Detecting Dense Foreground Stripes in Arabic Handwriting for Accurate Baseline Positioning

Size: px
Start display at page:

Download "Detecting Dense Foreground Stripes in Arabic Handwriting for Accurate Baseline Positioning"

Transcription

1 Detecting Dense Foreground Stripes in Arabic Handwriting for Accurate Baseline Positioning Felix Stahlberg Qatar Computing Research Institute, HBKU Doha, Qatar Stephan Vogel Qatar Computing Research Institute, HBKU Doha, Qatar Abstract Since Arabic script has a strong baseline, many state-of-the-art recognition systems for handwritten Arabic make use of baseline-dependent features. For printed Arabic, the baseline can be detected reliably by finding the maximum in the horizontal projection profile or the Hough transformed image. However, the performance of these methods drops significantly on handwritten Arabic. In this work, we present a novel approach to baseline detection in handwritten Arabic which is based on the detection of stripes in the image with dense foreground. Such a stripe usually corresponds to the area between lower and upper baseline. Our method outperforms a previous method by 22.4% relative for the task of finding acceptable baselines in Tunisian town names in the IFN/ENIT database. I. INTRODUCTION Robust recognition of handwritten Arabic usually relies on sophisticated normalization procedures in the preprocessing step. Common preprocessing operations include line thinning, skew correction, and size normalization. For Arabic, many recognition systems also require the extraction of the baselines. Arabic script features two baselines: The lower and the upper baseline. Figure 1 gives an example. The presence or absence of ascenders (above upper baseline) or descenders (below lower baseline) are useful features for recognition systems. Furthermore, concavity and foreground pixel distribution features turned out to be useful when calculated separately for the regions below and above the lower baseline [1]. Therefore, the accuracy of baseline estimation has direct impact on the final performance of many state-of-the-art recognition systems. In this paper, we automatically extract the lower baseline in handwritten Tunisian town names in the IFN/ENIT database [2]. Our method assumes that the zone between both baselines (core zone) usually includes a large fraction of all foreground pixels i.e. is a dense foreground region in the image. It is justified by the fact that ascenders and descenders usually contribute relatively little to all foreground pixels compared to the core zone. This assumption may hold even for scripts other than Arabic which have two baselines (e.g. Latin). Fig. 1. Upper and lower baselines in Arabic script. II. RELATED WORK Since the position of the lower baseline gives valuable hints to the recognition system, much research effort has been made in the past to automatically find the (lower) baseline in an image. Recently, neural nets have been used successfully for baseline estimation in Latin script [3], [4]. The baseline in horizontal printed Arabic text can be estimated reliably using the horizontal projection method. If the orientation of the printed text line is unknown, the maximum in the Hough space [5] is often assumed to correspond to the baseline. Both methods exploit the fact that most Arabic letters contain many pixels along the baseline [1]. However, for short and/or handwritten Arabic words, these approaches are no longer valid. Authors in [6] report that for more than a fifth of the words in the IFN/ENIT database the maximum in the Hough space does not match the baseline position. More accurate baselines can be found based on the approximation of the script skeleton with piecewise linear curves [7], [8]. Other more recent approaches include [9], [10], [11], [12], [13], [14]. A comparative overview of different methods for Arabic can be found in [15]. Unfortunately, the evaluation of the presented methods is often limited to example images but does not include standardized test sets or error measures. Thus, comparing them is difficult. Authors in [6] define the Baseline Error as the difference between detected baseline and ground truth in pixel, normalized by the image width. According the judgement of Arabic native speakers, they classify a baseline error greater than 7 pixels as insufficient. They also present their skeleton-based baseline estimation method that finds insufficient baselines in only 12.5% of the words in the IFN/ENIT database [2]. In this work, we suggest a novel method for baseline estimation that can reduce the fraction of insufficient baselines to 9.7% on the same dataset (22.4% relative improvement). III. BASELINE DETECTION BASED ON DENSE FOREGROUND STRIPES Horizontal projection and Hough space maximization assume that the straight line in the image which accumulates the most foreground pixels corresponds to the lower baseline. In contrast, we relax this assumption and search for stripes in the image that contain a certain fraction of all foreground pixels. We thereby assume that the area between lower and upper baseline (core zone) should contain at least a certain amount of foreground pixels (ɛ). We additionally require that the maximum of the projection profile of the stripe in the

2 stripe direction is in the lower half of the stripe. From all stripes that satisfy both constraints we select one according some optimization criterion (for example the narrowest stripe). We detect the lower baseline at the bottom border of the stripe. Fig. 2 illustrates the basic idea of our method. The highlighted area marks a stripe which includes 20% of the foreground pixels (ɛ = 20%). The bottom border of the stripe fits the lower baseline in Fig. 1 very accurately. In contrast, the maximum in the Hough space (shown as dashed line in Fig. 2) does not correspond to the correct baseline position. Note that the upper border of the stripe does not match the upper baseline, i.e. ɛ = 20% underestimates the number of pixels in the core zone. However, underestimation is not critical since we are only interested in the lower baseline which is placed correctly. A. Formal Algorithm Description Since diacritics and dots are irrelevant for baseline estimation we remove them in the image by detecting small connected components. Let P {0, 1} h w be the matrix of pixel values in the h w image after removing small connected components. We set P y,x = 0 for all background pixels and P y,x = 1 for all foreground pixel. (0, 0) corresponds to the top left image corner. We formalize a stripe as triple (y 1, y 2, m) where y 1 is the intersection of the upper stripe border and the y-axis, y 2 is the intersection of the lower stripe border and the y-axis (y 1 y 2 ), and m is the slope of the stripe. The function count( ) is the number of foreground pixels in the stripe: count : (y 1, y 2, m) w y 2+ m x x=1 y=y 1+ m x { Py,x if y [1, h] 0 otherwise The projection project( ) in m direction to the y-axis can be formulated as follows: (1) project : (y, m) count(y, y, m) (2) Let S be the set of all stripes which satisfy the requirements postulated above: First, the stripe needs to contain at least ɛ R of all foreground pixels: (y 1, y 2, m) S : ɛ count(y 1, y 2, m) w h x=1 y=1 P y,x Second, the maximum of the projection within the stripe must be in the lower half of the stripe: (3) (y 1, y 2, m) S : y 1 + y 2 2 arg max(project(y, m)) (4) y [y 1,y 2] The lower baseline bl = (y 2, m ) is detected at the lower border of the best stripe (y 1, y 2, m ) S. We explore two different optimization criteria: height minimizes the height of the stripe and thereby maximizes its foreground pixel density. height : S arg min y 2 y 1 (5) (y 1,y 2,m ) S min maximizes the minimum in the projection. The motivation behind the min criterion is that all lines within the core zone in slope direction m usually accumulate many foreground pixels. A small value of project(y, m ) within the stripe indicates misplaced stripe boundaries. B. Search min : S arg max (y 1,y 2,m ) S Algorithm 1 search(ɛ, ρ) min project(y, y [y 1,y 2 ] m ) (6) 1: S 2: n ɛ w h x=1 y=1 Py,x {Required number of pixel} 3: for i 0 to ρ do 4: Θ = i ρ {Stripe angle in [ 45, +45 ]} 5: m tan Θ 6: π (project(1, m),..., project(h, m)) 7: for y 1 1 to h do 8: if π y1 > 0 then 9: acc π y1 10: y 2 y 1 11: while acc < n or y 1+y : y 2 y > arg max y [y1,y 2 ] πy do 13: acc acc + π y2 14: end while 15: S S {(y 1, y 2, m)} 16: end if 17: end for 18: end for 19: return criterion(s) {Either height or min.} Our algorithm for finding the best stripe in an image is listed in Alg. 1. The search is challenging since S is an infinite uncountable set: Decreasing y 1 or increasing y 2 of a feasible stripe again results in a feasible stripe, and the slope m R is real-valued. We address the latter challenge by equidistant quantization of the slope angle Θ in the interval (a) Hough transform. (b) Axis aligned projection. Fig. 2. Baseline detection based on dense foreground stripes (ɛ = 0.2, min criterion). Fig. 3. Line representation in our algorithm compared to the polar coordinate representation used by the Hough transformation.

