Search

Optical Character Recognition OCR is an essential part of many real-world applications such as digital archiving, automatic number plate recognition, handle cheques, etc. However, developing an OCR for printed Arabic text is still a challenging and open research field due to the special characteristics of Arabic cursive script. In this paper, we propose a segmentation-based, omnifont, open-vocabulary OCR for printed Arabic text. The proposed approach doesn’t require an explicit font type recognition stage. It uses an explicit, indirect character segmentation method. The presented segmentation method is baseline dependent and employs a hybrid, three-steps character segmentation algorithm to handle the problem of character overlapping. Besides, it uses a set of topological features that are designed and generalized to make the segmentation approach font independent. The segmented characters are fed as an input to a convolutional neural network for feature extraction and recognition. The APTID-MF data set has been used for testing and evaluation. The average accuracy of the proposed segmentation stage is 95%, while the average accuracy of the recognition stage is 99.97%. The whole approach achieves an average accuracy of 95% without using font-type recognition or any post-processing techniques.

Segmentation is a significant stage for the recognition of old newspapers. Text-line extraction in the documents like newspaper pages which have very complex layouts poses a significant challenge. Old newspaper documents printed in Gurumukhi script present several forms of hurdles in segmentation due to noise, degradation, bleed-through of ink, multiple font styles and sizes, little space between neighboring text lines, overlapping of lines, etc. Because of the low quality and the complexity of these documents, automatic text line segmentation remains an open research field. Very few researches are available in the literature to segment news articles in Gurumukhi script. This is one of the first few attempts to recognize Gurumukhi newspaper text. The goal of this paper is to present a new methodology for text-line extraction by integrating median calculation and strip height calculation techniques. Non-suitability of existing techniques to segment newspaper text lines have also been discussed with results in the article. The efficiency of the proposed algorithm is demonstrated by experimentation directed on two diverse own made datasets: (a) on the data set of single-column documents with headlines block (b) on the dataset of multi-column documents with headlines block.

The CAPTCHA technology can be used to ensure big multimedia data security, which includes CAPTCHA design and CAPTCHA recognition. For the existing methods are difficult to achieve high breaking accuracy for complex handwritten text CAPTCHA, a handwritten CAPTCHA recognizer is proposed, which is a text CAPTCHA breaking method based on style transfer network. Firstly, different from the traditional viewpoints that font structure and font style of characters are inseparable in this field, a new idea of separating font structure and font style of characters is proposed, and it is pointed out that character recognition mainly depends on font structure rather than font style. Secondly, based on this idea, a style transfer network for text CAPTCHA is constructed to convert complex and variable handwritten CAPTCHA into easy-to-recognize printed CAPTCHA. Finally, based on deep convolutional neural network, a text CAPTCHA recognition network is constructed to identify the converted printed CAPTCHAs. With CAPTCHAs from three real websites: eBay, Google and reCAPTCHA, experimental results show that the recognizer has higher breaking accuracy for handwritten CAPTCHA compared with the methods proposed in NDSS’16, CCS’18 and “Science” in 2017.

In linguistic steganography, languages’ features are employed to hide information. The Arabic language has a rich set of features which have not been utilized until now in this area. In particular, Arabic calligraphy contains multiple fonts and multiple shapes of Arabic alphabet letters. In this paper, a framework that uses Arabic calligraphy to hide information is proposed. The phases of the framework are preparation phase, embedding phase and extraction phase. The embedding phase uses string matching to generate stego text and accompanying letter shapes according to a secret message. The framework also includes corpus creation and a modification of the Aho-Corasick string-matching algorithm. The Arabic font Naskh was used as a case study. A set of Arabic poetry and proverbs were used as a dataset. The framework was evaluated on capacity and security. Because the visual difference between the cover and the stego-cover must be unnoticeable to the human in any stego-system, the security in this framework is satisfying due there is no cover used. The cover represents the secret message itself and it provides high capacity to hide data also. The evaluation showed the potential of using the multiple shapes of Arabic letters to satisfy steganography requirements.

While optical character recognition for Latin based scripts have seen near human quality performance, the accuracy for the rounded scripts of South Asia still lags behind. Work on Sinhala OCR has mainly reported on performance on constrained classes of font faces and so been inconclusive. This paper provides a comprehensive series of experiments using conventional machine learning as well as deep learning on texts and font faces of diverse types and in diverse resolutions, in order to present a realistic estimation of the complexity of recognizing the rounded script of Sinhala. While texts of both old and contemporary books can be recognized with over 87% accuracy, those in old newspapers are much harder to recognize owing to poor print quality and resolution.

