Text Extraction

Scene text recognition brings various new challenges occurs in recent years. Detecting and recognizing text in scenes entails some of the equivalent problems as document processing, but there are also numerous novel problems to face for recognizing text in natural scene images. Recent research in these regions has exposed several promise but present is motionless much effort to be entire in these regions. Most existing techniques have focused on detecting horizontal or near-horizontal texts. In this paper, we propose a new scheme which detects texts of arbitrary directions in natural scene images. Our algorithm is equipped with two sets of characteristics specially designed for capturing both the natural characteristics of texts using MSER regions using Otsu method. To better estimate our algorithm and compare it with other existing algorithms, we are using existing MSRA Dataset, ICDAR Dataset, and our new dataset, which includes various texts in various real-world situations. Experiments results on these standard datasets and the proposed dataset shows that our algorithm compares positively with the modern algorithms when using horizontal texts and accomplishes significantly improved performance on texts of random orientations in composite natural scenes images.

This paper presents a new enhanced text extraction algorithm from degraded document images on the basis of the probabilistic models. The observed document image is considered as a mixture of Gaussian densities which represents the foreground and background document image components. The EM algorithm is introduced in order to estimate and improve the parameters of the mixtures of densities recursively. The initial parameters of the EM algorithm are estimated by the k-means clustering method. After the parameter estimation, the document image is partitioned into text and background classes by the means of ML approach. The performance of the proposed approach is evaluated on a variety of degraded documents comes from the collections of the National library of Tunisia.

In multimedia-based e-Learning systems, there are strong needs for segmenting lecture videos into topic units in order to organize the videos for browsing and to provide search capability. Automatic segmentation is highly desired because of the high cost of manual segmentation. While a lot of research has been conducted on topic segmentation of transcribed spoken text, most attempts rely on

Translators play an important role in conveying ideas and thoughts from one language to another. A contextual based approach to translation of English words from images to the equivalent Hindi words is described in this paper. Text segmentation is an integral and critical process in text extraction which have the major control over the accuracy of any translator. The paper describes a study on text segmentation algorithm with Maximally Stable Extremal Regions (MSER) which is implemented as a preprocessing module of the translation system. Initially the system learns through user disambiguation of appropriate words in the sentence. However, once the system has learned enough contextual information, it performs the automatic translation. The experimental results of segmentation are compared and analyzed. Such a system has a number of uses, including elearning of the English language as well as for Indian tourists from overseas countries. It can be applied for learning any courses. The system can also be extended to include other Indian languages.

— Handwritten character recognition and extraction is an most important part of image processing and pattern recognition. Handwritten Devnagari Characters are more difficult for recognition than any other language or English characters. Devnagari character has so many possible variations in order, direction and shape of the constituent strokes. Each and every person has its own handwriting style. Those are written in very bad manner, such words cannot be easily read by a machine. Due to very bad styles of writing, a lot of difficulties are faced in recognition process. We have developed a system to extract text from handwritten documents. We used 2 languages Marathi and English language. The main purpose is to introduce a method for recognition and extraction of handwritten Devnagari characters using segmentation and recognition. The whole process of text extraction includes main phases-segmentation of characters into line, lines into words and words into characters and then recognition through neural network.

Text in images contain important contents for information indexing and retrieval, automatic annotation and structuring of images. The text extraction from an image involves several steps including preprocessing, segmentation, feature extraction, classification, post processing. Preprocessing covers basic operation on input image like binarization which convert gray Scale image into Binary Image, noise reduction which remove the noisy signal from image. The segmentation stage for segment the given image into line by line and segment each character from segmented line. The Feature extraction defines the characteristics of character. A classification contains the database and does the comparison. It is very useful in today's word, a many time we have to recognized characters, written on sign boards with an big image, road side signals, number plates etc.

Currently, ontology plays an important role in semantic web technology. Ontology learning approach is to distinguish the type of input such as text, dictionary, knowledge, policies, schemes and semi-structured schemes relations. Ontology learning can be explained as information extraction subtask and its objectives are to dig the relevant concepts and relationships from the corpus or a particular type of data sets. In this project, an ontology learning of text extraction from Qur'anic text as input data was assessed using a newly developed support system. The algorithms used to extract Qur'anic text in this project are Alfonseca & Manandhar's and Gupta & Colleagues's approach. The support system will assess and evaluate these two algorithms and compare with the manually text extraction (Gold Standard) in order to come out an appropriate method or technique which suitable to extract the ontologies from Qur'anic text which can help more people to understand the true meaning from Qur'an teaching.