3 [ 45, +45 ]: First, we choose a resolution ρ. The outer loop in Alg. 1 (lines 3-18) tries out ρ different values for Θ. Then, the projection of the image in Θ direction is calculated and stored in the h-dimensional vector π. The vector π is similar to a single column in a Hough transformed image. However, the Hough transformation represents lines with the help of polar coordinates (r, Θ) (Fig. 3(a)). In contrast, π r represents the line in Θ-direction that intersects the y-axis at y = r (Fig. 3(b)). The advantage of the different representation is that we avoid quantization errors for r since only integer values need to be considered. The disadvantage is that we are not able to represent all possible lines (e.g. lines parallel to the y-axis cannot be represented). However, this is not an issue since the baseline cannot be vertical. Then, for each slope m = tan Θ and each possible stripe start y 1, we find the narrowest feasible stripe and add it to S (lines in Alg. 1). This gives us the best stripe according both criteria (height and min) for the given y 1 and m. In line 19, we return the best stripe in S according either to the height or min criterion. The set S has at most h ρ elements and thus can be traversed efficiently. IV. A. Baseline Error Measure EXPERIMENTS Authors in [6] suggest the Baseline Error measure to assess baseline detection methods. It is defined as the area between estimated and ground truth baseline divided by the width of the image w (Fig. 4). Based on human judgement, they classify a baseline error up to 5 pixels as excellent, between 5 and 7 pixels as acceptable, and higher than 7 pixels as insufficient. Our goal is to reduce the number of insufficient baselines. B. The IFN/ENIT Database The IFN/ENIT database [2] is a widely used corpus for Arabic handwriting recognition research.it consists of 26,459 images of 937 handwritten Tunisian town names (sets a to d). Many groups working on baseline estimation for handwritten Arabic use this corpus [15] because the database provides manually verified baseline information. However, baseline estimation on the IFN/ENIT database is challenging [6]: The writing styles of different writers differ largely. Moreover, the words are sometimes very short or the baseline is discontinuous or curved. We address the latter challenges in Sec. IV-E. We show in the next sections that our approach is able to cope with different writing styles and short words. C. Impact of the Parameters ɛ and ρ The central parameter of our method is the ɛ parameter which defines the required fraction of foreground pixels covered by the stripe. Fig. 5 shows the fraction of insufficient Insufficient Baselines Fig % 40 % 35 % 30 % 25 % 20 % 15 % width criterion min criterion 10 % 10 % 20 % 30 % 40 % 50 % 60 % ε Baseline error over ɛ for both optimization criteria width and min. baselines depending on the ɛ parameter for both the height and min criterion. The minima of both curves are not at the same position (ɛ = 25% for min and ɛ = 35% for width): There is no single value for ɛ which optimizes both criteria. The best result is achieved with the width criterion and ɛ = 35% producing no more than 14.5% insufficient baselines. Insufficient Baselines Fig % 80 % 60 % 40 % 20 % 0 % Insufficient baselines Runtime ρ 4 h 3 h 2 h 1 h 0 h Impact of the resolution for the slope angle (ρ parameter). The parameter ρ controls the resolution of the slope angle Θ and has major impact on the runtime. Fig. 6 plots the fraction of insufficient baselines and the runtime over ρ using ɛ = 35% and the height criterion. Our implementation 1 is written in Java and based on OpenCV [16]. Our testing platform was a Kubuntu with Linux kernel on a 8-core Intel R Core TM i7-3635qm processor at 2.40 GHz, HDD and sufficient RAM. The runtime measurements include loading all 26,459 images from the HDD, binarization, diacritic removal and baseline estimation. The runtime of the baseline estimation is linear in ρ. The time complexity of the other steps is low and independent of ρ (see green curve at ρ = 0). The error drops quickly with increasing ρ and stays at the same level for ρ 40. In latter experiments we choose ρ = 130 to eliminate accuracy degradations due to ρ quantization errors. D. Combination of Multiple Baselines We showed in the previous section that the exact value for the resolution ρ is not decisive for the baseline quality as long as it is not too small. However, how can we know the best value for ɛ in advance? Our solution is to combine multiple baselines resulting from different values for ɛ. This approach has two advantages: First, we cover a whole range for ɛ and do not have to commit to a single value. Second, the combination works even better than the optimal individual ɛ value. Fig. 7 demonstrates the improvements through combination. First, we decide a range for the ɛ parameter (visualized in Fig. 7 by the two horizontal axes). Second, we estimate a baseline Runtime Fig. 4. Baseline error measure. 1 Download from

4 (a) Baseline estimation using the height criterion. (b) Baseline estimation using the min criterion. Fig. 7. Fraction of insufficient baselines when multiple baselines are combined. The horizontal axes define the boundaries of the ɛ range. for each multiple of 5% in that range i.e. we sample ɛ with 5% resolution. The combination is done by selecting the baseline with the most horizontal slope (i.e. with the smallest absolute slope value m ). This simple criterion is justified by the fact that the slope angle of most baselines in the IFN/ENIT database is close to 0. Note that the time complexity is not significantly higher than with only a single ɛ value since all baselines can be estimated in a single pass with only minor changes to Alg. 1. As shown in Fig. 7 the fraction of insufficient baselines can be reduced to 10.6% with the height criterion (ɛ-range from 20% to 50%) and to 10.4% with the min criterion (ɛ-range from 15% to 45%). Changing the ɛ-range has only minor impact: For example, all ranges with a lower bound between 15% and 30% and an upper bound between 40% and 60% result in no more than 12% insufficient baselines using the height criterion (Fig. 7(a)). This enables us to use the same ɛ-ranges for a variety of resolutions, writing styles and pen sizes (Sec. IV-E): the exact boundary values are not critical for the accuracy. The error surface for the min criterion is steeper than for the height criterion indicating that min is slightly less tolerant against changes of the ɛ parameter. The best result is achieved when we combine baselines from both optimization criteria. In Fig. 8 the ɛ-range for the height criterion is fixed to [25%, 45%]. Like in Fig. 7(b) the horizontal axes define the upper and lower bound for the ɛ parameter for min. Setting the ɛ-range for min to [20%, 40%] reduces the fraction of insufficient baselines to 9.7%. Tab. I compares our work with other methods in the literature. To the best of our knowledge, the skeleton-based approach by Pechwitz et. al. [6] with 12.5% insufficient baselines is the best in the literature evaluating on the IFN/ENIT database with the baseline error evaluation scheme presented in Sec. IV-A. We are able to improve their results by 22.4% relative to our final result of 9.7% (relative improvement of TABLE I. Method COMPARISON WITH RELATED WORK Related work Insufficient baseline rate Hough space maximization [5], [6] 21.9% Skeleton-based method by Pechwitz et. al. [6] 12.5% This work Without combination (ɛ = 35%, height criterion) 14.5% With combination (ɛ [25%, 45%] for height, 9.7% ɛ [20%, 40%] for min) 55.7% compared to traditional Hough space maximization). E. Curved or Discontinuous Baselines If the baseline is straight, we can apply our approach to a whole line instead of words without modification. If the baseline is curved or discontinuous, our method fails since it tries to fit a straight dense stripe to the image. Authors in [18] suggest a collection of skeleton-based geometric and topological features for curved baseline extraction. In this work, we vertically split the image into smaller segments and apply our baseline estimation method on each segment separately to deal with curved baselines. Our splitting procedure ensures that the width of each segment is greater or equal twice the height and that we do not cut any foreground strokes. Fig. 9 gives an example. The first splitting position s 1 is at the left most vertical line with distance to the left image corner greater than 2 h that does not cross any foreground strokes. Starting from s h we look for the next vertical line that does not cut foreground and assign the next splitting position s 2 to it. We proceed until we reach the right corner of the image. The algorithm description for curved baseline extraction follows: 1) Estimate the baseline from the whole image and rotate it such that the estimated baseline is horizontal. This is a first rough normalization step. 2) Split the image into smaller segments. 3) Estimate the baseline for each of the segments separately and rotate/translate them such that the baseline is horizontal at the center of the image. 4) Concatenate the segments to a single image. Fig. 8. Combining both optimization criteria. The ɛ-range for height is fixed to [25%, 45%]. The ɛ-range for min is defined by the horizontal axes. Fig. 9. Image of curved text lines into smaller segments.