In our research, we study the problem of font synthesis using an end-to-end conditional deep adversarial network with a small sample of Korean characters (Hangul). Hangul comprises of 11,172 characters and is composed by writing in multiple placement patterns. Traditionally, font design has required heavy-loaded human labor, easily taking one year to finish one style set. Even with the help of programmable approaches, it still takes a long time and cannot escape the limitations around the freedom to change parameters. Many trials have been attempted in deep neural network areas to generate characters without any human intervention. Our research focuses on an end-to-end deep learning model, the Skeleton-Driven Font generator (SKFont): when given 114 samples, the system automatically generates the rest of the characters in the same given font style. SKFont involves three steps: First, it generates complete target font characters by observing 114 target characters. Then, it extracts the skeletons (structures) of the synthesized characters obtained from the first step. This process drives the system to sustain the main structure of the characters throughout the whole generation processes. Finally, it transfers the style of the target font onto these learned structures. Our study resolves long overdue shortfalls such as blurriness, breaking, and a lack of delivery of delicate shapes and styles by using the ‘skeleton-driven’ conditional deep adversarial network. Qualitative and quantitative comparisons with the state-of-the-art methods demonstrate the superiority of the proposed SKFont method.

This study aims to develop an automatic transliteration application for the Balinese palm leaf manuscripts into the Latin/Roman alphabet. The input for this system is the digital image of the original text from the ancient Balinese palm leaf manuscripts, not from the Balinese script, which is printed using a font on a computer. In this study, a segmentation-free transliteration machine using the LSTM model was implemented. In addition, the implementation of the AKSALont application is carried out for the interactions on a web-based platform using cross-platform interoperability. The experimental results show that the machine can transliterate Balinese characters on the Balinese palm-leaf manuscript images properly with a CER of 19.78 % using 10.475 test data. With a web-based online platform, AKSALont has been able to open wider access for the public to the web-based content with an online platform collection.

The modern world is impossible to imagine without the exchange of information. Newspapers, magazines, books, printed advertising products play an important role and are constantly being improved and require more and more resources and opportunities. At the moment, due to the sharp growth of digital technology, there is a problem between the capabilities of hardware and software. That is, when transferring data from the electronic world to paper, the difference between the computing capabilities of hardware and software is displayed. This applies to print media that cannot convey the full volume and depth of what is displayed on the screen. Raster fonts are used for printing faster, the main drawback of which is their non-scalability. When studying fonts in embedded systems, there were problems with the fact that there are quite a few ready-made options for converting text to a bitmap without certain restrictions. The font displayed on the screen and printed directly by the printer or other device may differ significantly because the number of pixels that can be printed is too large. Therefore, there is a need to optimize the conversion of the font into a bitmap, ie reducing the number of points without losing the base of the glyph. The paper presents the conversion of a font into a bitmap using a program based on a ruler, ie the division of the image into lines, which is used in printing. In order to create the required bitmap image in the program, you need to specify a specific glyph and its size, as well as the device for which it will be used. The scan converter takes a path and applies a set of rules to determine which pixels will be part of the glyph image when printed or displayed on the screen. The obtained results allow us to print the fonts we like, instead of the standard ones.

This research proposed a method for the affixed word transliteration to the Balinese Script since there has not been studied yet and it is important since the affixed word needs to be transliterated, inevitably. This research is one of the efforts to preserve digitally the endangered Balinese local language knowledge in Indonesia through the multi-discipline collaboration between Computer Science and Language discipline. The proposed method was taken care of two related aspects, i.e.; (1) A Latin root word has its related Balinese Script root word by using default or special transliteration rule; and (2) A Latin root word with a special transliteration rule for its Balinese Script root word, also need a special transliteration rule for its affixed word. This study was conducted on the pioneering web-based transliteration learning application, BaliScript, that receives the Latin text input and outputs the Balinese Script by using the Noto Serif Balinese font with its dedicated Balinese Unicode. Through the experiment, the proposed method gave the expected transliteration results, added a certain perspective, and strengthened the transliteration knowledge. Future work is to enhance and reuse this method on the mobile computing device, as a part of the Balinese Language ubiquitous learning that supports Balinese Language education, which is a mandatory local subject from the elementary school to the high school in Bali Province.

The expiry dates printed on the merchandise have a distinct background, font, alignment, and color in comparison with the available handwritten digit datasets. In this paper, an expiry date dataset is used, and also a convolutional neural network (CNN) model is proposed to recognize expiry dates out of images. This model may be employed together with our previously proposed smart expiry architecture to get an automated notification to the smartphone for the foods which are expiring soon. The suggested deep learning model is tested and has a classification accuracy of 90%.

Is it possible for a robot to navigate an unknown area without the ability to understand verbal instructions? This work proposes the use of pictorial cues (hand drawn sketches) to assist navigation in scenarios where verbal instructions seem less practical. These scenarios include verbal instructions referring to novel objects or complex instructions describing fine details. Furthermore, there are patterns (textures, languages) which are difficult to describe verbally. Given a single sketch, our novel “draw in 2D and match in 3D” algorithm spots the desired content under large view variations. We show that off-the-shelf deep features, for sketch matching, have limited view point invariance. Additionally, this work exposes the challenges of using the scene text as a pictorial cue. We propose a novel strategy to overcome these challenges across multiple languages. Our “just draw it” method overcomes the language understanding barrier. We show that sketch based text spotting works, without alteration, for arbitrary font shapes, which standard text detectors find hard to spot. Even in case of custom made text detector (for arbitrary shaped fonts), sketch based text spotting demonstrates complimentary performance. We provide extensive evaluation on public datasets. We also provide a fine grained dataset “Crossroads” which includes tough scenarios for generating navigational instructions. Finally we demonstrate the performance of our view invariant sketch detectors in robotic navigation scenarios using MINOS simulator which contains reconstructed indoor environments.

With the development of deep learning technology, many deep learning methods have been applied to font recognition and generation. However, few studies focus on font inpainting problems. This paper is dedicated to repairing damaged fonts based on style to repair damaged fonts in a better way. In this paper, we propose a CGAN (Conditional Generative Adversarial Nets)-based font repair method. This paper uses the content accuracy and style similarity of the repaired image as an evaluation index to evaluate the accuracy of the restored style font. The font content proposed by the paper based on CGAN network repair style is similar with the correct content.

Signboard detection and recognition is an important task in automated context-aware marketing. Recently many scripting languages like Latin, Japanese, and Chinese have been effectively detected by several machine learning algorithms. As compared to other languages, outdoor Urdu text needs further attention in detection and recognition due to its cursive nature. Urdu detection and recognition are also difficult due to a wide variety of illuminations, low resolution, inconsistent font styles, color, and backgrounds. To overcome the deficiency of Urdu text detection from the outdoor environment, we have proposed a new Urdu-text signboard dataset with 467 ligature categories, containing a 30 + K images for recognition and 700 base images with annotation are created for detection. We also propose a methodology, that consists of 3-phases. In first phase text regions containing Urdu ligatures from shop-signboard images are detected by a faster regional convolutional neural network (FasterRCNN) using pre-trained CNNs like Alexnet and Vgg16. In the second phase detected regions from the first phase are clustered to identify unique ligatures in a dataset. Lastly in the third phase, all detected regions are recognized by 18-layer convolutional neural network trained model. The proposed system has successfully achieved the precision and recall of 87% and 96% respectively using vgg16 model for detection. For the classification of ligatures, a recognition rate of 97.50% is achieved. Recognition of ligatures was also evaluated using bilingual evaluation understudy (BLEU), and achieved an encouraging score of 0.96 on the newly developed Urdu-Signboard dataset.

Scene text recognition is an important task in computer vision. Despite tremendous progress achieved in the past few years, issues such as varying font styles, arbitrary shapes and complex backgrounds etc. have made the problem very challenging. In this work, we propose to improve text recognition from a new perspective by separating the text content from complex backgrounds, thus making the recognition considerably easier and significantly improving recognition accuracy. To this end, we exploit the generative adversarial networks (GANs) for removing backgrounds while retaining the text content . As vanilla GANs are not sufficiently robust to generate sequence-like characters in natural images, we propose an adversarial learning framework for the generation and recognition of multiple characters in an image. The proposed framework consists of an attention-based recognizer and a generative adversarial architecture. Furthermore, to tackle the issue of lacking paired training samples, we design an interactive joint training scheme, which shares attention masks from the recognizer to the discriminator, and enables the discriminator to extract the features of each character for further adversarial training. Benefiting from the character-level adversarial training, our framework requires only unpaired simple data for style supervision. Each target style sample containing only one randomly chosen character can be simply synthesized online during the training. This is significant as the training does not require costly paired samples or character-level annotations. Thus, only the input images and corresponding text labels are needed. In addition to the style normalization of the backgrounds, we refine character patterns to ease the recognition task. A feedback mechanism is proposed to bridge the gap between the discriminator and the recognizer. Therefore, the discriminator can guide the generator according to the confusion of the recognizer, so that the generated patterns are clearer for recognition. Experiments on various benchmarks, including both regular and irregular text, demonstrate that our method significantly reduces the difficulty of recognition. Our framework can be integrated into recent recognition methods to achieve new state-of-the-art recognition accuracy.

Optical Character Recognition (OCR) is a technique that generates text from an image. Recognizing the importance of OCR in real-world settings, a plethora of techniques have been developed for Western, as well as Asian languages. Urdu is a prominent South Asian language and a number of different solutions for Urdu OCR have been proposed. However, fewer attempts have been made to develop end-to-end deep learning-based solutions for recognizing printed Urdu text. Furthermore, several benchmark corpora for Urdu OCR have been developed that can be used for training and evaluation of different OCR techniques. However, there are a number of limitations of the existing Urdu corpora: firstly, most of them have either character or word or text images, which are usually rendered using only a single font, Nastaleeq. Secondly, the volume of the existing datasets is so small that it is not suitable for working with the deep-learning techniques that have achieved groundbreaking results for OCRs. To that end, in this study, we have proposed a very large Multi-level and Multi-script Urdu corpus (MMU-OCR-21). It is the largest-ever Urdu corpus of printed text that is effectively suitable to work with deep learning techniques. In total, the corpus is composed of over 602,472 images, including text-line and word images in three prominent fonts, and their respective ground truth. Also, we have performed experiments using multiple state-of-the-art deep learning techniques for text-line and word level images.