The text data present in images and video contain certain useful information for automatic annotation, indexing, and structuring of images. However variations of the text due to differences in text style, font, size, orientation, alignment as well as low image contrast and complex background make the problem of automatic text extraction extremely difficult and challenging job. A large number of techniques have been proposed to address this problem and the purpose of this paper is to design algorithms for each phase of extracting text from a video using java libraries and classes. Here first we frame the input video into stream of images using the Java Media Framework (JMF) with the input being a real time or a video from the database. Then we apply pre processing algorithms to convert the image to gray scale and remove the disturbances like superimposed lines over the text, discontinuity removal, and dot removal. Then we continue with the algorithms for localization, segmentation and recognition for which we use the neural network pattern matching technique. The performance of our approach is demonstrated by presenting experimental results for a set of static images.

Complex color documents with both graphics and text, where the text varies in color and size, call for specialized binarization techniques. We propose a novel method for binarization of color documents, whereby the foreground text is output as black and the background as white, regardless of the polarity of foreground and background shades. The method employs an edge-based connected component approach to determine text-like components and binarize them individually. The threshold for binarization and the logic for inverting the output are derived from the image data and do not require any manual tuning. Unlike existing binarization methods, our technique can handle documents with multi-colored texts with different background shades. The method is applicable to documents having text of widely varying sizes, usually not handled by local binarization methods. Experiments on a broad domain of target document types illustrate the effectiveness and adaptability of the method.

— Self-Dependency for disabled persons is of prime importance. This paper presents a cost effective prototype system to help blind persons to shop independently. A camera based prototype system is proposed in which label of the product which a blind user wishes to buy will be extracted from the image and will be pronounced as voice. To isolate the object of interest from the captured image sequence or video, the user is asked to shake the product. Background subtraction will be used to detect the object of interest that is nothing but Region of Interest (ROI). Text extraction will be performed on ROI to obtain the label of product. The prototype system will be implemented on Raspberry Pi board.

This paper presents a system for Arabic Part.Of.Spe ech Tagging, which combines morphological analysis with Hidden Markov Model (HMM) and relies on the Arabic sentence structure. On the one hand, the morphological analysis is used to reduce the size of the tags lexicon by segmenting Arabic words in their prefixes, stems, and suffixes due to the fact that Arabic is

Maps convey valuable information by relating names to their positions. In this paper we present a new method for text extraction from raster maps using color space quantization. Previously, most researches in this field were focused on Latin texts and the results for Persian or Arabic texts were poor. In our proposed method we use a Mean-Shift algorithm with proper parameter adjustment and consequently, we apply color transformation to make the maps ready for K-Means algorithm which quantizes the colors in maps to six levels. By comparing to a threshold the text layer candidates are then limited to three. The best layer can afterwards be chosen by user. This method is independent of font size, direction and the color of the text and can find both Latin and Persian/Arabic texts in maps. Experimental results show a significant improvement in Persian text extraction.

A document image, beside text, may contain pictures, graphs, signatures, logos, bar-codes, hand-drawn sketches and/or seals.  Further, the text blocks in an image may be in Manhattan or any complex layout. Document Layout Analysis is an important preprocessing step before subjecting any such image to OCR. Here, the image with complex layout and content is segmented into its constituent components. For many present day applications, separating the text from the non-text blocks is sufficient. This enables the conversion of the text elements present in the image to their corresponding editable form.

In this work, an effort has been made to separate the text areas from the various kinds of possible non-text elements. The document images may have been obtained from a scanner or camera. If the source is a scanner, there is control on the scanning resolution, and lighting of the paper surface. Moreover, during the scanning process, the paper surface remains parallel to the sensor surface. However, when an image is obtained through a camera, these advantages are no longer available. Here, an algorithm is proposed to separate the text present in an image from the clutter, irrespective of the imaging technology used. This is achieved by using both the structural and textural information of the text present in the gray image. A bank of Gabor filters characterizes the statistical distribution of the text elements in the document. A connected component based technique removes certain types of non-text elements from the image.

When a camera is used to acquire document images, generally, along with the structural and textural information of the text, color information is also obtained. It can be assumed that text present in an image has a certain amount of color homogeneity. So, a graph-theoretical color clustering scheme is employed to segment the iso-color components of the image. Each iso-color image is then analyzed separately for its structural and textural properties. The results of such analyses are merged with the information obtained from the gray component of the image. This helps to separate the colored text areas from the non-text elements.