5 Fig. 10. Baseline normalization in images with curved baselines and various writing styles from the KHATT database [17]. The described procedure results in a normalized image with a rectified straight and horizontal baseline at the center. Fig. 10 shows some examples taken from the KHATT database [17]. This corpus consists of line images from handwritten forms filled out by a variety of different writers with different writing styles. The slope of the text often changes multiple times within a single text line. However, our approach is able to normalize the text reliably using the same ɛ-ranges which performed best on the IFN/ENIT database (ɛ [25%, 45%] for height, ɛ [20%, 40%] for min). Since KHATT does not provide ground truth baseline information, we have to restrict the evaluation to these visual inspections and are not able to report Baseline Errors as for the IFN/ENIT database. V. CONCLUSION In this work we studied baseline estimation in handwritten Arabic words. Our method finds stripes in the image that cover areas of dense foreground. Realizing that a large amount of foreground pixels in Arabic script are between the upper and the lower baseline, we place the lower baseline at the bottom border of the best stripe. We studied two optimization criteria: height minimizes the height of the stripe. min maximizes the minimum in the projection on the y-axis in the stripe direction. Our method requires two parameters. The ρ parameter controls the trade-off between accuracy and runtime. The ɛ parameter defines the required fraction of foreground pixels in the stripe. Using the height criterion with ɛ = 35% results in no more than 14.5% insufficient baselines on the IFN/ENIT [2] database consisting of handwritten Tunisian town names. The accuracy can be improved even more by combining a set of baselines resulting from different values for ɛ. Since multiple baselines can be estimated in a single pass, the time complexity is not significantly higher as with only a single value for ɛ. The best result is achieved when combining baselines estimated using ɛ [25%, 45%] for height and ɛ [20%, 40%] for min resulting in 9.7% insufficient baselines i.e. more than 9 of 10 estimated baselines have acceptable quality. In case of curved or discontinuous baselines, we segment the image into smaller parts and apply our baseline estimation method to each segment separately. By rotating and translating each segment according to its estimated baseline, we can rectify the complete baseline to a straight horizontal line. In the future, we plan to use our new baseline estimation algorithm for feature extraction for a recognition system to improve the recognition accuracy. REFERENCES [1] L. Likforman-Sulem, R. A. H. Mohammad, C. Mokbel, F. Menasri, A. Bianne-Bernard, and C. Kermorvant, Features for HMM-based Arabic handwritten word recognition systems, in Guide to OCR for Arabic Scripts, [2] M. Pechwitz, S. S. Maddouri, V. Märgner, N. Ellouze, H. Amiri et al., IFN/ENIT-database of handwritten Arabic words, in CIFED, [3] S. Espana-Boquera, M. J. Castro-Bleda, J. Gorbe-Moya, and F. Zamora- Martinez, Improving offline handwritten text recognition with hybrid HMM/ANN models, Pattern Analysis and Machine Intelligence, IEEE Transactions on, [4] O. Morillot, L. Likforman-Sulem, and E. Grosicki, New baseline correction algorithm for text-line recognition with bidirectional recurrent neural networks, Journal of Electronic Imaging, [5] R. O. Duda and P. E. Hart, Use of the Hough transformation to detect lines and curves in pictures, Communications of the ACM, vol. 15, no. 1, [6] M. Pechwitz, H. El Abed, and V. Märgner, Handwritten Arabic word recognition using the IFN/ENIT-database, in Guide to OCR for Arabic Scripts, [7] M. Pechwitz and V. Märgner, Baseline estimation for Arabic handwritten words, in ICFHR, [8] F. Farooq, V. Govindaraju, and M. Perrone, Pre-processing methods for handwritten Arabic documents, in ICDAR, [9] M. Ziaratban and K. Faez, A novel two-stage algorithm for baseline estimation and correction in Farsi and Arabic handwritten text line, in ICPR, [10] H. Boubaker, M. Kherallah, and A. M. Alimi, New algorithm of straight or curved baseline detection for short Arabic handwritten writing, in ICDAR, [11] H. Boukerma and N. Farah, A novel Arabic baseline estimation algorithm based on sub-words treatment, in ICFHR, [12] M. Blumenstein, C. K. Cheng, and X. Y. Liu, New preprocessing techniques for handwritten word recognition, in VIIP, [13] P. Nagabhushan and A. Alaei, Tracing and straightening the baseline in handwritten Persian/Arabic text-line: A new approach based on painting-technique, International Journal on Computer Science and Engineering, vol. 2, no. 4, [14] T. Abu-Ain, S. N. H. S. Abdullah, B. Bataineh, K. Omar, and A. Abu- Ein, A novel baseline detection method of handwritten Arabic-script documents based on sub-words, in Soft Computing Applications and Intelligent Systems, [15] A. Al-Shatnawi and K. Omar, A comparative study between methods of Arabic baseline detection, in ICEEI, [16] G. Bradski and A. Kaehler, Learning OpenCV: Computer vision with the OpenCV library. O Reilly Media, [17] S. A. Mahmoud, I. Ahmad, M. Alshayeb, W. G. Al-Khatib, M. T. Parvez, G. A. Fink, V. Märgner, and H. E. Abed, KHATT: Arabic offline handwritten text database, in ICFHR, [18] H. Boubaker, M. Kherallah, and A. M. Alimi, New algorithm of straight or curved baseline detection for short Arabic handwritten writing, in ICDAR, 2009.