Farsi font detection is considered as the first stage in the Farsi optical character recognition (FOCR) of scanned printed texts. To this aim, this paper proposes an improved version of the speeded-up robust features (SURF) algorithm, as the feature detector in the font recognition process. The SURF algorithm suffers from creation of several redundant features during the detection phase. Thus, the presented version employs the redundant keypoint elimination method (RKEM) to enhance the matching performance of the SURF by reducing unnecessary keypoints. Although the performance of the RKEM is acceptable in this task, it exploits a fixed experimental threshold value which has a detrimental impact on the results. In this paper, an Adaptive RKEM is proposed for the SURF algorithm which considers image type and distortion, when adjusting the threshold value. Then, this improved version is applied to recognize Farsi fonts in texts. To do this, the proposed Adaptive RKEM-SURF detects the keypoints and then SURF is used as the descriptor for the features. Finally, the matching process is done using the nearest neighbor distance ratio. The proposed approach is compared with recently published algorithms for FOCR to confirm its superiority. This method has the capability to be generalized to other languages such as Arabic and English.

Word clouds are widely used for non-analytic purposes, such as introducing a topic to students, or creating a gift with personally meaningful text. Surveys show that users prefer tools that yield word clouds with a stronger emotional impact. Fonts and color palettes are powerful typographical signals that may determine this impact. Typically, these signals are assigned randomly, or expected to be chosen by the users. We present an affect-aware font and color palette selection methodology that aims to facilitate more informed choices. We infer associations of fonts with a set of eight affects, and evaluate the resulting data in a series of user studies both on individual words as well as in word clouds. Relying on a recent study to procure affective color palettes, we carry out a similar user study to understand the impact of color choices on word clouds. Our findings suggest that both fonts and color palettes are powerful tools contributing to the affects evoked by a word cloud. The experiments further confirm that the novel datasets we propose are successful in enabling this. We also find that, for the majority of the affects, both signals need to be congruent to create a stronger impact. Based on this data, we implement a prototype that allows users to specify a desired affect and recommends congruent fonts and color palettes for the word.

User attentional analyses on web elements help in synthesis and rendering of webpages. However, majority of the existing analyses are limited in incorporating the intrinsic visual features of text and images. This study aimed to analyze the influence of elements’ visual features (font-size, font-family, color, etc., for text; and brightness, color, intensity, etc., for images) besides their position on users’ free-viewing visual attention. The investigation includes: (i) user’s position-based attention allocation on text and image web elements, (ii) identification of informative visual features with respect to the attention, (iii) performance of informative visual features in predicting the ordinal visual attention (fixation-indices). Towards the study, an eye-tracking experiment was conducted with 42 participants on 36 real-world webpages. The analyses revealed: (i) Though users predominantly allocate the initial attention to MiddleCenter}, MiddleLeft, TopCenter, TopLeft regions, the elements in Right and Bottom regions are not completely ignored; (ii) Space -related (column-gap, line-height, padding) and font Size -related (font-size, font-weight) intrinsic text features, and Mid-level Color Histogram intrinsic image features are informative, while position and size are informative for both the types; (iii) the informative visual features predict the ordinal visual attention on an element with 90% average accuracy and 70% micro-F1 score. Our approach finds applications in element-granular web-designing and user attention prediction.

For Chinese font images, when all their strokes are replaced by pattern elements such as flowers and birds, they become flower–bird character paintings, which are traditional Chinese art treasures. The generation of flower–bird painting requires professional painters’ great efforts. How to automatically generate these paintings from font images? There is a huge gap between the font domain and the painting domain. Although many image-to-image translation frameworks have been proposed, they are unable to handle this situation effectively. In this study, a novel method called font-to-painting network (F2PNet) is proposed for font-to-painting translation. Specifically, an encoder equipped with dilated convolutions extracts features of the font image, and then the features are fed into the domain translation module for mapping the font feature space to the painting feature space. The acquired features are further adjusted by the refinement module and utilised by the decoder to obtain the target painting. The authors apply adversarial loss and cycle-consistency loss to F2PNet and further propose a loss term, which is called recognisability loss and makes the generated painting have font-level recognisability. It is proved by experiments that F2PNet is effective and can be used as an unsupervised image-to-image translation framework to solve more image translation tasks.

The performance of document text recognition depends on text line segmentation algorithms, which heavily relies on the type of language, author’s writing style, pen type, and document quality. In this paper, we present a novel unsupervised text-line segmentation algorithm for printed Arabic documents with and without diacritics. The presented approach employs a projection profile along with connected components in an iterative manner to detect text-lines. The primary benefits of the presented algorithm are (i) it is not threshold dependent, (ii) it is not required a training phase for threshold selection, and (iii) it is robust towards page rotation, font type, size, and style variation for both with and without diacritics documents. The extensive computational simulations on manually collected dataset prove the efficiency of the proposed scheme compared with several baseline and states of the art methods, including, Voronoi, X-Y Cut, Docstrum, Smearing and Seam-carving methods. Computational time analysis also presented.

