Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.



Published on

Published in: Technology, Art & Photos
  • Be the first to comment

  • Be the first to like this


  1. 1. International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print), INTERNATIONAL JOURNAL OF COMPUTER ENGINEERING & ISSN 0976 - 6375(Online), Volume 4, Issue 5, September - October (2013), © IAEME TECHNOLOGY (IJCET) ISSN 0976 – 6367(Print) ISSN 0976 – 6375(Online) Volume 4, Issue 5, September – October (2013), pp. 285-291 © IAEME: Journal Impact Factor (2013): 6.1302 (Calculated by GISI) IJCET ©IAEME EFFECTIVE THRESHOLDING OF ANCIENT DEGRADED MANUSCRIPT FOLIO IMAGES Lalit Saxena Department of Computer Science, University of Mumbai, Mumbai, India ABSTRACT Thresholding is an essential procedure used in image segmentation and binarization applications. In this paper, segmentation methods applied on document images for separating the text from background presents pure binarization and filtering combined with image processing algorithms. This paper describes a contrast based thresholding method for old degraded manuscript images. It is an approach for degraded manuscript and document images by introducing an estimation of the threshold value. This technique effectively segments the texts from badly degraded document background. The method is suitable for segmentation of document images with complex and uneasy background having unreadable text. Proposed method performs segmentation using contrast estimating a threshold and exhaustively uses discrete gray level values. The proposed method broadly evaluated on more than 100 degraded manuscript images. The result shows the readable text in the improved images produced by the proposed method. Experiments confirm the effectiveness of the proposed method compared to standard thresholding methods. In research, the proposed method produced better results than standard thresholding methods for original manuscript images. Keywords: Degradation, Folio Images, Manuscripts, Segmentation, Thresholding. I. INTRODUCTION Thresholding is a rapid and precise procedure of segmentation of color and gray scale images. It is a sufficiently accurate and high processing speed segmentation approach to monochrome image. Over the years several thresholding techniques developed; but all of them aimed to have a generic approach to deal with different kinds of documents. There are two kinds of thresholding methods: global and local thresholding. Global thresholding algorithms use a discrete threshold for an image. These are intending to locate a discrete threshold to remove all pixels from the image background, while preserving all possible pixels in foreground. When there is a good 285
  2. 2. International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print), ISSN 0976 - 6375(Online), Volume 4, Issue 5, September - October (2013), © IAEME separation between background and foreground, global thresholding algorithms achieve high efficacy. For manuscript folio images, complex backgrounds and weak image foregrounds (many foreground pixels cover gray values close to those of some background pixels) toughens this procedure. In such cases, it is not possible to find a single threshold that separates the foreground from the image background. Thus, if this approach decides to binarize all the background pixels, then it also binarizes some of the foreground pixels. This results in broken texts that it not preserved the connectivity of strokes of the characters. On the other hand, local approaches estimate a separate threshold for background and foreground, on the basis of pixel neighborhoods. However, many document images have complex backgrounds that make the separation not so simple. Local or adaptive thresholding presents a better performance when treating documents with complex backgrounds. By contrast, adaptive thresholding methods fail in preserving stroke connectivity. II. STATE OF THE ART Despite of all the efforts made to restore degraded document images, recovery of the texts requires more efforts. The algorithm proposed by [1], initially binarizes the image using global method, and later invokes a comparable refinement method on each connected component to generate the absolute precise binary image. The document degradations happened because of shadows, non-uniform illumination, low contrast, large signal-dependent noise, smear and strain, handled by an approach developed by [2]. A nonparametric optimal threshold selection for image segmentation maximizing the separation of the gray level classes suggested by [3]. [4] proposed maximum entropy algorithm using probability distributions separating an image into objects and background on the basis of gray levels histogram. The method in [5] creates a threshold surface to find exact object boundaries for local threshold values using a gradient map of the image. A new concept