A method for the rectification of figures and photos in printeddocuments using a single camera-captured image. The algorithm requires a bounding box for the objects in a single-view image. On receiving the bounding box, the desired image can be cropped for further processing. The main feature of segmentation method is that, it exploits the properties of printed figures to detect the boundaries using an optimization scheme. The boundary is extracted to obtain the boundary points and rectifying the distorted image is carried out using a dewarping method. The method improves the quality of output by largely removing perspective distortions. The rectified image having low resolution is converted to a higher resolution image using an edge directed interpolation method. Thus, an enhanced rectified output image is obtained as a result.
We have collected Myanmar old printeddocuments such as different books, newspapers and magazines around 1938. Firstly, document images might be acquired by scanning or capturing. Preprocessing step involves grayscale conversion and noises are removed by applying median filter. In binarization process, gray scale images are converted into black and white image by using local thresholding method. Distortion alignment of input text, scanned or photo that needs to align it by performing skew correction method. And then lines and words are segmented. Characters are segmented by using block definition method. Six structural features such as aspect ratio, termination points, bifurcation points, horizontal strokes, vertical strokes, weight direction features are extracted. These six extracted features for each characters
Abstract— Bilingual document recognition has been the subject of intensive research and our focus is on the recognition of an Oriya-English bilingual documents. In most of our official papers, school text books, it is observed that English words interspersed within the Indian languages. So there is need for an Optical Character Recognition (OCR) system which can recognize these bilingual documents and store it for future use. In this paper we present an OCR system developed for the recognition of Indian language i.e. Oriya and Roman scripts for printeddocuments. For such purpose, it is necessary to separate different scripts before feeding them to their individual OCR system. Firstly, we need to correct the skew followed by segmentation. Here we propose the script differentiation line-wise. We emphasize on Upper and lower matras associated with Oriya and absent in English. We have used horizontal histogram for line distinction belonging to different script. After separation different scripts are sent to their individual recognition engines. Recognition of bilingual script in an image of a document page is of primary importance for a system processing bilingual document. Earlier we had communicated a paper using a single classifier and now three classifiers such as k-nearest neighbor (KNN), convolutional neural networks (CNN) and Support Vector Machines (SVM) schemes have been proposed for analyzing the accuracies for recognition. It has been observed that SVM outperform among all the classifiers.
This layout analysis system was submitted by Tanmoy Nandi & Sumit Kumar Saha, Gnosis Lab, Kolkata, India, Chandranath Adak, School of ICT, Griffith University, Aus- tralia, Durjoy Sen Maitra, Decimal Point Analytics, India, and Bidyut B. Chaudhuri, CVPR Unit, Indian Statistical In- stitute, India. It works with only printed Bangla (or, Bengali) script. Since the REID2017 dataset contains old printed doc- uments, some rigorous preprocessing is required, using fol- lowing steps:
substrate, data can now be sent directly to the press from a computer and from the computer file; it can then be printed within minutes with little set-up time. Digital press technology is quite different than that of traditional offset; yet it has become strikingly similar to offset in quality (Digital Offset Introduction, n.d.). With electrophotography, as stated before, the ink sits atop the paper and does not sink into the actual substrate, as other printing processes are designed to do (Fischer, 2005). All three commercial digital presses studied here share this attribute. While this process yields bright and vibrant colors, it also exacerbates the permanence problem because the ink is not actually part of the paper; it is merely fused to it. The finished product from a digital press is susceptible to scratching and can be altered by finishing techniques or other elements, such as mailing, ultraviolet exposure, chemical exposure, and humidity (Krasne, 2002).
In this paper, we describe the initial stages of our project, the goal of which is to create an integrated archive of the recordings, scanned documents, and photographs that would be accessible online and would provide multifaceted search capabilities (spoken content, biographical information, relevant time period, etc.). The recordings contain retrospective interviews with the witnesses of the totalitarian regimes in Czechoslovakia, where the vocabulary used in such interviews consists of many archaic words and named entities that are now quite rare in everyday speech. The scanned documents consist of text materials and photographs mainly from the home archives of the interviewees or the archive of the State Security. These documents are usually typewritten or even handwritten and have really bad optical quality. In order to build an integrated archive, we will employ mainly methods of automatic speech recognition (ASR), automatic indexing and search in recognized recordings and, to a certain extent, also the optical character recognition (OCR). Other natural language processing techniques like topic detection are also planned to be used in the later stages of the project. This paper focuses on the processing of the speech data using ASR and the scanned typewritten documents with OCR and describes the initial experiments.