The proposed scheme is computationally intensive, because the separation of the text from non-text entities is performed at the pixel level. Since any entity is represented by a connected set of pixels, it makes more sense to carry out the separation only at specific points, selected as representatives of their neighborhood. Harris' operator evaluates an edge-measure at each pixel and selects pixels, which are locally rich on this measure. These points are then employed for separating text from non-text elements.

Many documents in India and at other places of the world are bi-lingual or tri-lingual in nature. In these documents, successive words in a line of text may be of different scripts (languages). Hence, for OCR of these documents, the script must be recognized at the level of words, rather than lines or paragraphs. A database of about 20,000 words each from 11 Indian scripts, namely, Bengali, Devanagari, Gujarati, Kannada, Malayalam, Odiya, Punjabi, Roman, Tamil, Telugu and Urdu, is created. This is so far the largest database of Indian words collected and deployed for script recognition purpose. Here again, a bank of 36 Gabor filters is used to extract the feature vector which represents the script of the word. The effectiveness of Gabor features is compared with that of DCT and it is found that Gabor features marginally outperform the DCT. Simple, linear and non-linear classifiers are employed to classify the word in the feature space.

It is assumed that a scheme developed to recognize the script of the words would work equally fine for sentences and paragraphs. This assumption has been verified with supporting results. A systematic study has been conducted to evaluate and compare the accuracy of various feature-classifier combinations for word script recognition. We have considered the cases of bi-script and tri-script documents, which are largely available. Average recognition accuracies for bi-script and tri-script cases are 98.4% and 98.2%, respectively. A hierarchical blind script recognizer, involving all eleven scripts has been developed and evaluated, which yields an average accuracy of 94.1%.

Hindi is the lingua-franca of India. Although all non-native speakers can communicate well in Hindi, there are only a few who can read and write in it. In this work, we aim to bridge this gap by building transliteration systems that could transliterate Hindi into at-least 7 other Indian languages. The transliteration systems are developed as a reading aid for non-Hindi readers. The systems are trained on the transliteration pairs extracted automatically from a parallel corpora. All the transliteration systems perform satisfactorily for a non-Hindi reader to understand a Hindi text.

Extracting text from an image and reproducing them can often be a laborious task. We took it upon ourselves to solve the problem. Our work is aimed at designing a robot which can perceive an image shown to it and reproduce it on any given area as directed. It does so by first taking an input image and performing image processing operations on the image to improve its readability. Then the text in the image is recognized by the program. Points for each letter are taken, then inverse kinematics is done for each point with MATLAB/Simulink and the angles in which the servo motors should be moved are found out and stored in the Arduino. Using these angles, the control algorithm is generated in the Arduino and the letters are drawn.

This paper describes a novel method for extraction of colored text from natural scene images. The core segment of the method involves a two-fold smoothing of the color pixels performed along the row and columns sequentially. The 0-1 and 1-0 transitions in the edge image are analyzed to obtain the runlengths of 0s and 1s in each row and column of the image. The smoothed image is obtained by replacing each edge segment by the median of the color values of pixels in that particular segment. A color-based connected component labeling is then performed on the smoothed image and the stroke width and color information
of the resulting components are used to identify the foreground
text components. The method is tested on word images from the
ICDAR 2003 robust reading competition dataset and is found to
enhance the recognition accuracy significantly.

Nowadays the majority of TV channels are performing digitization of their archival materials. The digitization enhances preservation of records and makes the access to documents easier. However, when the quantity of documents become important the digitization is not enough to insure an efficient access. Indeed, we need to extract semantic information to help users to find what we need quickly. The text included in video sequences is an important needful for indexing and searching system. However, this text is difficult to detect and recognize because of the variability of it size, their low resolution characters and the complexity of the backgrounds. To solve these problems, we propose a system performing in two main tasks: first we extract the textual information from the video sequence and second we recognize this text.

In this paper we address the issue of enhancement in the quality of scanned images of old manuscripts. Small portions of the text in these manuscripts have degraded with time and are not readable. We propose a segmentation based histogram matching scheme for enhancing these degraded text regions. To automatically identify the degraded text we use a matched wavelet based text extraction algorithm followed by MRF(Markov Random Field) post processing. Additionally we do background clearing to improve the quality of results. This method does not require any a priori information about the font, font size, background texture or geometric transformation. We have tested our method on a variety of manuscript images. The results show proposed method to be a robust, versatile and effective tool for enhancement of manuscript images.