Word cloud is a common text visualization technique. With the ability of presenting the keywords of a document in a direct way, it has been widely applied in many real-world situations. However, to better represent the main idea of a document, a critical aspect for word cloud design is to set an appropriate font size to facilitate semantic expression. In this paper, we explore the influence of font scale on semantic expression and evaluate font size of word cloud in a more systematic approach. To quantify semantic information of a document, we utilize an LDA ensemble-based method to support interactive selection of topics and obtain the semantics of documents in a scientific way. We conducted two pilot studies to decide important attributes of word clouds for the formal study. Through formal study 1, we find that the scale affects the semantic expression of word cloud, including accuracy, time and confidence in making judgments. In study 2, we explored different semantic expression patterns of word clouds under different document categories. Our findings aimed at optimizing the scale of word cloud and improving its semantic expressing ability.

The perceived quality of a document is affected by various factors, including grammat- icality, readability, stylistics, and expertise depth, making the task of document quality assessment a complex one. In this paper, we explore this task in the context of assessing the quality of Wikipedia articles and academic papers. Observing that the visual rendering of a document can capture implicit quality indicators that are not present in the document text — such as images, font choices, and visual layout — we propose a joint model that combines the text content with a visual rendering of the document for document qual- ity assessment. Our joint model achieves state-of-the-art results over five datasets in two domains (Wikipedia and academic papers), which demonstrates the complementarity of textual and visual features, and the general applicability of our model. To examine what kinds of features our model has learned, we further train our model in a multi-task learning setting, where document quality assessment is the primary task and feature learning is an auxiliary task. Experimental results show that visual embeddings are better at learning structural features while textual embeddings are better at learning readability scores, which further verifies the complementarity of visual and textual features.

With the tens of thousands of fonts that are now readily available, it is non-trivial to select the most suitable font for a given use case. Considering the impact of the choice of font on human perception of the text, there is a strong need for semantic font search and recommendation. Aiming to fulfill this need, we induce a typographical lexicon providing associations between words and fonts. For this purpose, we determine font vectors for basic and complex emotions, based on word similarities, antonymy information, and Plutchik’s Wheel of Emotions. We create a large font lexicon, named FontLex, relying on emotion associations between the words and the fonts. We evaluate our results through user studies and find that for the majority of the evaluated words, the fonts recommended by FontLex are preferred. We also further extend the dataset using synonyms of font attributes and emotion names. Finally, using CNN embeddings of the fonts, we expand our attribute score assignment to new fonts. The resulting FontLex resource provides mappings between 6.7K words and 2K fonts. Our proof of concept application demonstrates how FontLex can be invoked to obtain semantic font recommendation for poster design.

When repeatedly selected features have predictive value, an observer can learn to prioritize them. However, relatively little is known about the mechanisms underlying this persistent statistical learning. In two experiments, we investigated the boundary conditions of statistical learning. Each task included a training phase where targets appeared more frequently in one of two target colors, followed by a test phase where targets appeared equally in both colors. A posttest survey probed awareness of target color probability differences. Experiment 1 tested whether statistical learning requires the predictive feature to be inherently bound to the target. Participants searched for a horizontal or vertical line among diagonal distractors and reported its length (long or short). In the bound condition, targets and distractors were colored, whereas targets were presented in white font and surrounded by colored boxes in the unbound condition. Experiment 2 tested whether reducing task difficulty by simplifying the judgment (horizontal or vertical) would eliminate statistical learning. The results suggested that statistical learning is robust to manipulations of binding, but is attenuated when task difficulty is reduced. Finally, we found evidence that explicit awareness may contribute to statistical learning, but its effects are small and require large sample sizes for adequate detection. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

Current software tools for computer generation of Persian calligraphy can be mostly described as conventional fonts and typesetting software, which basically neglect the ‘variations’ of real calligraphy performed by hand, in terms of personalization to different calligraphers’ styles, as well as their statistical characteristics. In this paper, we address the problem of natural-looking Persian calligraphy synthesis via a machine learning based approach, at the level of subwords. Given images of samples written by a calligrapher, we train a parametric model to imitate the style. The core idea is to make use of templates (fonts) as a source of background knowledge, and learn a probabilistic mapping from them to personal styles of calligraphers, which is posed as transformation of attributed graphs using neural networks with sliding windows. This can be understood as adding ‘naturalness’ to a Persian calligraphy font, in essence. We report both objective and subjective evaluations, including the model performance in writer (calligrapher) identification task and Visual Turing Test. The results of the latter suggest that humans are unable to distinguish the calligraphy synthesized by our approach from real calligraphy in many cases.

Building a robust Optical Character Recognition (OCR) system for languages, such as Arabic with cursive scripts, has always been challenging. These challenges increase if the text contains diacritics of different sizes for characters and words. Apart from the complexity of the used font, these challenges must be addressed in recognizing the text of the Holy Quran. To solve these challenges, the OCR system would have to undergo different phases. Each problem would have to be addressed using different approaches, thus, researchers are studying these challenges and proposing various solutions. This has motivate this study to review Arabic OCR dataset because the dataset plays a major role in determining the nature of the OCR systems. State-of-the-art approaches in segmentation and recognition are discovered with the implementation of Recurrent Neural Networks (Long Short-Term Memory-LSTM and Gated Recurrent Unit-GRU) with the use of the Connectionist Temporal Classification (CTC). This also includes deep learning model and implementation of GRU in the Arabic domain. This paper has contribute in profiling the Arabic text recognition dataset thus determining the nature of OCR system developed and has identified research direction in building Arabic text recognition dataset.

