Text Localization Research Articles

A Text Location Method for Web Images

This paper proposes a text localization method with multi-features based on cascade classifier for a variety of web images. Specifically, first, the original image is divided into sub-images with different scales, which form more satisfactory edge image blocks after being pretreated respectively; then, the researchers determine in the classifier whether the text area is contained in the candidate image blocks according to the edge connectivity characteristics, stroke density characteristics and text arrangement characteristics of text area; finally, the location results of sub-images with different scales are mixed together to obtain the final result. The experiments show that this location method has the relatively high precision and recall rate and quite strong robustness, which is suitable for a variety of web images.

End-to-end scene text recognition using tree-structured models

Detecting and recognizing text in natural images are quite challenging and have received much attention from the computer vision community in recent years. In this paper, we propose a robust end-to-end scene text recognition method, which utilizes tree-structured character models and normalized pictorial structured word models. For each category of characters, we build a part-based tree-structured model (TSM) so as to make use of the character-specific structure information as well as the local appearance information. The TSM could detect each part of the character and recognize the unique structure as well, seamlessly combining character detection and recognition together. As the TSMs could accurately detect characters from complex background, for text localization, we apply TSMs for all the characters on the coarse text detection regions to eliminate the false positives and search the possible missing characters as well. While for word recognition, we propose a normalized pictorial structure (PS) framework to deal with the bias caused by words of different lengths. Experimental results on a range of challenging public datasets (ICDAR 2003, ICDAR 2011, SVT) demonstrate that the proposed method outperforms state-of-the-art methods both for text localization and word recognition.

글자가 움직이는 동영상 자막 편집 어플리케이션 개발

본 논문에서는 사용자가 스마트폰을 이용하여 직접 찍은 동영상에 움직이는 자막을 즉석에서 간편하게 편집할 수 있는 안드로이드 어플리케이션인 VIVID를 개발하였다. VIVID를 통해 휴대폰에 저장된 동영상에 원하는 자막의 시간, 텍스트, 위치, 모션(자막의 움직임)을 간편하게 설정할 수 있으며, 결과물을 html 형식으로 웹 서버에 업로드하여 다른 사람들과 함께 공유할 수 있게 하였다. In this paper, we developed an Android application named VIVID where users can edit the moving captions easily on smartphone videos. This makes it convenient to set the time range, text, location and motion of caption text on the video. The editing result is uploaded to web server in html and can be shared with other users.

Geo-Visualization of Agricultural Products Price Data Based on WebGIS

In China, there is a large amount of data about agricultural products price released by thousands of agricultural websites on the Internet. These records are basically displayed in the form of text, table, and statistics chart but location information cannot be displayed graphically. Rich client and WebGIS techniques can improve representation of agricultural products price and help to provide effective technological solutions for increasing data. First, we describe the methods of standardizing agricultural products price data, which is crawled from agricultural websites through vertical search engine. Forward maximum matching algorithm is then explained and applied for spatializing data. After that, technologies of WebGIS based on rich client (JavaScript/Ajax and Flex) are summarized and analyzed and the map visualization methods of agricultural products price are presented, taking Baidu Map API and ArcGIS API for Flex for example. The methods can provide visual, interactive, user-friendly representation of agricultural products price data in geography information system as well as share visualization services through Internet or Intranet to improve data exchange and utilization. Compared to traditional WebGIS, WebGIS based on RIA improves user satisfaction and increases response speed. The results demonstrate that WebGIS by using RIA technology is a feasible solution method for map visualization of agricultural products price.

Detection and recognition of text superimposed in images base on layered method

Detection and recognition of text superimposed in complex background has been considered as a challenging problem. Most of the existing methods first locate the text regions and then feed them into OCR package for recognition. However, these methods cannot achieve good recognition performance due to the complex background. For this purpose, this paper proposes a novel text detection and recognition method by using color clustering to divide images into multiple layers according to main color class. In the proposed method, we exploited a connected component analysis to obtain the candidate text regions from each color layer, and then a cascade Adaboost classifier is adopted to determine whether the candidate text regions is real text regions in the corresponding image layer. Because the monochrome color exists in each layer, the interference of the background can be effectively reduced, which can significantly improve the accuracy of text regions localization. Afterwards, an OCR package is used to recognize the text regions which have been located by the cascade Adaboost classifier. Since the text region has a monochrome color, it helps to greatly improve the recognition rate. Finally, the relationship between different layers is used to verify the recognition results by the text location. The experimental results show that the proposed approach significantly outperforms the existing methods.