An Arabic Baseline Estimation Method Based on Feature Points Extraction

An Arabic Baseline Estimation Method Based on Feature Points Extraction , July 5-7, 2017, London, U.K. An Arabic Baseline Estimation Method Based on Feature Points Extraction Arwa AL-Khatatneh, Sakinah Ali Pitchay and Musab Al-qudah Abstract Baseline estimation is an important

More information

Tracing and Straightening the Baseline in Handwritten Persian/Arabic Text-line: A New Approach Based on Painting-technique

Tracing and Straightening the Baseline in Handwritten Persian/Arabic Text-line: A New Approach Based on Painting-technique Tracing and Straightening the Baseline in Handwritten Persian/Arabic Text-line: A New Approach Based on Painting-technique P. Nagabhushan and Alireza Alaei 1,2 Department of Studies in Computer Science,

More information

Novel Sub-Character HMM Models for Arabic Text Recognition

Novel Sub-Character HMM Models for Arabic Text Recognition Published version: http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=6628700 Novel Sub-Character Models for Arabic Text Recognition Irfan Ahmad Information and Computer Science King Fahd University

More information

Bag-of-Features Representations for Offline Handwriting Recognition Applied to Arabic Script

Bag-of-Features Representations for Offline Handwriting Recognition Applied to Arabic Script 2012 International Conference on Frontiers in Handwriting Recognition Bag-of-Features Representations for Offline Handwriting Recognition Applied to Arabic Script Leonard Rothacker, Szilárd Vajda, Gernot

More information

A Novel Approach to Correction of a Skew at Document Level Using an Arabic Script

A Novel Approach to Correction of a Skew at Document Level Using an Arabic Script ISSN:0975-9646 Amani Ali Ahmed Ali et al, / (IJCSIT) International Journal of Computer Science and Information Technologies, Vol. 8 (5), 2017,569-573 A Novel Approach to Correction of a Skew at Document

More information

ICDAR2015 Writer Identification Competition using KHATT, AHTID/MW and IBHC Databases

ICDAR2015 Writer Identification Competition using KHATT, AHTID/MW and IBHC Databases ICDAR2015 Writer Identification Competition using KHATT, AHTID/MW and IBHC Databases Handwriting is considered to be one of the commonly used modality to identify persons in commercial, governmental and

More information

Word Slant Estimation using Non-Horizontal Character Parts and Core-Region Information

Word Slant Estimation using Non-Horizontal Character Parts and Core-Region Information 2012 10th IAPR International Workshop on Document Analysis Systems Word Slant using Non-Horizontal Character Parts and Core-Region Information A. Papandreou and B. Gatos Computational Intelligence Laboratory,

More information

ICDAR 2009 Arabic Handwriting Recognition Competition

ICDAR 2009 Arabic Handwriting Recognition Competition 2009 10th International Conference on Document Analysis and Recognition ICDAR 2009 Arabic Handwriting Recognition Competition Volker Märgner, Haikal El Abed Technische Universitaet Braunschweig, Institute

More information

A Hybrid NN/HMM Modeling Technique for Online Arabic Handwriting Recognition

A Hybrid NN/HMM Modeling Technique for Online Arabic Handwriting Recognition A Hybrid NN/HMM Modeling Technique for Online Arabic Handwriting Recognition Najiba Tagougui, Houcine Boubaker, Monji Kherallah, Adel M. ALIMI REGIM-LAB: Research Groups on Intelligent Machines Laboratory

More information

Fine Classification of Unconstrained Handwritten Persian/Arabic Numerals by Removing Confusion amongst Similar Classes

Fine Classification of Unconstrained Handwritten Persian/Arabic Numerals by Removing Confusion amongst Similar Classes 2009 10th International Conference on Document Analysis and Recognition Fine Classification of Unconstrained Handwritten Persian/Arabic Numerals by Removing Confusion amongst Similar Classes Alireza Alaei

More information

Invariant Recognition of Hand-Drawn Pictograms Using HMMs with a Rotating Feature Extraction

Invariant Recognition of Hand-Drawn Pictograms Using HMMs with a Rotating Feature Extraction Invariant Recognition of Hand-Drawn Pictograms Using HMMs with a Rotating Feature Extraction Stefan Müller, Gerhard Rigoll, Andreas Kosmala and Denis Mazurenok Department of Computer Science, Faculty of

More information

Keywords Connected Components, Text-Line Extraction, Trained Dataset.

Keywords Connected Components, Text-Line Extraction, Trained Dataset. Volume 4, Issue 11, November 2014 ISSN: 2277 128X International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: www.ijarcsse.com Language Independent

More information

CS 223B Computer Vision Problem Set 3

CS 223B Computer Vision Problem Set 3 CS 223B Computer Vision Problem Set 3 Due: Feb. 22 nd, 2011 1 Probabilistic Recursion for Tracking In this problem you will derive a method for tracking a point of interest through a sequence of images.

More information

A New Approach to Detect and Extract Characters from Off-Line Printed Images and Text

A New Approach to Detect and Extract Characters from Off-Line Printed Images and Text Available online at www.sciencedirect.com Procedia Computer Science 17 (2013 ) 434 440 Information Technology and Quantitative Management (ITQM2013) A New Approach to Detect and Extract Characters from

More information

CS 231A Computer Vision (Fall 2012) Problem Set 3

CS 231A Computer Vision (Fall 2012) Problem Set 3 CS 231A Computer Vision (Fall 2012) Problem Set 3 Due: Nov. 13 th, 2012 (2:15pm) 1 Probabilistic Recursion for Tracking (20 points) In this problem you will derive a method for tracking a point of interest

More information

Initial Results in Offline Arabic Handwriting Recognition Using Large-Scale Geometric Features. Ilya Zavorin, Eugene Borovikov, Mark Turner

Initial Results in Offline Arabic Handwriting Recognition Using Large-Scale Geometric Features. Ilya Zavorin, Eugene Borovikov, Mark Turner Initial Results in Offline Arabic Handwriting Recognition Using Large-Scale Geometric Features Ilya Zavorin, Eugene Borovikov, Mark Turner System Overview Based on large-scale features: robust to handwriting

More information

Text Line Detection for Heterogeneous Documents

Text Line Detection for Heterogeneous Documents Text Line Detection for Heterogeneous Documents Markus Diem, Florian Kleber and Robert Sablatnig Computer Vision Lab Vienna University of Technology Email: diem@caa.tuwien.ac.at Abstract Text line detection

More information

Part-Based Skew Estimation for Mathematical Expressions

Part-Based Skew Estimation for Mathematical Expressions Soma Shiraishi, Yaokai Feng, and Seiichi Uchida shiraishi@human.ait.kyushu-u.ac.jp {fengyk,uchida}@ait.kyushu-u.ac.jp Abstract We propose a novel method for the skew estimation on text images containing

More information

VOL. 3, NO. 7, Juyl 2012 ISSN Journal of Emerging Trends in Computing and Information Sciences CIS Journal. All rights reserved.

VOL. 3, NO. 7, Juyl 2012 ISSN Journal of Emerging Trends in Computing and Information Sciences CIS Journal. All rights reserved. Arabic Hand Written Character Recognition Using Modified Multi-Neural Network Farah Hanna Zawaideh Irbid National University, Computer Information System Department dr.farahzawaideh@inu.edu.jo ABSTRACT

More information

Word-Based Arabic Handwritten Recognition Using SVM Classifier with a Reject Option

Word-Based Arabic Handwritten Recognition Using SVM Classifier with a Reject Option Word-Based Arabic Handwritten Recognition Using SVM Classifier with a Reject Option Bouchra EI qacimy #l, Mounir Ait kerroum* 2, Ahmed Hammouch #3 # Laboratory LRGE, ENSET of Rabat, Mohamed V University

More information

Chapter 3 Image Registration. Chapter 3 Image Registration

Chapter 3 Image Registration. Chapter 3 Image Registration Chapter 3 Image Registration Distributed Algorithms for Introduction (1) Definition: Image Registration Input: 2 images of the same scene but taken from different perspectives Goal: Identify transformation

More information

Comparison of Bernoulli and Gaussian HMMs using a vertical repositioning technique for off-line handwriting recognition

Comparison of Bernoulli and Gaussian HMMs using a vertical repositioning technique for off-line handwriting recognition 2012 International Conference on Frontiers in Handwriting Recognition Comparison of Bernoulli and Gaussian HMMs using a vertical repositioning technique for off-line handwriting recognition Patrick Doetsch,

More information

Handwritten text segmentation using blurred image

Handwritten text segmentation using blurred image Handwritten text segmentation using blurred image Aurélie Lemaitre, Jean Camillerapp, Bertrand Coüasnon To cite this version: Aurélie Lemaitre, Jean Camillerapp, Bertrand Coüasnon. Handwritten text segmentation

More information

Equation to LaTeX. Abhinav Rastogi, Sevy Harris. I. Introduction. Segmentation.