The clusters of historical documents are of great importance in terms of cultural and scientific. In order to access the documents, originality should be maintained. So conversion of digital form is highly required for recognition. While converting, those documents may be due to poor quality, overlapping of characters, complex background and so on. In this paper, an efficient system for recognizing English characters from degraded historical document images is proposed. Initially, modified Adaptive Thresholding based binarization process is performed to eliminate the noise content in the input image. The characters are segmented through the rectangular bounding box method. Then Local binary pattern (LBP) algorithm is enforced to extricate the features of each characters. Finally, Spatial Pyramid Matching (SPM) classifier is used for texture classification. HDLA 2011 dataset is employed to validate the proposed method. The proposed method achieves 94.6% recognition accuracy and 0.34 s computation time for Lucida Black-letter font. This method also outperforms better than the existing recognition techniques.

In analytical systems of Arabic text recognition, segmentation is a critical and decisive stage. In this stage, words after being extracted from the text document are segmented into individual characters to allow the feature extraction and classification. The absence of standard corpus for printed Arabic text does not allow a good and objective comparison between different segmentation systems. In this paper, we propose two contributions. The first one is a multi-font reference corpus of printed Arabic text, in which we grouped all segmentation problems and which can be used as a reference to compare different segmentation systems. The second is a hybrid method for segmentation of printed multi-font Arabic text. The proposed method is based on two of the most known techniques in the field: contour based and Template matching techniques. This method is insensitive to the font variability and the characters overlapping. To evaluate the results of this method, we have studied, implemented and tested other methods in literature by using the proposed corpus. The experimental results show that our method gives better segmentation rates.

Verification of font style followed in a file is a difficult task to classify. An artificial intelligence based algorithm network can effectively perform this task in reduced time. Capsule network is one among such algorithm and an emerging technique implemented for so many classification process with limited datasets. The proposed font style classification algorithm is enforced with Capsule Network (CapsNet) algorithm for executing the font style classification task. The proposed method is confirmed by classifying times new roman, Arial black and Algerian font style in English letters along with the performance evaluation in terms of accuracy and confusion matrix parameters. The proposed network structure is also compared with the existing Naive Bayes (NB), Decision Tree (DT) and K nearest neighbor (KNN) algorithms for comparative study and the evaluation result indicates that the proposed font style classification model based on CapsNet is classifying the images with better accuracy, F1 score and Gmean.

Text effect transfer aims at learning the mapping between text visual effects while maintaining the text content. While remarkably successful, existing methods have limited robustness in font transfer and weak generalization ability to unseen effects. To address these problems, we propose FET-GAN, a novel end-to-end framework to implement visual effects transfer with font variation among multiple text effects domains. Our model achieves remarkable results both on arbitrary effect transfer between texts and effect translation from text to graphic objects. By a few-shot fine-tuning strategy, FET-GAN can generalize the transfer of the pre-trained model to the new effect. Through extensive experimental validation and comparison, our model advances the state-of-the-art in the text effect transfer task. Besides, we have collected a font dataset including 100 fonts of more than 800 Chinese and English characters. Based on this dataset, we demonstrated the generalization ability of our model by the application that complements the font library automatically by few-shot samples. This application is significant in reducing the labor cost for the font designer.

The automatic style translation of Chinese characters (CH-Char) is a challenging problem. Different from English or general artistic style transfer, Chinese characters contain a large number of glyphs with the complicated content and characteristic style. Early methods on CH-Char synthesis are inefficient and require manual intervention. Recently some GAN-based methods are proposed for font generation. The supervised GAN-based methods require numerous image pairs, which is difficult for many chirography styles. In addition, unsupervised methods often cause the blurred and incorrect strokes. Therefore, in this work, we propose a three-stage Generative Adversarial Network (GAN) architecture for multi-chirography image translation, which is divided into skeleton extraction, skeleton transformation and stroke rendering with unpaired training data. Specifically, we first propose a fast skeleton extraction method (ENet). Secondly, we utilize the extracted skeleton and the original image to train a GAN model, RNet (a stroke rendering network), to learn how to render the skeleton with stroke details in target style. Finally, the pre-trained model RNet is employed to assist another GAN model, TNet (a skeleton transformation network), to learn to transform the skeleton structure on the unlabeled skeleton set. We demonstrate the validity of our method on two chirography datasets we established.