The systems currently available for contentbased image and video retrieval work without semantic knowledge, i. e. they use image processing methods to extract low level features of the data. The similarity obtained by these approaches does not always correspond to the similarity a human user would expect. A way to include more semantic knowledge into the indexing process is to

This research is conducted to accommodate the needs of visually impaired people through an intelligent system, which reads textual information on papers and produces corresponding voice. Indonesian Automated Document Reader (I-ADR) is operated via a voice-based user interface to scan a document page. Textual information from the scanned page is then extracted using Optical Character Recognition (OCR) techniques. A user can then choose to have the system read the whole page, or they can opt to listen to a summary of the information in page. SIDoBI (Sistem Ikhtisar Dokumen untuk Bahasa Indonesia) is integrated into the system to provide summarization feature. The result of either the whole-page reading or summarization is converted to speech through a text-to-speech synthesizer. This whole system is developed under the Free Open Source Software policy and will be distributed openly to all users in need without any cost. This paper is focused on the text segmentation algorithm implemented in I-ADR to extract text from documents with complex layout. We implemented I-ADR text segmentation module using Enhanced CRLA and propose an improved algorithm for text extraction. Evaluation of the proposed system with various page layouts showed promising results.

Legal documents often have a complex layout with many different headings, headers and footers, side notes, etc. For the further processing, it is important to extract these individual components correctly from a legally binding document, for example a signed PDF. A common approach to do so is to classify each (text) region of a page using its geometric and textual features. This approach works well, when the training and test data have a similar structure and when the documents of a collection to be analyzed have a rather uniform layout. We show that the use of global page properties can improve the accuracy of text element classification: we first classify each page into one of three layout types. After that, we can train a classifier for each of the three page types and thereby improve the accuracy on a manually annotated collection of 70 legal documents consisting of 20,938 text elements. When we split by page type, we achieve an improvement from 0.95 to 0.98 for single-column pages with left marginalia and from 0.95 to 0.96 for double-column pages. We developed our own feature-based method for page layout detection, which we benchmark against a standard implementation of a CNN image classifier.

The approach presented here is based on corpus of freely available German contracts and general terms and conditions. Both the corpus and all manual annotations are made freely available. The method is language agnostic.

The research presented in this thesis addresses the problem of Text Segmentation in Web images. Text is routinely created in image form (headers, banners etc.) on Web pages, as an attempt to overcome the stylistic limitations of HTML. This text however, has a potentially high semantic value in terms of indexing and searching for the corresponding Web pages. As current search engine technology does not allow for text extraction and recognition in images, the text in image form is ignored. Moreover, it is desirable to obtain a uniform ...

— The goal of this paper is to detect the text from video. With the increasing popularity of practical vision systems and smart phones, text detection in natural scenes becomes a critical yet challenging task. In a society driven by visual information and with the drastic expansion of low-priced cameras, text recognition is nowadays a fast changing field. The goal of this paper is develop a robust algorithm that extracts text from video with cluttered background. We propose a simple but efficient methodology for text detection in video frames. The method is based on the gradient information and edge map selection. In the proposed method we first find the gradient of the image and then enhance the gradient information. We later binarized the enhanced gradient image and select the edges by taking the intersection of the edge map with the binary information of the enhanced gradient image. We use the edge detector for generating the edge map. The selected edges are then morphologically dilated and opened using suitable structuring elements and used for text regions. We then perform the projection profile analysis to identify the boundary of the text region. At the end, we implement a false positive elimination methodology to improve the text detection results. Then we make the video from image frames which contain the detection of text.