Equation to LaTeX. Abhinav Rastogi, Sevy Harris. I. Introduction. Segmentation. Equation to LaTeX Abhinav Rastogi, Sevy Harris {arastogi,sharris5}@stanford.edu I. Introduction Copying equations from a pdf file to a LaTeX document can be time consuming because there is no easy way

More information

Handwritten Text Recognition

Handwritten Text Recognition Handwritten Text Recognition M.J. Castro-Bleda, S. España-Boquera, F. Zamora-Martínez Universidad Politécnica de Valencia Spain Avignon, 9 December 2010 Text recognition () Avignon Avignon, 9 December

More information

Automatic Recognition and Verification of Handwritten Legal and Courtesy Amounts in English Language Present on Bank Cheques

Automatic Recognition and Verification of Handwritten Legal and Courtesy Amounts in English Language Present on Bank Cheques Automatic Recognition and Verification of Handwritten Legal and Courtesy Amounts in English Language Present on Bank Cheques Ajay K. Talele Department of Electronics Dr..B.A.T.U. Lonere. Sanjay L Nalbalwar

More information

CS 231A Computer Vision (Winter 2014) Problem Set 3

CS 231A Computer Vision (Winter 2014) Problem Set 3 CS 231A Computer Vision (Winter 2014) Problem Set 3 Due: Feb. 18 th, 2015 (11:59pm) 1 Single Object Recognition Via SIFT (45 points) In his 2004 SIFT paper, David Lowe demonstrates impressive object recognition

More information

Pixels. Orientation π. θ π/2 φ. x (i) A (i, j) height. (x, y) y(j)

Pixels. Orientation π. θ π/2 φ. x (i) A (i, j) height. (x, y) y(j) 4th International Conf. on Document Analysis and Recognition, pp.142-146, Ulm, Germany, August 18-20, 1997 Skew and Slant Correction for Document Images Using Gradient Direction Changming Sun Λ CSIRO Math.

More information

LECTURE 6 TEXT PROCESSING

LECTURE 6 TEXT PROCESSING SCIENTIFIC DATA COMPUTING 1 MTAT.08.042 LECTURE 6 TEXT PROCESSING Prepared by: Amnir Hadachi Institute of Computer Science, University of Tartu amnir.hadachi@ut.ee OUTLINE Aims Character Typology OCR systems

More information

DATABASE DEVELOPMENT OF HISTORICAL DOCUMENTS: SKEW DETECTION AND CORRECTION

DATABASE DEVELOPMENT OF HISTORICAL DOCUMENTS: SKEW DETECTION AND CORRECTION DATABASE DEVELOPMENT OF HISTORICAL DOCUMENTS: SKEW DETECTION AND CORRECTION S P Sachin 1, Banumathi K L 2, Vanitha R 3 1 UG, Student of Department of ECE, BIET, Davangere, (India) 2,3 Assistant Professor,

More information

A New Algorithm for Detecting Text Line in Handwritten Documents

A New Algorithm for Detecting Text Line in Handwritten Documents A New Algorithm for Detecting Text Line in Handwritten Documents Yi Li 1, Yefeng Zheng 2, David Doermann 1, and Stefan Jaeger 1 1 Laboratory for Language and Media Processing Institute for Advanced Computer

More information

ROBUST LINE-BASED CALIBRATION OF LENS DISTORTION FROM A SINGLE VIEW

ROBUST LINE-BASED CALIBRATION OF LENS DISTORTION FROM A SINGLE VIEW ROBUST LINE-BASED CALIBRATION OF LENS DISTORTION FROM A SINGLE VIEW Thorsten Thormählen, Hellward Broszio, Ingolf Wassermann thormae@tnt.uni-hannover.de University of Hannover, Information Technology Laboratory,

More information

with Profile's Amplitude Filter

with Profile's Amplitude Filter Arabic Character Segmentation Using Projection-Based Approach with Profile's Amplitude Filter Mahmoud A. A. Mousa Dept. of Computer and Systems Engineering, Zagazig University, Zagazig, Egypt mamosa@zu.edu.eg

More information

HTR Part II: Handwritten Text Recognition

HTR Part II: Handwritten Text Recognition HTR Part II: Handwritten Text Recognition Preprocessing and Feature Extraction for Off-Line Continuous HTR Alejandro H. Toselli & PRHLT-Group Departamento de Sistemas Informáticos y Computación Universidad

More information

Multi-scale Techniques for Document Page Segmentation

Multi-scale Techniques for Document Page Segmentation Multi-scale Techniques for Document Page Segmentation Zhixin Shi and Venu Govindaraju Center of Excellence for Document Analysis and Recognition (CEDAR), State University of New York at Buffalo, Amherst

More information

Restoring Warped Document Image Based on Text Line Correction

Restoring Warped Document Image Based on Text Line Correction Restoring Warped Document Image Based on Text Line Correction * Dep. of Electrical Engineering Tamkang University, New Taipei, Taiwan, R.O.C *Correspondending Author: hsieh@ee.tku.edu.tw Abstract Document

More information

Segmentation of Kannada Handwritten Characters and Recognition Using Twelve Directional Feature Extraction Techniques

Segmentation of Kannada Handwritten Characters and Recognition Using Twelve Directional Feature Extraction Techniques Segmentation of Kannada Handwritten Characters and Recognition Using Twelve Directional Feature Extraction Techniques 1 Lohitha B.J, 2 Y.C Kiran 1 M.Tech. Student Dept. of ISE, Dayananda Sagar College

More information

ABJAD: AN OFF-LINE ARABIC HANDWRITTEN RECOGNITION SYSTEM

ABJAD: AN OFF-LINE ARABIC HANDWRITTEN RECOGNITION SYSTEM ABJAD: AN OFF-LINE ARABIC HANDWRITTEN RECOGNITION SYSTEM RAMZI AHMED HARATY and HICHAM EL-ZABADANI Lebanese American University P.O. Box 13-5053 Chouran Beirut, Lebanon 1102 2801 Phone: 961 1 867621 ext.