Robotic calligraphy, as a typical application of robot movement planning, is of great significance for the inheritance and education of calligraphy culture. The existing implementations of such robots often suffer from its limited ability for font generation and evaluation, leading to poor writing style diversity and writing quality. This paper proposes a calligraphic robotic framework based on the generative adversarial nets (GAN) to address such limitation. The robot implemented using such framework is able to learn to write fundamental Chinese character strokes with rich diversities and good quality that is close to the human level, without the requirement of specifically designed evaluation functions thanks to the employment of the revised GAN. In particular, the type information of the stroke is introduced as condition information, and the latent codes are applied to maximize the style quality of the generated strokes. Experimental results demonstrate that the proposed model enables a calligraphic robot to successfully write fundamental Chinese strokes based on a given type and style, with overall good quality. Although the proposed model was evaluated in this report using calligraphy writing, the underpinning research is readily applicable to many other applications, such as robotic graffiti and character style conversion.

This paper presents a novel approach for Indian Number Plate Recognition using the Convolution Neural Network (CNN). We generated and augmented synthetic data in Indian standards with variant background and font. We also collected and labelled real data of Indian License plates to create a genuine Indian oriented dataset. We compared existing models like SVM, KNN and CRNN in terms of accuracy and inference time, and selected LPRNet. We first trained the existing LPRNet model, which is designed for Chinese License plates on synthetic data and secured 85% accuracy. We then modified the LPRNet model and secured 93% accuracy on the same synthetic data. We applied the post-processing technique of pattern recognition rules for Indian standards of License plates to further improve accuracy. The Proposed system achieved 95% accuracy after fine-tuning with real data. Our model is very lightweight and can easily be deployed on Nvidia-Jetson nano, Intel Neural Compute Stick.

Arabic printed script segmentation and recognition techniques change from font to other i.e. each font has particular properties calligraphic and structural which differ with other. Majority of segmentation system suffer in word or sub word segmentation into characters because they consider one algorithm to segment all kind of Arabic printed font, style and size. The goal of this work is to prepare a system of word or sub word Optical Font Arabic Recognition (OFAR) for different font size and style of Arabic printed script, in order to integrate it in global Arabic Optical Character Recognition (AOCR) to choose preferred and good segmentation algorithm. APTI database was used to extract last ten pixels for each word or sub word to build new database of last 10 pixels for each word; OFAR is based upon this new database and our extraction approach called Pixels Continuity (PC) algorithm in different matrix direction and some histogram statistics to extract 20 features. Three KNN classifiers with K=5 and three different distances using Cityblock, Euclidean and Correlation based upon majority-vote are used to evaluate the system robustness. This classifier is compared in the first time with Back propagation Neural Network and Steerable Pyramid (SP) algorithm to re cognize three font families, then in the second time with Gaussian Mixture Models (GMMs) to recognize font and size. The average recognition results obtained was 99.55% about font and size and 98.17% for font, size and style recognition.

Emotion plays a critical role in calligraphy composition, which makes the calligraphy artwork impressive and have a soul. However, previous research on calligraphy generation all neglected the emotion as a major contributor to the artistry of calligraphy. Such defects prevent them from generating aesthetic, stylistic, and diverse calligraphy artworks, but only static handwriting font library instead. To address this problem, we propose a novel cross-modal approach to generate stylistic and diverse Chinese calligraphy artwork driven by different emotions automatically. We firstly detect the emotions in the text by a classifier, then generate the emotional Chinese character images via a novel modified Generative Adversarial Network (GAN) structure, finally we predict the layout for all character images with a recurrent neural network. We also collect a large-scale stylistic Chinese calligraphy image dataset with rich emotions. Experimental results demonstrate that our model outperforms all baseline image translation models significantly for different emotional styles in terms of content accuracy and style discrepancy. Besides, our layout algorithm can also learn the patterns and habits of calligrapher, and makes the generated calligraphy more artistic. To the best of our knowledge, we are the first to work on emotion-driven discourse-level Chinese calligraphy artwork composition.

In the field of graphology, the fundamental idea is to establish a connection between the handwriting style of people and their personality traits. In this study, an analysis was made on the science of graphology using machine learning and psychology. A basic personality test was applied to the users who voluntarily participated in our study, some distinctive writing characteristics were extracted from the text samples taken from the same users, and machine learning-based prediction models were developed based on the features constructed over these font characteristics. Here, the aim is to examine how successful matching can be implemented automatically from the user's writing style with regard to personality analysis. The results show that even though not very sensitive, the personality traits can be predicted to a certain extent from the writing style by the machine learning models.

Text detection from the natural scene is considered to be a challenging problem due to the complex background, varied light intensity at different locations, a large variety of colors, diverse font style and size. This paper focusses on detecting candidate text objects from the scene. The image is initially preprocessed to remove the noise and enhance the contrast. Then the various objects of the scene are marked and extracted forming a pool of objects. A set of candidate text objects are extracted from this pool of objects and given as output. In order to locate text candidates among these objects, we use Least-Square Support Vector Machine Technique, which trains the model using Char 74K character dataset and CIFAR 10 non-text image dataset. Finally, the trained model was applied to perform a binary classification of text and non-text objects. The results were evaluated over ICDAR 2015 scene images, MSRA500 and SVT datasets and also have been compared to other approaches acquiring encouraging results. Keywords: Text detection, Support Vector Machine, Least Square Support Vector Machine, Machine Learning, Natural scene text extraction