The present paper apart from experimenting the methods that performs binarization and character segmentation, also compares other methods present in the literature for same kind of problem. The algorithm was tested on document images representing both handwritten and printeddocuments, SLGS and FGM methods are used for image document analysis. Analysis shows that Fast global minimization algorithm(FGM) is better than the selective local or global segmentation method(SLGS). FGM algorithm works faster while SLGS algorithm takes some time. Fast minimization algorithm can provide more accuracy in binarization as well as in segmentation
Many existing approaches on word spotting rely on ac- curate segmentation of words or connected components . Rath and Manmatha  used vertical projection profile, up- per and lower boundary projection profile and applied dy- namic time warping (DTW) distance for matching similar words. To take care the word segmentation problem, Ley- dier et al.  have used differential features and a cohe- sive elastic distance. Gatos and Pratikakis  proposed a segmentation-free word spotting approach for historical printeddocuments using salient region detection by tem- plate matching at an initial stage. Often, word shape coding is used to encode the words in normal printeddocuments. Lu et al.  proposed a technique to retrieve document images by a set of topological codes based on shape fea- tures including character ascenders/descenders, holes, wa- ter reservoirs information etc.
Because the proposed method works with both the handwritten and the printeddocuments, by using the same procedures can be discussed in the skew detection and correction algorithms the applicability, and the generality of the proposed method can be satisfied, while each of the Hough transform and Projection Profile in other hand can work with both types of document images, whether handwritten or, printed but the results remain poor unless it is supported by some of the conditions for working with each method on its own.
discussion on Orientalism. According to the authors, the “fleeting and work-in-progress nature” (Xavier and Z ̌ upanov 2015:319) of Catholic Orientalist knowledge, lack of printeddocuments and stable information channels, the abrupt end of the Portuguese empire in South Asia (p. 288, p. 319) and the destruction of archives by natural calamities (p. 339) are some of the reasons for the neglect of the Catholic knowledge practices discussed in this work. The book attempts to bring Catholic knowledge practices into the discussion on Orientalism and emphasize its role in the formation of all later “Orientalisms.” The authors endeavor to prove that these practices of constructing India for the European imagination “fed into” all future practices of Orientalism, including British efforts to produce knowledge about India (p. 289).
Lots of work has been done for OCRs for decent quality printeddocuments in English, Gurumukhi, Devanagari, etc. Many researchers had handled the degradation problem of the Devanagari script, though punishing degradation was not considered. This work attempts to progress the quality of highly degraded printed Marathi documents during the preprocessing phase, so that further phases of OCR can be applied effortlessly. This work produced better quality images with higher PSNR values and good values of MI as well. Future attempt will do word segmentation and character segmentation in such a highly degraded printed Marathi document.
3) Segmentation: The Segmentation phase is the most important process in Optical Character Recognition. Segmentation is done by separation from the individual characters from the main image document so that each character is taken separately for recognition process. Segmentation of handwritten characters into different zones (upper, middle and lower zone) and characters is more difficult than that of printeddocuments that are in standard form as handwritten characters have very complex shapes and sizes. This is mainly because of variability in paragraph, words of line and characters of a word, skew, slant, size and curvedness of these characters. The components of two adjacent characters in the handwritten text may be touched or overlapped and this situation creates difficulties in the segmentation process. Touching or overlapping problem occurs frequently in the handwritten characters, because of modified characters in upper-zone and lower-zone.
tion (ACR) technique and a page layout analysis (PLA) procedure to extract text in given document image. ACR technique is used to obtain optimal number of colors (principal colours) in the document. Then, using the principal colors, the document image is split into the separable color planes. On each color plane, PLA procedure is applied to identify the text regions. A merging procedure is applied in the final stage to merge the text regions derived from each of the color planes and to produce the final document. The disadvantage of this approach is that selection of optimal numbers is difficult for a printed newspaper documents as they contain many colors. This method proposed in  introduces a new color reduction technique to decrease number of colors in the document image. This technique estimates the dominant colors in the document and re-assigns the colors of original image to reduced set. Each dominant color defines a color plane in which the connected components (CCs) are extracted. Next, in each color plane a CC filtering procedure is applied which is followed by a grouping procedure. At the end of this stage, blocks of CCs are constructed which are next redefined by obtaining the direction of connection (DOC) property for each CC. Using the DOC property, the blocks of CCs are classified as text or nontext. The identified text blocks are binarized properly using suitable binarization techniques, considering the rest of the pixels as background. The final result is a binary image which contains always black characters in white background independent of the original colors of each text block. The paper , discusses about region-based thresholding for color document images and other paper extracting halftones from printeddocuments using texture analysis are also been studied as part of literature survey. From the above discussion on different methods present in the literature one could say that there is no standard technique to binarize a color document.