More information

Digital Image Processing Fundamentals

Digital Image Processing Fundamentals Ioannis Pitas Digital Image Processing Fundamentals Chapter 7 Shape Description Answers to the Chapter Questions Thessaloniki 1998 Chapter 7: Shape description 7.1 Introduction 1. Why is invariance to

More information

A Comparison of Sequence-Trained Deep Neural Networks and Recurrent Neural Networks Optical Modeling For Handwriting Recognition

A Comparison of Sequence-Trained Deep Neural Networks and Recurrent Neural Networks Optical Modeling For Handwriting Recognition A Comparison of Sequence-Trained Deep Neural Networks and Recurrent Neural Networks Optical Modeling For Handwriting Recognition Théodore Bluche, Hermann Ney, Christopher Kermorvant SLSP 14, Grenoble October

More information

A System to Retrieve Text/Symbols from Color Maps using Connected Component and Skeleton Analysis

A System to Retrieve Text/Symbols from Color Maps using Connected Component and Skeleton Analysis A System to Retrieve Text/Symbols from Color Maps using Connected Component and Skeleton Analysis Partha Pratim Roy 1, Eduard Vazquez 1, Josep Lladós 1, Ramon Baldrich 1, and Umapada Pal 2 1 Computer Vision

More information

Practical Image and Video Processing Using MATLAB

Practical Image and Video Processing Using MATLAB Practical Image and Video Processing Using MATLAB Chapter 18 Feature extraction and representation What will we learn? What is feature extraction and why is it a critical step in most computer vision and

More information

Robust line segmentation for handwritten documents

Robust line segmentation for handwritten documents Robust line segmentation for handwritten documents Kamal Kuzhinjedathu, Harish Srinivasan and Sargur Srihari Center of Excellence for Document Analysis and Recognition (CEDAR) University at Buffalo, State

More information

Segmentation of Characters of Devanagari Script Documents

Segmentation of Characters of Devanagari Script Documents WWJMRD 2017; 3(11): 253-257 www.wwjmrd.com International Journal Peer Reviewed Journal Refereed Journal Indexed Journal UGC Approved Journal Impact Factor MJIF: 4.25 e-issn: 2454-6615 Manpreet Kaur Research

More information

Text Line Detection in Corrupted and Damaged Historical Manuscripts

Text Line Detection in Corrupted and Damaged Historical Manuscripts Text Line Detection in Corrupted and Damaged Historical Manuscripts Irina Rabaev, Ofer Biller, Jihad El-Sana, Klara Kedem Department of Computer Science Ben-Gurion University Beer-Sheva, Israel rabaev,billero,el-sana,klara@cs.bgu.ac.il

More information

On-line handwriting recognition using Chain Code representation

On-line handwriting recognition using Chain Code representation On-line handwriting recognition using Chain Code representation Final project by Michal Shemesh shemeshm at cs dot bgu dot ac dot il Introduction Background When one preparing a first draft, concentrating

More information

Segmentation of Arabic handwritten text to lines

Segmentation of Arabic handwritten text to lines Available online at www.sciencedirect.com ScienceDirect Procedia Computer Science 73 (2015 ) 115 121 The International Conference on Advanced Wireless, Information, and Communication Technologies (AWICT

More information

Layout Segmentation of Scanned Newspaper Documents

Layout Segmentation of Scanned Newspaper Documents , pp-05-10 Layout Segmentation of Scanned Newspaper Documents A.Bandyopadhyay, A. Ganguly and U.Pal CVPR Unit, Indian Statistical Institute 203 B T Road, Kolkata, India. Abstract: Layout segmentation algorithms

More information

Skew Detection and Correction Technique for Arabic Document Images Based on Centre of Gravity

Skew Detection and Correction Technique for Arabic Document Images Based on Centre of Gravity Journal of Computer Science 5 (5): 363-368, 2009 ISSN 1549-3636 2009 Science Publications Skew Detection and Correction Technique for Arabic Document Images Based on Centre of Gravity Atallah Mahmoud Al-Shatnawi

More information

Character Recognition

Character Recognition Character Recognition 5.1 INTRODUCTION Recognition is one of the important steps in image processing. There are different methods such as Histogram method, Hough transformation, Neural computing approaches

More information

A Review of Skew Detection Techniques for Document

A Review of Skew Detection Techniques for Document 2015 17th UKSIM-AMSS International Conference on Modelling and Simulation A Review of Skew Detection Techniques for Document Arwa AL-Khatatneh, Sakinah Ali Pitchay Faculty of Science and Technology Universiti

More information

Line Detection and Segmentation in Historical Church Registers

Line Detection and Segmentation in Historical Church Registers Line Detection and Segmentation in Historical Church Registers Markus Feldbach and Klaus D. Tönnies Computer Vision Group Dement of Simulation and Graphics Otto-von-Guericke University of Magdeburg P.O.

More information

EE368 Project Report CD Cover Recognition Using Modified SIFT Algorithm

EE368 Project Report CD Cover Recognition Using Modified SIFT Algorithm EE368 Project Report CD Cover Recognition Using Modified SIFT Algorithm Group 1: Mina A. Makar Stanford University mamakar@stanford.edu Abstract In this report, we investigate the application of the Scale-Invariant

More information

FUZZY BASED PREPROCESSING USING FUSION OF ONLINE AND OFFLINE TRAIT FOR ONLINE URDU SCRIPT BASED LANGUAGES CHARACTER RECOGNITION

FUZZY BASED PREPROCESSING USING FUSION OF ONLINE AND OFFLINE TRAIT FOR ONLINE URDU SCRIPT BASED LANGUAGES CHARACTER RECOGNITION International Journal of Innovative Computing, Information and Control ICIC International c 2012 ISSN 1349-4198 Volume 8, Number 5(A), May 2012 pp. 3149 3161 FUZZY BASED PREPROCESSING USING FUSION OF ONLINE

More information

Auto-Digitizer for Fast Graph-to-Data Conversion

Auto-Digitizer for Fast Graph-to-Data Conversion Auto-Digitizer for Fast Graph-to-Data Conversion EE 368 Final Project Report, Winter 2018 Deepti Sanjay Mahajan dmahaj@stanford.edu Sarah Pao Radzihovsky sradzi13@stanford.edu Ching-Hua (Fiona) Wang chwang9@stanford.edu

More information

Recognition-based Segmentation of Nom Characters from Body Text Regions of Stele Images Using Area Voronoi Diagram

Recognition-based Segmentation of Nom Characters from Body Text Regions of Stele Images Using Area Voronoi Diagram Author manuscript, published in "International Conference on Computer Analysis of Images and Patterns - CAIP'2009 5702 (2009) 205-212" DOI : 10.1007/978-3-642-03767-2 Recognition-based Segmentation of

More information

HOUGH TRANSFORM. Plan for today. Introduction to HT. An image with linear structures. INF 4300 Digital Image Analysis

HOUGH TRANSFORM. Plan for today. Introduction to HT. An image with linear structures. INF 4300 Digital Image Analysis INF 4300 Digital Image Analysis HOUGH TRANSFORM Fritz Albregtsen 14.09.2011 Plan for today This lecture goes more in detail than G&W 10.2! Introduction to Hough transform Using gradient information to

More information

IDIAP. Martigny - Valais - Suisse IDIAP

IDIAP. Martigny - Valais - Suisse IDIAP R E S E A R C H R E P O R T IDIAP Martigny - Valais - Suisse Off-Line Cursive Script Recognition Based on Continuous Density HMM Alessandro Vinciarelli a IDIAP RR 99-25 Juergen Luettin a IDIAP December

More information

Solving Word Jumbles

Solving Word Jumbles Solving Word Jumbles Debabrata Sengupta, Abhishek Sharma Department of Electrical Engineering, Stanford University { dsgupta, abhisheksharma }@stanford.edu Abstract In this report we propose an algorithm

More information

A Feature based on Encoding the Relative Position of a Point in the Character for Online Handwritten Character Recognition

A Feature based on Encoding the Relative Position of a Point in the Character for Online Handwritten Character Recognition A Feature based on Encoding the Relative Position of a Point in the Character for Online Handwritten Character Recognition Dinesh Mandalapu, Sridhar Murali Krishna HP Laboratories India HPL-2007-109 July

More information

Keywords OCR, NoteMate, Text Mining, Template Matching, Android.