An effective graph-cut scene text localization with embedded text segmentation

This paper presents an effective and efficient approach to extracting scene text from images. The approach first extracts the edge information by the local maximum difference filter (LMDF), and at the same time a given image is decomposed into a group of image layers by color clustering. Then, through combining the characteristics of geometric structure and spatial distribution of scene text with the edge map, the candidate text image layers are identified. Further, in character level, the candidate text connected components are identified using a set of heuristic rules. Finally, the graph-cut computation is utilized to identify and localize text lines with arbitrary directions. In the proposed approach, the segmentation of text pixels is efficiently embedded into the computation of text localization as a part. The comprehensive evaluation experiments are performed on four challenging datasets (ICDAR 2003, ICDAR 2011, MSRA-TD500 and The Street View Text (SVT)) to verify the validation of our approach. In the comparison experiments with many state-of-the-art methods, the results demonstrate that our approach can effectively handle scene text with diverse fonts, sizes, colors, different languages, as well as arbitrary orientations, and it is robust to the influence of illumination change.

An Efficient Method of Text Localization in Complicated Background Scenes

We propose a novel method for text localization in low illumination images with complicated background scenes. Low illumination image, complex background and variations of text script make text localization problem challenging. The proposed method use density-based information and rectangle window in the residual edge image. The proposed method firstly is color space conversion from RGB to Ycbcr color space, secondly we extract Y component from Ycbcr color space and get image enhancement from Y component, then extract out the vertical edges of enhanced Y component, thirdly, long curve and random noise in complicated background are removed, finally, search the text localization and segment the text out from the original low illumination images. We evaluate our method using natural image data set and ICDAR data set. The experimental results show that the proposed method has more robust to interference characters and more accurate when compared with other methods.

A New Method For Text Location in News Video Based on Ant Colony Algorithm

Abstract—Text in video is a very compact and accurate clue for video indexing and summarization. The paper presents a new method for text location in news video with ant colony algorithm. Three features of characters are extracted as a basis for the formation of heuristic function. In order to balance the weight of the three features, three functions are introduced to transform them. The ants will be randomly put on sub-blocks of video frames for searching text areas. Therefore, ants would leave pheromone in each sub-block. After the ant colony algorithm is finished, it produces a pheromone matrix. By binarizing the pheromone matrix, the text blocks can be located. The result has proved that the binarization method proposed in this paper is more accurate than otsu’s method. At last, to reduce the false detection rate, the different directions of edge intensity ratio of text areas are computed, as the real text areas’ edge intensity ratio is much smaller than the false one.

Simple and Efficient Text Localization for Compressed Image in Mobile Phone

Extraction of the text data present in images involves Text detection, Text localization, Text tracking, Text extraction, Text Enhancement and Text Recognition. Due to its inherent complexity, traditional text localization algorithms in natural scenes, especially in multi-context scenes, are not implementable under low computational resources architectures such as mobile phones. In this paper, we proposed a simple method to automatically localize signboard texts within JPEG mobile phone camera images. Taking into account the information provided by the Discrete Cosine Transform (DCT) used by the JPEG compression format, we delimitate the borders of the most important text region. This system is simple, reliable, affordable, easily implementable, and quick even working under architectures with low computational resources.

Fuzzy-Based Segmentation for Variable Font-Sized Text Extraction from Images/Videos

Textual information embedded in multimedia can provide a vital tool for indexing and retrieval. A lot of work is done in the field of text localization and detection because of its very fundamental importance. One of the biggest challenges of text detection is to deal with variation in font sizes and image resolution. This problem gets elevated due to the undersegmentation or oversegmentation of the regions in an image. The paper addresses this problem by proposing a solution using novel fuzzy-based method. This paper advocates postprocessing segmentation method that can solve the problem of variation in text sizes and image resolution. The methodology is tested on ICDAR 2011 Robust Reading Challenge dataset which amply proves the strength of the recommended method.

Automatic Inpainting Scheme for Video Text Detection and Removal

We present a two stage framework for automatic video text removal to detect and remove embedded video texts and fill-in their remaining regions by appropriate data. In the video text detection stage, text locations in each frame are found via an unsupervised clustering performed on the connected components produced by the stroke width transform (SWT). Since SWT needs an accurate edge map, we develop a novel edge detector which benefits from the geometric features revealed by the bandlet transform. Next, the motion patterns of the text objects of each frame are analyzed to localize video texts. The detected video text regions are removed, then the video is restored by an inpainting scheme. The proposed video inpainting approach applies spatio-temporal geometric flows extracted by bandlets to reconstruct the missing data. A 3D volume regularization algorithm, which takes advantage of bandlet bases in exploiting the anisotropic regularities, is introduced to carry out the inpainting task. The method does not need extra processes to satisfy visual consistency. The experimental results demonstrate the effectiveness of both our proposed video text detection approach and the video completion technique, and consequently the entire automatic video text removal and restoration process.

Image Retrieval System for Composite Images using Directional Chain Codes

Most of the image retrieval systems published and implemented have focused on basic features like color, shape and texture of an image with little or no consideration of the included text region of the image. In this work, we have developed an image retrieval system that works on finding similar composite images containing graphical shapes as well as text from a database of thousands of images. By proposing a novel method for text localization, extraction followed by detection, we have demonstrated how this method outperforms commercial OCR tools. The significant feature of this work is its handling the requirements of invariance to font size, design, text region orientation and its ability to give accurate result even in the presence of complex background and graphical elements. The methodology has been tested for English text but is capable to handle any other language.