—Text-line extraction in handwritten documents is an important step for document image understanding, and a number of algorithms have been proposed to address this problem. In order to overcome this limitation, we develop text-line extraction algorithm for cursive handwriting. Our method is based on connected components (CCs), however, unlike conventional methods, we analysed strokes and partition under-segmented CCs into normalized ones. Due to this normalization, the proposed method is able to estimate the states of CCs for a range of different languages and writing styles. I. INTRODUCTION TEXT-LINE extraction in document images is an essential step for various document image processing tasks such as layout analysis and optical character recognition (OCR).Therefore, there have been a lot of researches in this area, and a number of algorithms have been proposed for the extraction of text-lines in machine-printed document images. However, text-line extraction in handwritten documents is still considered a challenging problem: the scale and orientation of characters are spatially varying, inter-line distances are irregular, and characters may touch across words and/or text-lines. Handwriting detection is a technique or ability of computer to receive & interpret intelligible handwritten input from source. Handwriting recognition is comparatively difficult, because different people have different handwriting style. In optical character recognition, segmentation is a significant phase and accuracy of character recognition highly depends on accuracy of segmentation. Incorrect segmentation leads to incorrect character recognition. Segmentation phase includes text line, word, and character segmentation. Text line detection and separation in digital image documents is a challenging job for handwritten document analysis and character recognition. The problem becomes compounded if the text lines in the text image are connected or overlapped. Emergence of these problems is common in handwritten documents in comparison of printed documents because of individual's varying handwriting styles. Researchers are continuously working on these problems for different languages. Text-line extraction in handwritten documents is an important step for document image understanding, we develop a language-independent text-line extraction algorithm. However, most conventional work focused on specific character sets. That is, conventional algorithms address the variations caused by individual writers by exploiting language-specific features. The situation is worse for Indian scripts where most characters are connected. On the other hand, character components are placed in a one-dimensional way in cursive Latin-based and Indian scripts, allowing us to develop horizontal bottom-up clustering rules. Our method is based on connected components (CCs), however, unlike conventional methods; we analyze strokes and partition under-segmented CCs into normalized ones. Due to this normalization, the proposed method is able to estimate the states of CCs for a range of different languages and writing styles. From the estimated states, we build a cost function whose minimization yields text-lines. We develop an effective CC segmentation method: by partitioning under-segmented CCs into normalized ones, we can estimate states reliably in a variety of documents.

This paper presents an approach for knowledge discovery in texts extracted from the Web. Instead of analyzing words or attribute values, the approach is based on concepts, which are extracted from texts to be used as characteristics in the mining process. Statistical techniques are applied on concepts in order to find interesting patterns in concept distributions or associations. In this way, users can perform discovery in a high level, since concepts describe real world events, objects, thoughts, etc. For identifying concepts in ...

We present an automated method of generating human-readable summaries from a variety of text documents including newspaper articles, business reports, government documents, even broadcast news transcripts. Our approach exploits an em- pirical observation that much of the written text display certain regularities of organization and style, which we call the Discourse Macro Structure (DMS). A summary is therefore created to reflect the components of a given DMS. In order to pro- duce a coherent and readable summary we select continuous, well-formed passages from the source document and assemble them into a mini-document within a DMS template. In this paper we describe an automated summarizer that can generate both short indicative abstracts, useful for quick scanning of a list of documents, as well as longer informative digests that can serve as surrogates for the full text. The summarizer can assist the users of an informa- tion retrieval system in assessing the quality of the result...

Text data present in multimedia contain useful information for automatic annotation, indexing. Extracted
information used for recognition of the overlay or scene text from a given video or image. The Extracted
text can be used for retrieving the videos and images. In this paper, firstly, we are discussed the different
techniques for text extraction from images and videos. Secondly, we are reviewed the techniques for
indexing and retrieval of image and videos by using extracted text.

This paper presents a system for Arabic Part.Of.Spe ech Tagging, which combines morphological analysis with Hidden Markov Model (HMM) and relies on the Arabic sentence structure. On the one hand, the morphological analysis is used to reduce the size of the tags lexicon by segmenting Arabic words in their prefixes, stems, and suffixes due to the fact that Arabic is

Scene text recognition brings various new challenges occurs in recent years. Detecting and recognizing text in scenes entails some of the equivalent problems as document processing, but there are also numerous novel problems to face for recognizing text in natural scene images. Recent research in these regions has exposed several promise but present is motionless much effort to be entire in these regions. Most existing
techniques have focused on detecting horizontal or near-horizontal texts. In this paper, we propose a new scheme which detects texts of arbitrary directions in natural scene images. Our algorithm is equipped with two sets of characteristics specially designed for capturing both the natural characteristics of texts using MSER regions using Otsu method. To better estimate our algorithm and compare it with other existing algorithms, we are using existing MSRA Dataset, ICDAR Dataset, and our new dataset, which includes various texts in various real-world situations. Experiments results on these standard datasets and the proposed dataset shows that our algorithm compares positively with the modern algorithms when using horizontal texts and accomplishes significantly improved performance on texts of random orientations in
composite natural scenes images.