Keywords OCR, NoteMate, Text Mining, Template Matching, Android. Volume 5, Issue 3, March 2015 ISSN: 2277 128X International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: www.ijarcsse.com NoteMate - A Note

More information

Mono-font Cursive Arabic Text Recognition Using Speech Recognition System

Mono-font Cursive Arabic Text Recognition Using Speech Recognition System Mono-font Cursive Arabic Text Recognition Using Speech Recognition System M.S. Khorsheed Computer & Electronics Research Institute, King AbdulAziz City for Science and Technology (KACST) PO Box 6086, Riyadh

More information

Printed Arabic Text Recognition using Linear and Nonlinear Regression

Printed Arabic Text Recognition using Linear and Nonlinear Regression Printed Arabic Text Recognition using Linear and Nonlinear Regression Ashraf A. Shahin 1,2 1 College of Computer and Information Sciences, Al Imam Mohammad Ibn Saud Islamic University (IMSIU) Riyadh, Kingdom

More information

Identifying Layout Classes for Mathematical Symbols Using Layout Context

Identifying Layout Classes for Mathematical Symbols Using Layout Context Rochester Institute of Technology RIT Scholar Works Articles 2009 Identifying Layout Classes for Mathematical Symbols Using Layout Context Ling Ouyang Rochester Institute of Technology Richard Zanibbi

More information

Non-uniform Slant Correction using Generalized Projections

Non-uniform Slant Correction using Generalized Projections I J C T A, 9(17) 2016, pp. 8489-8497 International Science Press Non-uniform Slant Correction using Generalized Projections A. M. Hafiz * and G. M. Bhat * ABSTRACT Slant Correction is an important component

More information

A System for Joining and Recognition of Broken Bangla Numerals for Indian Postal Automation

A System for Joining and Recognition of Broken Bangla Numerals for Indian Postal Automation A System for Joining and Recognition of Broken Bangla Numerals for Indian Postal Automation K. Roy, U. Pal and B. B. Chaudhuri CVPR Unit; Indian Statistical Institute, Kolkata-108; India umapada@isical.ac.in

More information

Keyword Spotting in Document Images through Word Shape Coding

Keyword Spotting in Document Images through Word Shape Coding 2009 10th International Conference on Document Analysis and Recognition Keyword Spotting in Document Images through Word Shape Coding Shuyong Bai, Linlin Li and Chew Lim Tan School of Computing, National

More information

Scanning Neural Network for Text Line Recognition

Scanning Neural Network for Text Line Recognition 2012 10th IAPR International Workshop on Document Analysis Systems Scanning Neural Network for Text Line Recognition Sheikh Faisal Rashid, Faisal Shafait and Thomas M. Breuel Department of Computer Science

More information

Text lines and snippets extraction for 19th century handwriting documents layout analysis

Text lines and snippets extraction for 19th century handwriting documents layout analysis Author manuscript, published in "2009 10th International Conference on Document Analysis and Recognition, Barcelona : Spain (2009)" Text lines and snippets extraction for 19th century handwriting documents

More information

A New Method for Skeleton Pruning

A New Method for Skeleton Pruning A New Method for Skeleton Pruning Laura Alejandra Pinilla-Buitrago, José Fco. Martínez-Trinidad, and J.A. Carrasco-Ochoa Instituto Nacional de Astrofísica, Óptica y Electrónica Departamento de Ciencias

More information

Database for Arabic Printed Text Recognition Research

Database for Arabic Printed Text Recognition Research Database for Arabic Printed Text Recognition Research Faten Kallel Jaiem 1, Slim Kanoun 1, Maher Khemakhem 1, Haikal El Abed 2, and Jihain Kardoun 3 1 MIRACL laboratory, ISIMS, University of Sfax, Tunisia

More information

A Segmentation Free Approach to Arabic and Urdu OCR

A Segmentation Free Approach to Arabic and Urdu OCR A Segmentation Free Approach to Arabic and Urdu OCR Nazly Sabbour 1 and Faisal Shafait 2 1 Department of Computer Science, German University in Cairo (GUC), Cairo, Egypt; 2 German Research Center for Artificial

More information

HMM-Based Handwritten Amharic Word Recognition with Feature Concatenation

HMM-Based Handwritten Amharic Word Recognition with Feature Concatenation 009 10th International Conference on Document Analysis and Recognition HMM-Based Handwritten Amharic Word Recognition with Feature Concatenation Yaregal Assabie and Josef Bigun School of Information Science,

More information

Slant Correction using Histograms

Slant Correction using Histograms Slant Correction using Histograms Frank de Zeeuw Bachelor s Thesis in Artificial Intelligence Supervised by Axel Brink & Tijn van der Zant July 12, 2006 Abstract Slant is one of the characteristics that

More information

Skew Detection Technique for Binary Document Images based on Hough Transform

Skew Detection Technique for Binary Document Images based on Hough Transform Skew Detection Technique for Binary Document Images based on Hough Transform Manjunath Aradhya V N*, Hemantha Kumar G, and Shivakumara P Abstract Document image processing has become an increasingly important

More information

Segmentation of Bangla Handwritten Text

Segmentation of Bangla Handwritten Text Thesis Report Segmentation of Bangla Handwritten Text Submitted By: Sabbir Sadik ID:09301027 Md. Numan Sarwar ID: 09201027 CSE Department BRAC University Supervisor: Professor Dr. Mumit Khan Date: 13 th

More information

Colour Segmentation-based Computation of Dense Optical Flow with Application to Video Object Segmentation

Colour Segmentation-based Computation of Dense Optical Flow with Application to Video Object Segmentation ÖGAI Journal 24/1 11 Colour Segmentation-based Computation of Dense Optical Flow with Application to Video Object Segmentation Michael Bleyer, Margrit Gelautz, Christoph Rhemann Vienna University of Technology

More information

International Journal of Signal Processing, Image Processing and Pattern Recognition Vol.9, No.2 (2016) Figure 1. General Concept of Skeletonization

International Journal of Signal Processing, Image Processing and Pattern Recognition Vol.9, No.2 (2016) Figure 1. General Concept of Skeletonization Vol.9, No.2 (216), pp.4-58 http://dx.doi.org/1.1425/ijsip.216.9.2.5 Skeleton Generation for Digital Images Based on Performance Evaluation Parameters Prof. Gulshan Goyal 1 and Ritika Luthra 2 1 Associate

More information

On-Line Recognition of Mathematical Expressions Using Automatic Rewriting Method

On-Line Recognition of Mathematical Expressions Using Automatic Rewriting Method On-Line Recognition of Mathematical Expressions Using Automatic Rewriting Method T. Kanahori 1, K. Tabata 1, W. Cong 2, F.Tamari 2, and M. Suzuki 1 1 Graduate School of Mathematics, Kyushu University 36,

More information

CHAPTER 1 INTRODUCTION

CHAPTER 1 INTRODUCTION CHAPTER 1 INTRODUCTION 1.1 Introduction Pattern recognition is a set of mathematical, statistical and heuristic techniques used in executing `man-like' tasks on computers. Pattern recognition plays an