Dokumen digital yang dihasilkan dari alat pemindai tidak dapat diedit. Sehingga, untuk dapat mengedit dokumen digital tersebut dibutuhkan teknologi Optical Character Recognition (OCR). Penelitian ini dilakukan dengan tujuan untuk mengimplementasikan jaringan saraf tiruan backpropagationpada dokumen tercetak dan untuk mengetahui bagaimana tingkat akurasi dari implementasi jaringan saraf tiruan backpropagationpada dokumen tercetak. Penelitian inimenggunakan jaringan lapisan banyak (multilayer network)dengan tiga lapisan. Lapisan input terdiri dari 225 node dengan masukan berupa citra digital berukuran 15 × 15 piksel, lapisan tersembunyi terdiri dari 110 node, dan lapisan output terdiri dari 54 node yang merepresentasikan karakter A-Z, a-z, serta tanda baca titik (.) dan koma (,). Learning rateyang digunakan dalam penelitian ini adalah 0,29. Rata-rata tingkat akurasi yang didapat dari implementasi jaringan saraf tiruan backpropagationdalam penelitian ini adalah 94 % untuk jenis font Arial Unicode Ms, 96,6 % untuk jenis font Tahoma, dan 94 % untuk jenis font Times New Roman.

As a shining pearl in traditional Tibetan culture, historical Tibetan documents have received extensive attention from historians, linguists and Buddhist scholars. These documents are converted into digital form using Tibetan document segmentation and recognition methods. The document digitization is of great significance for the research, protection and inheritance of Tibetan history. This paper proposes an overall segmentation and recognition framework for historical Tibetan document images. Firstly, the historical Tibetan document image is preprocessed to correct imbalanced illumination, tilt and noises, and is further transformed into the binarized image. Secondly, we propose a layout segmentation method based on block projection to segment Tibetan document images into texts, lines and frames. Thirdly, in order to solve the problems of touching strokes between text-lines and curvilinear text-lines, we present a text-line segmentation method based on graph model for historical Tibetan text-line segmentation. Lastly, we present a touching segmentation method to segment touching Tibetan character string, and then recognize Tibetan characters. Experimental results show our proposed methods on layout segmentation, text-line segmentation and touching character string segmentation, achieve the satisfactory performance. The proposed methods can also be applied to other fonts in Tibetan font family.

Generating text to have a consistent style with only a few observed highly-stylized text samples is a difficult task for image processing. The text style involving the typography, i.e., font, stroke, color, decoration, effects, etc., should be considered for transfer. In this paper, we propose a novel approach to stylize target text by decoding weighted deep features from only a few referenced samples. The deep features, including content and style features of each referenced text, are extracted from a Convolutional Neural Network (CNN) that is optimized for character recognition. Then, we calculate the similarity scores of the target text and the referenced samples by measuring the distance along the corresponding channels from the content features of the CNN when considering only the content, and assign them as the weights for aggregating the deep features. To enforce the stylized text to be realistic, a discriminative network with adversarial loss is employed. We demonstrate the effectiveness of our network by conducting experiments on three different datasets which have various styles, fonts, languages, etc. Additionally, the coefficients for character style transfer, including the character content, the effect of similarity matrix, the number of referenced characters, the similarity between characters, and performance evaluation by a new protocol are analyzed for better understanding our proposed framework.

Mathematics is an effective way to measure or describe things objectively and accurately. It is of great significance to accurately master stroke trend, font structure and component position to improve the level of calligraphy art. This paper tries to use mathematical elements to analyze the art of calligraphy. It is concluded that through the knowledge of angle, golden section, structure and curve in mathematics, the rule of stroke trend, font structure, component position, and work shape as well as the overall layout of characters can be identified.

Oracle bone inscriptions are the earliest mature writing system discovered in China. Oracle is usually published in the form of glyphs and pictures, and the related oracle font database environment is closed and limited. Due to the lack of standardization of ancient oracle bone inscriptions, the structure and glyph of the same characters are not unified, and a large number of different shapes and complicated characters coexist. Even ancient character devotees need to complete the basic recognition of oracle with the help of professional reference books and experts. In addition, in the field of traditional literature, distinguishing variant characters of oracle-bone inscriptions needs strong expert knowledge and lacks efficiency. The recognition process consists of two stages. According to the characteristics of oracle-bone inscriptions, such as left-right symmetry, up-down symmetry and positive and negative coexistence, the corresponding data enhancement methods are used to get sufficient training samples in the first stage. In the second stage, the recognition candidate set under each threshold condition in the first stage are used to select the recognition candidate set which is empty under the specific threshold condition, and introduce the prior knowledge to distinguish the oracle variants. In the first stage, computer related methods are used to identify oracle variants. In the second stage, on the basis of the first stage, a set of recognition results that cannot be determined roughly is selected, and a priori knowledge is introduced to integrate multi-domain methods to identify oracle variants. The method proposed in this paper has achieved great results in the recognition of oracle variant characters.