Bursty Hot-Words Detection for Campus BBS

In the monitoring of campus public opinions, hot words often reflect the latest burst hot topics within a certain period of time. Therefore, this paper takes in-depth research for bursty hot words detection. In the process of words weight calculation, we consider not only traditional features such as TF, IDF, but also the burstiness, part of speech, length, location in text and other factors. Consequently, the measurement formula of burstiness and the weight calculating formula based on symphysic multi-features are proposed. The weight calculating formula can identify the bursty hot-words quickly and accurately, and then discover the bursty events, finally realizing the early warning of campus public opinions effectively. DOI: http://dx.doi.org/10.11591/telkomnika.v11i6.2671 Full Text: PDF

Characters Strings are Extracted Exhibit Morphology Method of an Image

In understanding an image, extraction of characters existing in the image is considered to be important. Scene images are different from document images, which are composed of characters and complicated background i.e. photo, picture, or painting etc. instead of white one that makes it difficult to be dealt with. Extraction and localization of scene text are used in many applications. In this paper, we have proposed a connected component based method to extract text from natural images. The proposed method uses colour space processing. Character recognition is done through OCR that accepts the input in form of text boxes, which are generated through text detection and localization stages. The Proposed method is robust with respect to font size, colour, orientation, and style. Results of the proposed algorithm, by taking the real scenes, including indoor and outdoor images, shows that proposed method efficiently extracts and localizes the scene text. In this paper, we have introduced a new method to extract characters from scene images using mathematical morphology.Â

Hybrid approach for Farsi/Arabic text detection and localisation in video frames

Video text detection plays an important role in semantic-based video analysis. In this study, a new Farsi/Arabic text detection and localisation approach is proposed. First, with the help of edge extraction, artificial corners are obtained and font size estimation is performed. Second, by combining discrete cosine transform coefficients, texture intensity picture is created. Afterwards, a new Local Binary Pattern (LBP) picture is introduced to describe the obtained texture pattern. The input image is then divided into macro blocks and some features are extracted from them and fed into Support Vector Machone (SVM) classifier to categorize them into text and non-text groups. Finally, the candidate text blocks undergo project profile analysis and empirical rules for text localisation. Experimental results demonstrate that the proposed hybrid approach can be used as an automatic text detection system, which is robust to font size, font colour and background complexity.

Text detection in chart images

Common OCR (Optical Character Recognition) systems fail to detect and recognize small text strings of few characters, in particular when a text line is not horizontal. Such text regions are typical for chart images. In this paper we present an algorithm that is able to detect small text regions regardless of string orientation and font size or style. We propose to use this algorithm as a preprocessing step for text recognition with a common OCR engine. According to our experimental results, one can get up to 20 times better text recognition rate, and 15 times higher text recognition precision when the proposed algorithm is used to detect text location, size and orientation, before using an OCR system. Experiments have been performed on a benchmark set of 1000 chart images created with the XML/SWF Chart tool, which contain about 14000 text regions in total.

A Survey of Text Reading from Scene Images

reading in images or video is important step to achieve content retrieval from images. The content retrieve from image or video content useful information it act as clue for many image based applications such as scene understanding, content based image retrieval text based image indexing, industrial automation. Locating or detecting text from complex background is a challenging task due to variation in size, orientation, style. This paper describes the techniques which try to solving this problem. Keywordimage text detection, text localization, text recognition.

Hypermedia Annotation Presentation

This study explored whether hypermedia annotation presentation in different locations of the text facilitates EFL vocabulary acquisition and reading comprehension. It also examined participants' preferences for the location of hypermedia-annotated text (at the end of the text, in the margin, at the bottom of the screen, or in a pop-up window). Data were collected from 80 intermediate adult EFL learners who read annotated texts and received support in different conditions of hypermedia annotations. A survey of prior knowledge was carried out, followed by vocabulary and reading comprehension tests, and a questionnaire. Results indicated that students who had access to hypermedia annotations outperformed their peers who used traditional glosses listed at the end of the text. There was also a positive impact for texts with hypermedia annotations located in the margin of the text. Learners also preferred hypermedia annotation presentation in the margin. Finally, no relationship was found between participants' preferences and their level of achievement.

Text localization and detection method for born-digital images

Text localization and detection method for born-digital images

Title Extraction from Book Cover Images Using Histogram of Oriented Gradients and Color Information

In this paper, we present a technique to extract the title areas from book cover images. A typical book cover image may contain text, pictures, diagrams as well as complex and irregular background. In addition, the high variability of character features such as thickness, font, position, background and tilt of the text also makes the text extraction task more complicated. Therefore, we propose a two steps efficient method that uses Histogram of Oriented Gradients and color information to find the title areas. Firstly, text localization is carried out to find the title candidates. Finally, refinement process is performed to find the sufficient components of title areas. To obtain the best result, we also use other constraints about the size, ratio between the length and width of the title. We achieve encouraging results of extracted title regions from book cover images which prove the advantages and efficiency of the proposed method.