More information

Isolated Curved Gurmukhi Character Recognition Using Projection of Gradient

Isolated Curved Gurmukhi Character Recognition Using Projection of Gradient International Journal of Computational Intelligence Research ISSN 0973-1873 Volume 13, Number 6 (2017), pp. 1387-1396 Research India Publications http://www.ripublication.com Isolated Curved Gurmukhi Character

More information

Automated Digital Conversion of Hand-Drawn Plots

Automated Digital Conversion of Hand-Drawn Plots Automated Digital Conversion of Hand-Drawn Plots Ruo Yu Gu Department of Electrical Engineering Stanford University Palo Alto, U.S.A. ruoyugu@stanford.edu Abstract An algorithm has been developed using

More information

Representations and Metrics for Off-Line Handwriting Segmentation

Representations and Metrics for Off-Line Handwriting Segmentation Representations and Metrics for Off-Line Handwriting Segmentation Thomas M. Breuel PARC Palo Alto, CA, USA tbreuel@parc.com Abstract Segmentation is a key step in many off-line handwriting recognition

More information

SKEW DETECTION AND CORRECTION

SKEW DETECTION AND CORRECTION CHAPTER 3 SKEW DETECTION AND CORRECTION When the documents are scanned through high speed scanners, some amount of tilt is unavoidable either due to manual feed or auto feed. The tilt angle induced during

More information

Handwritten Arabic Digits Recognition Using Bézier Curves

Handwritten Arabic Digits Recognition Using Bézier Curves www.ijcsi.org 57 Handwritten Arabic Digits Recognition Using Bézier Curves Aissa Kerkour El Miad and Azzeddine Mazroui University Mohammed First, Faculty of Sciences, Oujda, Morocco Abstract In this paper

More information

A Simplistic Way of Feature Extraction Directed towards a Better Recognition Accuracy

A Simplistic Way of Feature Extraction Directed towards a Better Recognition Accuracy International Journal of Engineering Research and Development e-issn: 2278-067X, p-issn: 2278-800X, www.ijerd.com Volume 3, Issue 7 (September 2012), PP. 43-49 A Simplistic Way of Feature Extraction Directed

More information

An Extension of the Multicut L-Shaped Method. INEN Large-Scale Stochastic Optimization Semester project. Svyatoslav Trukhanov

An Extension of the Multicut L-Shaped Method. INEN Large-Scale Stochastic Optimization Semester project. Svyatoslav Trukhanov An Extension of the Multicut L-Shaped Method INEN 698 - Large-Scale Stochastic Optimization Semester project Svyatoslav Trukhanov December 13, 2005 1 Contents 1 Introduction and Literature Review 3 2 Formal

More information

Classification of objects from Video Data (Group 30)

Classification of objects from Video Data (Group 30) Classification of objects from Video Data (Group 30) Sheallika Singh 12665 Vibhuti Mahajan 12792 Aahitagni Mukherjee 12001 M Arvind 12385 1 Motivation Video surveillance has been employed for a long time

More information

A Novel Approach for Rotation Free Online Handwritten Chinese Character Recognition +

A Novel Approach for Rotation Free Online Handwritten Chinese Character Recognition + 2009 0th International Conference on Document Analysis and Recognition A Novel Approach for Rotation Free Online andwritten Chinese Character Recognition + Shengming uang, Lianwen Jin* and Jin Lv School

More information

A semi-incremental recognition method for on-line handwritten Japanese text

A semi-incremental recognition method for on-line handwritten Japanese text 2013 12th International Conference on Document Analysis and Recognition A semi-incremental recognition method for on-line handwritten Japanese text Cuong Tuan Nguyen, Bilan Zhu and Masaki Nakagawa Department

More information

A Document Image Analysis System on Parallel Processors

A Document Image Analysis System on Parallel Processors A Document Image Analysis System on Parallel Processors Shamik Sural, CMC Ltd. 28 Camac Street, Calcutta 700 016, India. P.K.Das, Dept. of CSE. Jadavpur University, Calcutta 700 032, India. Abstract This

More information

EE368 Project: Visual Code Marker Detection

EE368 Project: Visual Code Marker Detection EE368 Project: Visual Code Marker Detection Kahye Song Group Number: 42 Email: kahye@stanford.edu Abstract A visual marker detection algorithm has been implemented and tested with twelve training images.

More information

2 ATTILA FAZEKAS The tracking model of the robot car The schematic picture of the robot car can be seen on Fig.1. Figure 1. The main controlling task

2 ATTILA FAZEKAS The tracking model of the robot car The schematic picture of the robot car can be seen on Fig.1. Figure 1. The main controlling task NEW OPTICAL TRACKING METHODS FOR ROBOT CARS Attila Fazekas Debrecen Abstract. In this paper new methods are proposed for intelligent optical tracking of robot cars the important tools of CIM (Computer

More information

Skew Detection and Correction of Document Image using Hough Transform Method

Skew Detection and Correction of Document Image using Hough Transform Method Skew Detection and Correction of Document Image using Hough Transform Method [1] Neerugatti Varipally Vishwanath, [2] Dr.T. Pearson, [3] K.Chaitanya, [4] MG JaswanthSagar, [5] M.Rupesh [1] Asst.Professor,

More information

Word Matching of handwritten scripts

Word Matching of handwritten scripts Word Matching of handwritten scripts Seminar about ancient document analysis Introduction Contour extraction Contour matching Other methods Conclusion Questions Problem Text recognition in handwritten

More information

Hierarchical Representation of 2-D Shapes using Convex Polygons: a Contour-Based Approach

Hierarchical Representation of 2-D Shapes using Convex Polygons: a Contour-Based Approach Hierarchical Representation of 2-D Shapes using Convex Polygons: a Contour-Based Approach O. El Badawy, M. S. Kamel Pattern Analysis and Machine Intelligence Laboratory, Department of Systems Design Engineering,

More information

RESEARCH ON OPTIMIZATION OF IMAGE USING SKELETONIZATION TECHNIQUE WITH ADVANCED ALGORITHM

RESEARCH ON OPTIMIZATION OF IMAGE USING SKELETONIZATION TECHNIQUE WITH ADVANCED ALGORITHM 881 RESEARCH ON OPTIMIZATION OF IMAGE USING SKELETONIZATION TECHNIQUE WITH ADVANCED ALGORITHM Sarita Jain 1 Sumit Rana 2 Department of CSE 1 Department of CSE 2 Geeta Engineering College 1, Panipat, India

More information

Image Normalization and Preprocessing for Gujarati Character Recognition

Image Normalization and Preprocessing for Gujarati Character Recognition 334 Image Normalization and Preprocessing for Gujarati Character Recognition Jayashree Rajesh Prasad Department of Computer Engineering, Sinhgad College of Engineering, University of Pune, Pune, Mahaashtra

More information

Skew Detection for Complex Document Images Using Fuzzy Runlength

Skew Detection for Complex Document Images Using Fuzzy Runlength Skew Detection for Complex Document Images Using Fuzzy Runlength Zhixin Shi and Venu Govindaraju Center of Excellence for Document Analysis and Recognition(CEDAR) State University of New York at Buffalo,

More information

Automatic removal of crossed-out handwritten text and the effect on writer verification and identification

Automatic removal of crossed-out handwritten text and the effect on writer verification and identification Automatic removal of crossed-out handwritten text and the effect on writer verification and identification (The original paper was published in: Proc. of Document Recognition and Retrieval XV, IS&T/SPIE

More information