A Survey paper on skew detection of offline handwritten character recognition system

Transcription

1 ABSTRACT A Survey paper on skew detection of offline handwritten character recognition system Bishakha Jain 1,Mrinaljit Borah 2 1 Sikkim Manipal Institute of Technology, Majhitar,Sikkim. 2 Jorhat Engineering College, Garmur,Jorhat,Assam. The paper emphasizes on the progress achieved in the field of skew detection of offline character recognition system. Character recognition is the ability of a computer to receive and interpret intelligible handwritten input from sources such as paper documents, photographs, touch-screens and other devices. The paper gives a brief view about the two types of character recognition technique. The paper also gives a brief introduction about the skew detection process and some of the methods implemented so far. Keywords Character recognition, preprocessing, skew, skew estimation, skew angle, scanned documents [1] INTRODUCTION Character recognition is the ability of the computer to recognize the text or documents that can be fed into the system via input device or online on the basis of how the text is entered. They are broadly classified into two sub-divisions: Online character recognition Offline character recognition The paper emphasizes mainly on offline character recognition technique. Online Character Recognition It recognizes character on the basis of the direction of the motion while writing character. This method is generally available on touchpad; touch screen cell phones etc [1]. In this process co-ordinate information of strokes is available with timing information which makes it easier. Also character recognition can be processed from both online and offline perspective but not vice-versa.

2 Offline Character Recognition Off-line handwriting recognition involves the automatic conversion of text in an image into letter codes which are usable within computer and text-processing applications [1]. The data obtained by this form is regarded as a static representation of handwriting. It is comparatively difficult, as different people have different handwriting styles. It does not have the advantage of recognizing direction of the movements while writing the character. [2] DATA PREPROCESSING After document scanning, sequences of preprocessing steps are applied to the data in order to put them in a suitable format ready for feature extraction. A generic character recognition system has different stages like noise removal, skew detection and correction, segmentation, feature extraction and classification. Results of the initial stages can affect the performance of the subsequent stages in the Optical Character Recognition process. To make the results of the subsequent stages more accurate, the skew detection and correction,the slant detection and correction and segmentation play an important role. [3] SMOOTHING AND NOISE REMOVAL Smoothing operations in gray level document images are used for blurring and for noise reduction. Blurring is used for removal of small details from an image. In binary (black and white) document images, smoothing operations straightens the edges of the characters, for example, it fills the small gaps and also removes the small bumps in the edges (contours) of the characters. Filtering also helps in smoothing and noise removal. Filtering is a neighborhood operation, in which the value of any given pixel in the output image is determined by applying some algorithm to the values of the pixels in the neighborhood of the corresponding input pixel. There are two types of filtering approaches: linear and nonlinear. A pixel s neighborhood is a set of pixels, defined by their locations relative to that pixel. [4] SLANT DETECTION AND CORRECTION The slant of the scanned document image specifies the deviation of its text lines from the vertical axis. Slant correction is an important step in the preprocessing stage of both handwritten words and numeral strings recognition. The general purpose of slant correction is to reduce the variation of the script and specifically to improve the quality of the segmentation candidates of the words or numerals in a string, which in turn can yield higher recognition accuracy. [5] SKEW DETECTION AND CORRECTION Skew detection of scanned document images is one of the most important stages of its recognition preprocessing. The skew of the scanned document image specifies the deviation of its text lines from the horizontal axis. The skew of the document image can be a global (all document s blocks have the same orientation), multiple (document s blocks have a different orientation) or nonuniform (multiple