about global thresholding proposed by [6] that separates an image into three regions, i.e., foreground, background, and a fuzzy area. Using multi-scale texture segmentation and spatial cohesion constraints to detect and extract text in images proposed by [7]. [8] introduced a method to binarize degraded and poor quality gray scale images having signal-dependent noise using logical adaptive thresholding. The method in [9], considered an image as a collection of subcomponents of text, background and picture for adaptive document image binarization. [6] proposed new thresholding technique and compared it against some existing algorithms. The experiment done using simple and complex images of postal envelopes by [7] used a multi-stage global thresholding approach followed by a local spatial thresholding. An image binarization method using [10] for low quality historical documents proposed by [11], calculates background surface by interpolating neighboring pixel intensities. A detailed survey on image thresholding methods with comparisons and categorization given by [12] and [13]. [14] introduced a local feature thresholding decompose algorithm, document sub regions using quad-tree decomposition and compared global and local thresholding techniques for degraded historical documents images. Considering that the text contains only 10% of the document image for binarization presented by [15]. [16] proposed a Kohonen adaptive neural network system for the binarization of normal and degraded documents for visualization and recognition of text characters. III. PROPOSED AND OTHER METHODS This paper describes an effective thresholding method for binarization of heavily degraded and poor quality gray scale manuscript images. This method can deal with complex signal-dependent noise and variable background intensity caused by non-uniform illumination, shadow, smear or smudge and very low contrast images. The outcome binary image has no observable loss of useful texts. The proposed method extracts the binary image adaptively from the degraded gray scale 286
  3. 3. International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print), ISSN 0976 - 6375(Online), Volume 4, Issue 5, September - October (2013), © IAEME document image with complex and inhomogeneous background. It estimates the value of threshold using contrast and gray values of the pixels. This method can threshold various poor quality gray scale document images without the need of any prior knowledge of the document. And it not requires any fine-tuning of parameters and also without taking into account characters geometric features. It keeps information accurately without over connected and broken strokes of the characters, and thus, has a wider range of applications. The block diagram of the proposed method is provided for precise understanding. Original Manuscripts Manuscripts images Gray scale conversion Threshold calculation Adaptive threshold Enhanced image Block diagram of the proposed method • Block 1: Original manuscripts: The original manuscripts collected in its native form without any external alterations. This is exceptional to possible deterioration removal. • Block 2: Manuscripts images: Camera with high resolution (this work used 14mega pixels) for clarity and format readable to latest computer. • Block 3: Gray scale conversion: Gray scale image of the color image produced to reduce the pixel processing complexity, since color image has three values; R, G, B. • Block 4: Threshold calculation: Threshold calculation involves gray scale values of the image pixels intensity, contrast used to understand the difference between foreground text and background. • Block 5: Adaptive thresholding: This adaptively thresholds gray scale image to binarized image. • Block 6: Enhanced image: The enhanced image with clear text, easy readability is the output of the proposed method. 287
  4. 4. International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print), ISSN 0976 - 6375(Online), Volume 4, Issue 5, September - October (2013), © IAEME 1. Proposed Method: It is obvious that a fixed value of the threshold estimation ܶሺ‫ݕ ,ݔ‬ሻ ൌ ܿ‫.ݐݏ݊݋‬ cannot yield satisfactory binarization results for images obtained under non-uniform illumination or with a non-uniform background. The proposed method calculates the local threshold value based in the mean value of the minimum and maximum intensities of pixels within a window [17]. If the window is centered at the pixel ሺ‫ݕ ,ݔ‬ሻ the threshold for ݂ሺ‫ݕ ,ݔ‬ሻ is defined by: ܶሺ‫ݕ ,ݔ‬ሻ ൌ ܶ௠௔௫ ൅ ܶ௠௜௡ 2 where ܶ௠௔௫ and ܶ௠௜௡ are the maximum and minimum intensity of the pixels in the window. This estimation of threshold value works correctly only when the contrast is sufficiently high. Also, the contrast is defined as ‫ ܥ‬ሺ‫ݕ ,ݔ‬ሻ ൌ ܶ௠௔௫ െ ܶ௠௜௡ [18]. It suggests that if the contrast is less than this value the pixels within the window will be assigned to background or foreground depending on the window. The proposed method is dependent on the size ܰ of the window defined by ܰ െ ܾ‫ ݕ‬െ ܰ. 2. Otsu's method: Suggested a discriminant analysis method for thresholding of the images. It is a formal pattern recognition procedure in which a criterion function used as a measure of statistical separation between classes. Calculations done for the two classes of intensity values (foreground and ଶ ଶ background) separated by an intensity threshold. The criterion function used here is ߪ஻௜ ⁄ߪ் for every ଶ ଶ intensity, ݅ ൌ 0, … , ‫ ܫ‬െ 1, where ߪ஻௜ is the between-class variance and ߪ் is the total variance. The intensity that maximizes this function said to be the optimal threshold. 3. Niblack's method: This method calculates the local mean and local standard deviation [10] of the image pixels in the window. It calculates the threshold value at pixel (x,y) by: ܶሺ‫ݕ ,ݔ‬ሻ ൌ ݉ሺ‫ݕ ,ݔ‬ሻ ൅ ݇. ‫ݏ‬ሺ‫ݕ ,ݔ‬ሻ where ݉ሺ‫ݕ ,ݔ‬ሻ and ‫ݏ‬ሺ‫ݕ ,ݔ‬ሻ are the mean and the standard deviation of a local area respectively. The size of the window must be large enough to suppress the noise in the image, but also small enough to preserve local details of the image. A window size 15 െ ܾ‫ ݕ‬െ 15 works efficiently. The value of k used to adjust the percentage of total pixels that belong to foreground object especially in the boundaries of the object. A value of ሾെ0.2ሿ produces objects separated well enough from background. 4. Sauvola's method: In this binarization method, the threshold ܶሺ‫ݕ ,ݔ‬ሻ calculated using the mean ݉ሺ‫ݕ ,ݔ‬ሻ and standard deviation ‫ݏ‬ሺ‫ݕ ,ݔ‬ሻ of the pixel intensities in a window centered around the pixel ሺ‫ݕ ,ݔ‬ሻ: ‫ݏ‬ሺ‫ݕ ,ݔ‬ሻ ܶሺ‫ݕ ,ݔ‬ሻ ൌ ݉ሺ‫ݕ ,ݔ‬ሻ ൅ ൤1 ൅ ݇ ൬ ൰൨ െ 1 ܴ where ܴ is the maximum value of the standard deviation (ܴ ൌ 128 for a gray scale document), and ݇ is a parameter which takes positive values in the range ሾ0.2 െ 0.5ሿ. The local mean ݉ሺ‫ݕ ,ݔ‬ሻ and standard deviation ‫ݏ‬ሺ‫ݕ ,ݔ‬ሻ adapt the value of the threshold according to the contrast in the local neighborhood of the pixel. When there is high contrast in some region of the image, ‫ݏ‬ሺ‫ݕ ,ݔ‬ሻ ൎ ܴ which results in ܶሺ‫ݕ ,ݔ‬ሻ ൎ ݉ሺ‫ݕ ,ݔ‬ሻ. 288
  5. 5. International Journal of Computer Engineering and Technology (IJCET), ISSN 0976 0976-6367(Print), ISSN 0976 - 6375(Online), Volume 4, Issue 5, September - October (2013), © IAEME IV. EXPERIMENTAL RESULTS While global thresholding algorithms are not good enough to treat complex backgrounds and local approaches do not preserve stroke connectivity (critical for digitization and preservation of manuscripts), the proposed approach successfully removes the background, yet keeping stroke connectivity untouched. Robust thresholding gives the opportunity of a correct separation of the drawn strokes or text from its background. E ective thresholding very easily separates the text Effective written on manuscripts from its background. This paper presents an e ective thresholding method effective for binarization of severely degraded and very low appearing gray scale manuscript images. The proposed method was tested with complex background images of old Indian manuscripts. The od method developed in this paper is to recover the textual information as much as possible. Literature presents implementation of several algorithms for thresholding on various types of document images. a b c d f e Figure 1: Thresholding results:a) original manuscript image, b) histogram, c) Sauvola method, d) Otsu method, e) Niblack method, f) Proposed method 289
  6. 6. International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print), ISSN 0976 - 6375(Online), Volume 4, Issue 5, September - October (2013), © IAEME V. CONCLUSIONS This paper described an algorithm that employs adaptive thresholding values to operate over the manuscript images. The purpose of this work on folio images is to threshold ancient manuscript images establishing an innovative method. Adjusting the threshold value according to the state of the image becomes reasonably in selecting gray scale values. It takes into account the improvement in the image quality as a whole and the increased readability of the texts. Results show that the proposed method performs better than other thresholding methods. Also, it is robust for document images in differences based on connectivity and background separation. Thus, no algorithm works better for all types of images but some work well than others for particular types of images. Hence, it suggests that for achieving improved performance, selection or combination of appropriate algorithm for the type of document image under investigation is necessary. The proposed method described a procedure that utilizes gray scale values of the pixels and image contrast. Many methods require intensive preprocessing steps to get proper data for working because document image segmentation techniques are still in infancy. The results show improved image quality of the manuscript images used in this work. However, this improvement is susceptible to noise, making the method unsuitable for heavy stained documents. ACKNOWLEDGEMENT The author wishes to thank Dr. Anjali Kade, Librarian, University of Mumbai, Mumbai for providing and allowing to take photographs of Original manuscripts folios used in this work. REFERENCES [1] I.B. Yosef. Input sensitive thresholding for ancient Hebrew manuscript. Pattern Recognition Letters, 26(8):1168–1173, June 2005. [2] B. Gatos, I. Pratikakis, and S.J. Perantonis. Adaptive degraded document image binarization. Pattern Recognition, 39(3):317–327, March 2006. [3] N. Otsu. A threshold selection method from gray–level histograms. IEEE Transaction on Systems, Man and Cybernetics, 9(1):62–66, January 1979. [4] J.N. Kapur, P.K. Sahoo, and A.K.C.Wong. A new method for gray–level picture thresholding using the entropy of the histogram. Computer Vision, Graphics, and Image Processing, 29(3):273–285, March 1985. [5] D.L. Yanowitz and A.M. Bruckstein. A new method for image segmentation. Computer Vision, Graphics, and Image Processing, 46(1):82–95, April 1989. [6] G. Leedham, C. Yan, K. Takru, J.H.N. Tan, and L. Mian. Comparison of some thresholding algorithms for text/background segmentation in difficult document images. In Proceedings of Seventh International Conference on Document Analysis and Recognition (ICDAR'03), pages 859–864. IEEE Computer Society, August 2003. [7] C.C. Wu, C.H. Chou, and F. Chang. A machine–learning approach for analyzing document layout structures with two reading orders. Pattern Recognition, 41(10), October 2008. [8] Y. Yang and H. Yan. An adaptive logical method for binarization of degraded document images. Pattern Recognition, 33(5):787–807, May 2000. [9] J. Sauvola and M. Pietikainen. Adaptive document image binarization. Pattern Recognition, 33(2):225–236, February 2000. [10] W. Niblack. An Introduction to Digital Image Processing, pages 115–116. Englewood Cliffs, N. J., Prentice Hall, 1986. 290
  7. 7. International Journal of Computer Engineering and Technology (IJCET), ISSN 0976-6367(Print), ISSN 0976 - 6375(Online), Volume 4, Issue 5, September - October (2013), © IAEME [11] S. Peratonis, B. Gatos, K. Ntzios, I. Pratikakis, I. Vrettaros, A. Drigas, C.E. mmanouilidis, A. Kesidis, and D. Kalomirakis. Digitisation processing and recognition of old Greek manuscripts (the d–scribe project). International Journal "Information Theories & Applications", 11(3):232–240, 2004. [12] M. Sezgin and B. Sankur. Survey over image thresholding techniques and quantitative performance evaluation. Journal of Electronic Imaging, 13(1):146–165, January 2004. [13] A. Vidyarthi and A. Kansal. A survey report on digital images segmentation algorithms. International journal of Computer Engineering & Technology (IJCET), 3(2):85–91, JulySeptember 2012. [14] Q. Chen, Q.S. Sun, P.A. Heng, and D.S. Xia. A double–threshold image binarization method based on edge detector. Pattern Recognition, 41(4):1254–1267, April 2008. [15] E. Kavallieratou and H. Antonopoulou. Advanced Concepts for Intelligent Vision Systems, volume LNCS 3708, chapter Cleaning and Enhancing Historical Document Images, pages 681–688. Springer-Verlag, Berlin Heidelberg, 2005. [16] E. Badekas and N. Papamarkos. Document binarization using Kohonen–som. IET Image Processing, 1(1):67–84, March 2007. [17] M.L. Feng and Y.P. Tan. Contrast adaptive binarization of low quality document images. IEICE Electronics Express, 1(16):501–506, November 2004. [18] R.C. Gonzalez and E.R. Woods. Digital Image Processing. Prentice Hall, Upper Saddle River, New Jersey, 2 edition, 2002. [19] Ratil Hasnat Ashique, Md Imrul Kayes, M T Hasan Amin and Badrun Naher Liya, “Speckle Noise Reduction from Medical Ultrasound Images using Wavelet Thresholding and Anisotropic Diffusion Method”, International Journal of Electronics and Communication Engineering & Technology (IJECET), Volume 4, Issue 4, 2013, pp. 283 - 290, ISSN Print: 0976- 6464, ISSN Online: 0976 –6472. [20] J.Rajarajan and Dr.G.Kalivarathan, “Influence of Local Segmentation in the Context of Digital Image Processing – A Feasibility Study”, International Journal of Computer Engineering & Technology (IJCET), Volume 3, Issue 3, 2012, pp. 340 - 347, ISSN Print: 0976 – 6367, ISSN Online: 0976 – 6375. [21] Mane Sameer S. and Dr. Gawade S.S., “Review on Vibration Analysis with Digital Image Processing”, International Journal of Advanced Research in Engineering & Technology (IJARET), Volume 4, Issue 3, 2013, pp. 62 - 67, ISSN Print: 0976-6480, ISSN Online: 09766499. [22] M. M. Kodabagi and S. R. Karjol, “Script Identification from Printed Document Images using Statistical Features”, International Journal of Computer Engineering & Technology (IJCET), Volume 4, Issue 2, 2013, pp. 607 - 622, ISSN Print: 0976 – 6367, ISSN Online: 0976 – 6375. 291