• Scanning a collection of printeddocuments, performing OCR, and then manually creating appropriate topics and relevance judgments. The size of a test collection created in this way will be limited by the resources available for the relevance judgment process. However, this technique can accurately model many aspects that may be present in real applications (e.g., unfamiliar fonts, damaged pages, and handwritten annotations). Taghva, et al. experimented with a 204-document English document image collection using this technique. The average length of the documents was 38 pages. He observed no significant effect of degradation on retrieval (Taghva et al., 1994). Tseng and Oard experimented with different combinations of n-grams on a Chinese collection of 8,438 document images. The documents images were scanned from printed material. They observed that combinations of character 1-grams and character 2-grams performed best. Further, they reported that blind relevance feedback did not improve retrieval effectiveness (Tseng & Oard, 2001). Darwish and Oard experimented with a variety of Arabic index terms on an Arabic collection 2,730 document images. The documents were scanned from a single book. They reported that 3-grams and 4-grams are the best index terms for OCR degraded Arabic text (Darwish & Oard, 2002).
Printeddocuments have very well defined gaps in between text lines and in between words. These can be used to separate lines first and consequently extract the words. Horizontal profiling is a technique in which the image is traversed row wise and for each row the count of black pixels are stored. Values less than a predefined threshold(which may be zero also)gives us the white line separating two text lines. The same method can be used to extract words from a text line line giving us the coordinates of words.
The results show that the alignment works prop- erly when using the same photos and take the photos from right above. If the photos differ, the alignment does not work properly. That is no problem, because only identical photos should pass the algorithm. This means that SURF will not only align the document photo with the chip photo, but also check if there is a difference be- tween the chip photo and the document photo. For photos taken from right above, the results show that SURF performs well, but not for pho- tos taken at an angle. To avoid to have a small angle, it might be better to scan the document photo instead of photographing. The document will then always be from right above. Also when scanning the document, the glare over the doc- ument photos of the residence permits would be less visible. This will make the printed photo
With increase in growing population, the number of blind people across the world is set to triple from about 36 million to 115 million by 2050. The uniqueness in characters were identified on physiological characteristics and behavioral characteristics. Uniqueness in signature, handwriting and voice comes under behavioral characteristics while face, iris, retina impression falls within physiological characteristics. We know that hand writing differs from person to person and even with respect to time which the person write. Hence a well- defined trained algorithm must be implemented to recognize the hand written documents. The efficiency of the system is 80-90% in order to recognize the slightly varied hand written documents. The document which has to be converted are captured as a image using USB camera connected with raspberry pi controller. Further the
Whilst the effect of protein fouling can be mitigated by the incorporation of permselective membranes and antifouling coatings, these can often have a negative impact on the sensitivity through reduced diffusion and surface area of the electrodes. Therefore, the ability of a commercially available fullerene–MWCNT mixture to enhance electroanalytical properties of the printed carbon electrodes was explored through the surface modification with 1-10mg/mL of the fullerene-MWCNT mixture cast in Nafion. Nafion concentrations were tested from 0.01-0.5% (total concentration in solution), a final working concentration of 0.05% Nafion was utilised in electrode development [data not shown]. Figure 4A illustrates a general trend of increasing peak height with increasing quantities of fullerene-MWCNT mixture up to 7.5-10mg/mL where the increase plateaus. In addition to the peak magnitude, as this voltammetric pH sensor has the peak position as its analytical measure, peak sharpness is a key consideration to allow for accurate and precise interpretation of the peak potentials. Figure 4B highlights the half height peak width of the oxidation peaks and indicates that 5mg/mL provides a balance of peak magnitude and peak sharpness. These two phenomena are related, and are attributed to an increased thickness of the fullerene-MWCNT film which impacts its conductivity, stability and results in increased background currents, as per MWCNT modified electrodes reported in literature (Guo et al. 2013; Jain and Sharma 2012). Therefore, 5mg/mL appears to be the optimal load of fullerene-MWCNT that can be cast before these negative characteristics are observed.