3 orientation in a text line) [2]. Hence, a criterion function for skew detection is obtained. After that using the function defined, the skew is estimated. The angle corresponding to the maximum or the minimum value of the function is usually considered as the skew. So, the maximum or the minimum of the criterion function is achieved [3]. Until now, many methods for skew detection of scanned document images have been proposed. These methods include projection profile analysis, Hough transform, nearest neighbor clustering, morphology, cross-correlation etc. In the following section, brief descriptions of these methods are given. [5.1] Projection Profile Analysis A straightforward solution to determining the skew angle of a document image uses a horizontal projection profile. This is a one-dimensional array with a number of locations equal to the number of rows in an image. Each location in the projection profile stores a count of the number of black pixels in the corresponding row of the image. This histogram has the maximum amplitude and frequency when the text in the image is skewed at zero degrees since the number of collinear black pixels is maximized in this condition [4]. In 1988, Postl used the horizontal projection profile for skew detection by using the sum of squared differences between adjacent elements of the projection profile as the criterion function [5]. In 1993, Bloomberg and Kopec employed the variance of the number of black pixels in each row as the criterion function values. They added a feature selection sub-step before calculating the projection profile and downsampling the image, to reduce the computational burden of this method [6]. In 2011, Papandreou and Gatos used vertical projection profile for skew detection. They considered the sum of squares of the projection profile elements as the value of the criterion function. This method is robust to noise and warp of the image. This method also works well for the languages where most of their letters include at least one vertical line, such as languages with Latin alphabets [7]. [5.2] Hough Transform Hough transform technique may be applied on the upper envelopes for skew estimation, but this is slow process. Sometimes digitized image may be skewed and for this situation skew correction is necessary to make text lines horizontal [8]. In 1989, Srihari used Hough Transform for skew detection. He considered the rate of change in the accumulator array values for each angle θ as a criterion function of the angle [9]. In 1996, Yu and Jain introduced the hierarchical Hough Transform approach [10]. In order to reduce the dimension of the search space, they performed a feature selection step where the range of angle θ is divided into large distance and the angle is obtained using Hough Transform. The interval around that angle is the desired output. After this stage, the Hough Transform is performed on the new search space. This new range is divided into small intervals and the final angle is obtained using Hough Transform. In 2011, Epshtein presented a Hough Transform based method that instead of estimating the direction of text lines; it estimated the direction of the white space between lines [11]. In 2012, Kumar and Singh applied Hough Transform on a set of pixels. Thus, they reduced the running time and maintained the accuracy of the method. They divided the spectrum of the Hough Transform

4 space (the skew can be between 0 and 45 degrees) to the distance by one tenth. Then, they selected the section including the final skew. Finally, they divided that section to the distance by one tenth and searched that for final skew [12]. [5.3] Nearest neighbor clustering This method uses vectors connecting the image connected components to the nearest neighbors are used as features for skew detection. In 1986, Hashizume proposed this method for skew detection by using the vector angle histogram as the criterion function [13]. In 1993, O'Gorman developed Hashizume s methodby using the vectors connecting each component to its k nearest neighbor [14]. In 2003, Lu and Tan proposed k-nearest-neighbor chain with language independent for improved accuracy of the nearest neighbor clustering method [15], [16]. [5.4] Morphological approach The Mathematical Morphology consists in comparing an unknown picture X with a pattern B, perfectly defined in terms of shape, size and grayscale, named structuring element [17]. Different methods based on morphology are proposed for skew detection [18]-[21]. For the first time, in 1994, Chen and Haralick used recursive morphological transforms to detect skew [18]. [5.5] Cross-correlation method This method calculate the document skew by finding the amount of vertical shifts needed to maximize the cross-correlation between pairs of narrow vertical columns of the documents. In 1997, Chundhuri, instead of finding the correlation for the entire image, calculated it for small randomly chosen areas. For this purpose, he used Monte-Carlo sampling to determine the number of regions to calculate the correlation. He considered the median maximum cross-correlation as the criterion to obtain the skew [22]. In 1999, Chen, like Chundhuri s method, calculated correlations for a randomly selected small area, but added the verification stage to determine the suitability of the area. Also, in order to solve the second problem of Yan s method, he calculated horizontal and vertical crosscorrelations of the image [23]. [6] CONCLUSION A lot of work has already been proposed and has been going on in the field of skew detection. An overview of some of the methods was discussed in this paper. Along with the overview, references to the detailed description of the processes are given. There has been massaive advancement in this field and more work is still under research. We hope this research based survey paper will prove to be a boon for the newcomers in this field to get a general overview on the skew detection process.

