine Jacquinet 1. Les termes suivis d'un astérisque (à leur première occurrence) sont définis dans le glossaire en fin d'ouvrage. 1. Arjun Appadurai, Géographie de la colère. La violence à l'âge de la globalisation (2006), trad. française F. Bouillot. Paris, Payot, 2007. 1. Centrale technique d'État pour les bibliothèques publiques : < >.

The most dangerous and rapidly spreading disease in the world is Tuberculosis. In the investigating for suspected tuberculosis (TB), chest radiography is the only key techniques of diagnosis based on the medical imaging So, Computer aided diagnosis (CAD) has been popular and many researchers are interested in this research areas and different approaches have been proposed for the TB detection. Image segmentation plays a great importance in most medical imaging, by extracting the anatomical structures from images. There exist many image segmentation techniques in the literature, each of them having their own advantages and disadvantages. The aim of X-ray segmentation is to subdivide the image in different portions, so that it can help during the study the structure of the bone, for the detection of disorder. The goal of this paper is to review the most important image segmentation methods starting from a data base composed by real X-ray images.

In this paper we present a survey of the literature on Arabic writer identification scheme and up-to date techniques employed in identification. The paper begins with an overview of the various writer identification schemes in Arabic and Persian languages. After that, an attempt is made to describe the complex character of Arabic strokes. Previous studies have used a number of Arabic datasets comprised of handwritten text samples that have been used over the years aimed to support handwriting recognition. In automated writer identification, two methods of identification are available text independent and text dependent methods. These are the key focus of this paper. For evaluation purposes we have analyzed the Arabic writer identification scheme by comparing the recognition rate of different methods.

Arad Ostracon 16 is part of the Elyashiv Archive, dated to ca. 600 b.c. It was published as bearing an inscription on the recto only. New multispectral images of the ostracon have enabled us to reveal a hitherto invisible inscription on the verso, as well as additional letters, words, and complete lines on the recto. We present here the new images and offer our new reading and reinterpretation of the ostracon.

This article discusses the quality assessment of binary images. The customary, ground truth based methodology, used in the literature is shown to be problematic due to its subjective nature. Several previously suggested alternatives are surveyed and are also found to be inadequate in certain scenarios. A new approach, quantifying the adherence of a binarization to its document image is proposed and tested using six different measures of accuracy. The measures are evaluated experimentally based on datasets from DIBCO and H-DIBCO competitions, with respect to different kinds of binarization degradations.

In this paper we present a hybrid approach to segment and classify contents of document images. A Document Image is segmented into three types of regions: Graphics, Text and Space. The image of a document is subdivided into blocks and for each block five GLCM (Grey Level Co-occurrence Matrix) features are extracted. Based on these features, blocks are then clustered into three groups using K-Means algorithm; connected blocks that belong to the same group are merged. The classification of groups is done using pre-learned heuristic rules. Experiments were conducted on scanned newspapers and images from MediaTeam Document Database Dans cet article nous présentons un modèle hybride de segmentation et classification du contenu d&#39;images de documents. Chaque image est segmentée en trois types de régions: graphiques, textes et espaces. L&#39;image de chaque document est subdivisée en blocs; pour chacun de ces blocs, cinq caractéristiques sont extraites de leur Matrice de Co-occurrence d...

We describe a simple convolutional network for blind unmixing of transient absorption microscopy data along with a model ensembling strategy. Our network is based on an autoencoder previously developed for blind unmixing of hyperspectral satellite images. Its advantages are (a) that it learns to unmix spectra by unsupervised learning, i.e. by learning to reconstruct imaging data, without knowledge of the underlying spectra or their abundances and (b) that the endmember spectra are directly encoded by the output layer's coefficients. Extensive modifications to both the network architecture and training loss functions were necessary to produce reasonable performance on transient absorption data. We demonstrate results from blind unmixing of transient absorption images of unstained muscle fibers, acquired at 520nm pump and 620nm probe, training an ensemble of 500 different networks (i.e. unmixing models), each starting from a different random initialization. Variability among resulting models was analyzed by principal component analysis on the recovered endmembers from all models, deriving from the projections a model probability density function. We found consistent models (predicting similar endmembers and abundance maps) near the most likely model and surrounding highprobability region, with more variability in low-probability regions. Then, a permutation-aligned average of the ensemble produced much better results than an unweighted ensemble average, or simple selection of one model based on maximum likelihood or best fit. We anticipate this approach of parametrizing models and ensembling based on relative probability to have applications in other chemical imaging modalities such as FLIM, Raman microscopy and mass spectroscopic imaging.

In this paper, we present a method for discriminating handwritten and printed text from document images based on shape features. The separation of handwritten and printed text from document image is essential to optimize the OCR accuracy and to activate an appropriate OCR engine. It leads to reduce the search space of the OCR and it also facilitates the retrieval of Handwritten and Printed text from document images. We have used IAM dataset 3.0 and with morphological transformations segmented 74 pages and obtained 10768 words and 2000 were used for experimentation and achieved average accuracy of 98.57% with only seven features. The proposed method is simple, have promising discrimination accuracy and less time complexity as compared to [10].

Document identification is used to extract information from a digital document such as Al-Quran, articles, agreement and so on. With increasing digital documents on the internet, it is important to identify that the document is genuine or not. There is existing research on document identification. However, the problem occurs when character recognition is done for particular language only and it is hard to recognize the character when the image dataset size was small. Therefore, the purpose of this research is to make use the similarities of each character language which is unoccupied space as the document feature extraction using triangle model. As for the preliminary work, the objective is to obtain a list of point selection that will be used for triangle model from the generated histogram. This research using an experimental design, the dataset was chosen is own dataset which document image will be used and IFN/ENIT dataset in order to handle the small size dataset. While the tech...

Abstract: In an automatic document conversion system, which builds digital documents from scanned articles, there is a need to perform various adjustments before the scanned image is fed to the layout analysis system. This is because the layout detection system is sensitive to errors ...

Religious minorities, particularly Coptic Christians, represent a vital component of Egypt’s social and historical fabric. Copts are the indigenous inhabitants of Egypt, tracing their origins to pre-Arab conquest periods. They maintain distinct linguistic, religious, and cultural traditions, including the Coptic language, liturgical practices, and community festivals

We present a new approach for recognition of complex graphic symbols in technical documents. Graphic symbol recognition is a well known challenge in the field of document image analysis and is at heart of most graphic recognition systems. Our method uses structural approach for symbol representation and statistical classifier for symbol recognition. In our system we represent symbols by their graph based signatures: a graphic symbol is vectorized and is converted to an attributed relational graph, which is used for computing a feature vector for the symbol. This signature corresponds to geometry and topology of the symbol. We learn a Bayesian network to encode joint probability distribution of symbol signatures and use it in a supervised learning scenario for graphic symbol recognition. We have evaluated our method on synthetically deformed and degraded images of presegmented 2D architectural and electronic symbols from GREC databases and have obtained encouraging recognition rates.

Handwriting recognition and analysis has been an active area of research in the last two decades. Handwriting analysis is being studied in various fields of science, such as graphology, neurology, psychology, and computer science. Furthermore, automated handwriting analysis has several applications, including forensic, security, medical, and disease prediction. This paper presents the most recent handwriting analysis techniques and advancements available in the literature for age and gender classification/detection. Different steps, including feature extraction and classification, frequently used in the literature for age and gender detection, are discussed, and the presented works are classified according to the applied feature extraction and classification methods. The online and offline benchmark databases are also reviewed. We used a text mining technique to perform a quantitative content analysis of the presented research and better understand the cooccurrence network diagrams of age and gender classification/detection. This study is a valuable resource that provides new research directions to students and researchers interested in this field for further research and investigation.

Die künstliche Intelligenz schafft völlig neue Möglichkeiten in der Erzeugung von Bildern. Erstmals sind alle Bildgebungstechniken der Kunstgeschichte in einem Werkzeug zusammengeführt, ausführbar von jedem Computerbesitzer ohne besondere Fachkenntnisse. Erstellen lassen sich damit Bilder, die aussehen wie Fotografien, aber keine sind. Was bedeutet das für die Fotografie im Allgemeinen und die dokumentarische Fotografie im Besonderen? Eine Diskussion über diese Fragen ist überfällig, wenn man bedenkt, dass das Vertrauen in die Authentizität von Fotos fast 200 Jahre lang die Menschheitsgeschichte mitgeprägt hat.
Bernd Arnold, Autor und Fotograf von „Das Kölner Heil“ und „Wahl Kampf Ritual“, ist überzeugt, dass wir eine Neue Welt betreten werden, wenn authentische Fotos und KI-generierte Fotografie-Imitationen bei der Bildung unseres Welt- und Geschichtsbildes miteinander konkurrieren. In drei Essays analysiert er die Technik- und Wahrnehmungsgeschichte der Fotografie, erklärt das Neue und Innovative der KI und befasst sich mit der zukünftigen Rolle der dokumentarischen Fotografie und deren Produzenten.

„Das Vertrauen in die Authentizität
 einer dokumentarischen Fotografie
 ist fundamental für eine Demokratie“

„Es sind nicht die sensationellen Bilder, die das
 Vertrauen untergraben, sondern die Bilder, die wie 
alltäglich gewohnte Fotografien erscheinen, aber
Fotografie imitieren, Realität modulieren und in den 
gewohnten Kanälen im Laufe der Jahre zunehmend
 verbreitet werden.“

The World of New Images: Documentary photography and AI Artificial intelligence creates completely new possibilities in the creation of images. For the first time, all the imaging techniques of art history have been brought together in a single tool that can be used by any computer user without any special expertise. A discussion about these issues is overdue, considering that trust in the authenticity of photographs has helped shape human history for almost 200 years. Images can be created that look like photographs but are not. What does this mean for photography in general and documentary photography in particular?
They can be used to create images that look like photographs but are not. What does this mean for photography in general and documentary photography in particular? A discussion about these questions is overdue, considering that trust in the authenticity of photographs has helped shape human history for almost 200 years.
Bernd Arnold (author and photographer) is convinced that we will enter a New World when authentic photos and AI-generated photographic imitations compete with each other in forming our view of the world and history. In three essays, he analyzes the technological and perceptual history of photography, explains the new and innovative nature of AI and looks at the future role of documentary photography and its producers. For the phenomenon of photographic imitations, he introduces the term “dichography”, which describes a parallel universe of fictional reality that will soon be indistinguishable from the photographic light traces of pastreality.

By proper exploitation of the structural characteristics existing in a compressed document, it is possible to speed up certain image processing operations. Alternatively, one can derive a compression scheme which would lend itself to an efficient manipulation of documents without compromising the compression factor. Here, a run-based compression technique is discussed for binary documents. The technique, in addition to achieving bit rates comparable to other compression schemes, preserves document features which are useful for analysis and manipulation of data. Algorithms are proposed to perform vertical run extraction, and similar operations in the compressed domain. These algorithms are implemented in software. Experimental results indicate that fast analysis of electronic data is possible if data is coded according to the proposed scheme.

AbstractÐAn algorithm, which performs connected components detection in the course of decoding ITU-T (former CCITT) facsimile Group 3/4, i.e., MH/MR/MMR compressed images is presented. New definitions of mode color and a new transition element are introduced that allow MR/MMR codes to analyze and derive information about connection of black runs in two adjacent scan lines in the course of decoding. The experiments on the standard set of eight CCITT documents have shown that, on the average, the complexity of direct processing of MR/MMR codes is lower by a factor of 20 and 2.5 than that for raster images and MH codes processing respectively. Data structures for image vector description are discussed. Index TermsÐDocument image, CCITT (ITU) Group 3/4 compression, MH/MR/MMR, connected components. ae 1 INTRODUCTION . E. Regentova and S. Latifi are with the Department of Electrical and Computer Engineering,

Incunabulaare the texts printed mainly during the second half of 15th century that are a key cultural element in a revolutionary period of the history and evolution of the book and the printing. In these books, the identification of their origin largely affects its academic, cultural, patrimonial, and economical value. This paper proposes a process to automate the identification of the origin of a digitised incunable document using the Proctor/Haebler method, a commonly established procedure in the field. This process has been validated with a selected dataset obtained from the incunabula collection at the digital repository of the University of Zaragoza.

Printed documents continue to be the most commonly used media for information transfer in official context. However, such documents may be subject to illegitimate modification or malicious purposes. Therefore, agencies must be able to verify their authenticity before accepting them. This work aims to identify and classify the current computational methods that are used to identify and detect a forgery in printed documents. We focus on methods that detect forger passively, i.e., no requirement of active information to be included in the document during or shortly after printing. Hence, we classify the methods into four general categories. First, detection based on printer and ink properties. The second is based on character-level features analysis that is based on detecting imperfections produced by scan, edit, and print forgery techniques. It is also possible to detect forgery through analysis of paper properties. Finally, forgery detection can be achieved by analyzing the pixel-level features.

The rapid growth of digital data has led to the widespread creation and storage of digital images containing text. The extraction and use of textual information might be advantageous for various kinds of domains. Text detection in natural images is primarily affected by noise, blur, distortions, font variation, alignments, and orientation. Government forms, mark cards, medical records, business receipts, and other increasingly common bilingual documents have a substantial impact about precision of text detection and recognition. The paper, "Enhanced Machine Learning Methods for Robust Real-Time Classification of Bilingual Documents," emphasizes these challenges and suggests a solution that uses image enhancement techniques to improve an image's appearance and quality. Digital documents are bilingual; hence, extracting information from them is challenging since a computer can read and interpret text written in many scripts, including English and Kannada, within the same documentation. It thus extracts text from documents using optical character recognition (OCR)-based problems. Natural language processing (NLP) to classify text, super-resolution methods to synthesize a highresolution (HR) image from several lowresolution (LR) images and machine learning models EasyOcr to detect and recognize Text.