5 REFERENCES [1] Vedgupt Saraf, D.S. Rao, Devnagari Script Character Recognition Using Genetic Algorithm for Get Better Efficiency, International Journal of Soft Computing and Engineering (IJSCE) ISSN: , Volume-2, Issue-4, April 2013 [2]O. Okun, M. Pietikäinen, and J. Sauvola, Document skew estimation without angle range restriction, International Journal on Document Analysis and Recognition, vol. 2, pp , [3] Sepideh Barekat Rezaei, Abdolhossein Sarrafzadeh, and Jamshid Shanbehzadeh, Skew Detection of Scanned Document Images, Proceedings of the International MultiConference of Engineers and Computer Scientists 2013 Vol I, IMECS 2013, March 13-15, 2013, Hong Kong. [4] JONATHAN J. HULL, DOCUMENT IMAGE SKEW DETECTION: SURVEY AND ANNOTATED BIBLIOGRAPHY, Document Analysis Systems Ii,pp 40-64, 1998 [5] W. Postl, Method for automatic correction of character skew in the acquisition of a text original in the form of digital scan results, United States Patent, [6] D. S. Bloomberg and G. E. Kopec, Method and apparatus for identification and correction of document skew, United States Patent, [7] A. Papandreou and B. Gatos, A Novel Skew Detection Technique Based on Vertical Projections, in Document Analysis and Recognition (ICDAR), 2011 International Conference on Document Analysis and Recognition, 2011, pp [8] Farjana Yeasmin Omee, Shiam Shabbir Himel and Md. Abu Naser Bikas, A Complete Workflow for Development of Bangla OCR, Int. Journal of Computer Applications by Foundation of Computer Science, 21(9):1-6, May [9] S. N. Srihari and V. Govindaraju, Analysis of textual images using the Hough transform, Machine Vision and Applications, vol. 2, pp , [10] B. Yu and A. K. Jain, A robust and fast skew detection algorithm for generic documents, Pattern Recognition, vol. 29, pp , [11] B. Epshtein, Determining Document Skew Using Inter-line Spaces, in Document Analysis and Recognition (ICDAR), 2011 International Conference, 2011, pp [12] D. Kumar and D. Singh, Modified Approach of Hough Transform for Skew Detection and Correction in Documented Images, International Journal of Research in Computer Science, vol. 2, pp , [13] A. Hashizume, P.-S. Yeh, and A. Rosenfeld, A method of detecting the orientation of aligned components, Pattern Recognition Letters, vol. 4, pp , [14] L. O'Gorman, The document spectrum for page layout analysis, 4Pattern Analysis and Machine Intelligence, IEEE Transactions on, vol. 15, pp , [15] Y. Lu and C. L. Tan, A nearest-neighbor chain based approach to skew estimation in document images, Pattern Recognition Letters, vol. 24, pp , [16] Y. Lu and C. L. Tan, Improved Nearest Neighbor Based Approach to Accurate Document Skew Estimation, presented at the Proceedings of the Seventh International Conference on Document Analysis and Recognition, Vol. 1, 2003

6 [17] MARISA E. MORITA 1 - FL AVIO BORTOLOZZI 2 - JACQUES FACON 3 ROBERT SABOURIN, Morphological approach of handwritten word skew correction, Anais do XI SIBGRAPI, outubro de [18] S. Chen and R. M. Haralick, An automatic algorithm for text skew estimation in document images using recursive morphological transforms, in Image Processing, Proceedings. ICIP-94., IEEE International Conference, 1994, pp vol.1. [19] A. K. Das and B. Chanda, A fast algorithm for skew detection of document images using morphology, International Journal on Document Analysis and Recognition, vol. 4, pp , [20] B. V. Dhandra, V. S. Malemath, H. Mallikarjun, and R. Hegadi, Skew Detection in Binary Image Documents Based on Image Dilation and Region labeling Approach, in Pattern Recognition, ICPR th International Conference, 2006, pp [21] T. Nguyen Due, B. Vo Dai, M. Nguyen Thi Tu, and G. Nguyen Thuy, A Robust Document Skew Estimation Algorithm Using Mathematical Morphology, in Tools with Artificial Intelligence, ICTAI th IEEE International Conference, 2007, pp [22] A. Chaudhuri and S. Chaudhuri, Robust detection of skew in document images, Image Processing, IEEE Transactions on Image Processing, vol. 6, pp , [23] M. Chen and X. Ding, A robust skew detection algorithm for grayscale document image, in Document Analysis and Recognition, ICDAR '99. Proceedings of the Fifth International Conference on Document Analysis and Recognition, 1999, pp Author[s] brief Introduction Bishakha Jain is a student from Department of Computer Science and Engineering of Sikkim Manipal Institute of Technology,Sikkim,India. Mrinaljit Borah is a faculty member from Department of Masters in Computer Applications of Jorhat Engineering College,Jorhat,Assam

7 Corresponding Address- Bishakha Jain D/O Binod Kumar Jain, Bagaria Complex,Flat No 3/3, Thana Chariali,Dibrugarh, Assam Contact number: Mrinaljit Borah C/O Mr. Kuladhar Borah Midhakhat Tiniali P.O.-Teok PIN Jorhat,Assam Contact number: